|
|
Home | Alpha Telephone | Domain Names | Web Hosting | Get Traffic | xrEvidence | xrSoccer United States Patent
Continuous casting method and apparatus for implementing same method A multi-articulated arm in a continuous casting machine which interacts with a melt surface condition within a mold, including melt surface abnormalities such as boiling and lack of powder, and a condition of slag beard formation by using image sensors, and also detects a condition of Deckel formation by using a load sensor through a sensor rod, thereby controlling a casting speed, a flow rate of gas blown into a pouring nozzle, a flow rate of a melt poured into the mold, scatter of powder to prevent the occurrence of the melt surface abnormalities. Additionally, said arm can remove a slag beard by a hitting oscillator and removing Deckels by the sensor rod. Those works can be performed following priority thereof by a multi-functional robot equipped with a multi-articulated arm having at its distal end an automatic tool changer mechanism.
Primary Examiner: Bradley; P. Austin Assistant Examiner: Pelto; Rex E. Attorney, Agent or Firm: This is a divisional of 700,118, filed on May 10, 1991 now U.S. Pat. No. 5,242,014, which is a continuation of Ser. No. 443,541, filed on Nov. 30, 1989, now abandoned. What is claimed is: 1. In a continuous casting apparatus including a mold and a multi-functional robot said robot further comprising: a multi-articulated support arm, said support arm being freely rotatable and vertically movable, and having at its distal end an automatic tool changer mechanism, and a drive controller for driving said support arm; at least one visual sensor disposed near a working region of said support arm for picking up images of the working region of said robot; a load sensor disposed near said automatic tool changer mechanism for detecting the load exerted on a tip end of a tool; a process detecting sensor for detecting operating conditions of said continuous casting process to be handled by said robot; and a synthetic judgement controller which includes: a) an input means for receiving input signals from said visual, load and process detecting sensors. b) a judging means for processing said input signals in accordance with a previously established sequence, and judging situations in working contents of said robot and the working region of said robot at a current time; c) a setting means for pre-setting priority of working positions and the working contents based on current operating conditions of the process to be handled by said robot and situations in the working region of said robot; d) a comparing and deciding means for comparing respective signals from said judging means and said setting section with each other, and scheduling and deciding the highest-priority or a most efficient working content dependent on the current situations; and e) a control means for issuing at least one of a tool change command signal to said automatic tool changer mechanism, a drive control signal to said support arm drive controller, and a control signal to control said process based on decision of the working contents made by said comparing and deciding means. 2. An apparatus according to claim 1 wherein: said working region is a region around said mold for continuous casting; said visual sensor includes at least one pair of image sensors for detecting a melt surface condition disposed at positions above said mold, which has a pouring nozzle being disposed in the center area of an upper opening portion of said mold, said image sensors being positioned on opposite side of said pouring nozzle; said load sensor includes a sensor for detecting said load exerted from a melt within said mold on a sensor rod which is supported by said support arm and dipped into the melt surface within said mold for continuous casting; said process detecting sensor includes a casting speed sensor, an in-mold melt surface level sensor, a melt flow rate sensor, and a flow rate sensor for gas blown into said pouring nozzle; said working contents includes adjustment of a casting speed, adjustment of a flow rate of the melt, blow of gas into said pouring nozzle, scatter of powder into said mold, removal of a slag beard, and remelting of a Deckel by pushing the Deckel into the melt; and said tool includes means for supplying powder into said mold, a slag beard removing unit, and a Deckel removing unit. 3. An apparatus according to claim 2 wherein; said powder supplying means has a powder reservoir for storing a preset amount of powder; said slag beard removing unit is provided at its tip end with an oscillation applying mechanism designed to crush or resonate a slag beard for peeling the slag beard from a mold wall; and said Deckel detecting unit has said sensor rod. 4. An apparatus according to claim 3 wherein said control means includes a casting speed controller, an in-mold melt surface level casting speed controller, a melt flow rate controller, and a blown gas flow rate controller. 5. An apparatus according to claim 4 wherein said in-mold melt surface lever controller includes thermocouples. 6. An apparatus according to claim 4 wherein said melt flow rate controller includes a sliding nozzle. 7. An apparatus according to claim 4 wherein said blown gas flow rate controller includes a part thereof fitted to said pouring nozzle extending from a tundish into said mold. 8. An apparatus according to claim 3 wherein said means for supplying powder includes: a powder supply unit for storing a preset amount of powder and capable of being opened at the bottom thereof or being turned; a feed unit for feeding powder to said powder supply unit; said powder supply unit being supported by said multi-articulated support arm at the distal end thereof; a drive unit for driving said support arm; and said powder supply unit and said drive unit being actuated in response to detected signals indicating lack of powder and a position of lack of powder. BACKGROUND OF THE INVENTION The present invention relates to a method and an apparatus for continuously casting metals while preventing the occurrence of various undesired abnormal conditions in the mold surface. Continuous casting is a method of drawing a molten metal being poured into a casting of constant cross-section, and allows to produce, for example, bars having a circular or rectangular cross-section, pipes and plate-like products. Such a casting method is used for producing castings of, for example, aluminum, copper alloys, cast iron, and steel. To describe continuous casting of steel, for example, a molten material (or melt) is poured from a ladle into a tundish and then from the tundish into a water-cooled mold. A casting emerging out of the mold is supported by a multiplicity of rolls when it is cooled by water. Pinch rolls are disposed under the supporting rolls for slightly pressing the casting to draw it into desired products. The drawn part is cut by a cutter upon reaching a certain length. On the other hand, a lubricant is supplied to inner surfaces of the mold to prevent the casting from sticking thereto. The tundish includes a nozzle to prevent entrainment of impurities into the melt when it enters from the tundish into the mold, so that the melt should be poured with the distal end portion of the nozzle kept always immersed in the melt within the mold. For example, continuous casting of steel, in particular, has various difficult problems occur which have impeded complete automation of work, especially in the process related to pouring of a melt into a mold. Such problems will be explained in more detail below. A. Problem of Melt Surface Abnormalities The associated system including a continuous casting mold (hereinafter referred to simply as mold) comprises 1) a mold, 2) a pouring nozzle, including a lower nozzle, an upper nozzle and a dipped nozzle, which is disposed at the center area of the mold and mounted to the bottom of a tundish, 3) a sliding nozzle (hereinafter referred to simply as SN), or melt (molten steel) flow rate controller utilizing, for example, a stopper, 4) a flow rate controller of gas blown into the pouring nozzle (hereinafter referred to simply as blown gas) for capturing and surfacing inclusions or deoxidation products present in the melt within the mold, and preventing the pouring nozzle from being clogged by the inclusions or deoxidation products, 5) a melt surface level controller, and the like. The blown gas is discharged into the melt surface, which means the surface of contents including molten steel, molten flux and non-molten flux, within the mold, while mold powder or mold flux is charged (or scattered) into the mold for the purposes of heat-keeping, thermal insulation and anti-oxidation of the melt within the mold, capture of deoxidation products or inclusions, and lubrication between a solidified shell and the mold, for instance. The powder or flux is melted upon contacting the melt (or molten steel) to form a molten layer, followed by flowing into gaps between the mold and the solidified shell to effect the various above explained functions. Further, vertical oscillations are applied to the mold to ensure smooth drawing of a casting, while the melt surface within the mold is constantly fluctuated or moved warily with the melt discharged through the pouring nozzle. The melt discharged through the pouring nozzle under such conditions is cooled by the mold and others to start solidifying while forming a meniscus at the top surface of the melt in conformity with the mold. Thus, in the associated apparatus components and mechanism including a mold, the above mentioned factors are present in a intermixed and complicated state, which factors are delicately balanced in a stable condition of the entire casting system. In other words, the associated apparatus components and mechanisms including a mold contain very complex fluctuating factors, and are always in such an unstable slate that the process is varied largely just by changing some operated variable to a small extent. For example, boiling may occur just by slightly changing a flow rate of the gas blown into the pouring nozzle. The casting system is therefore very sensitive to operational fluctuating factors such as fluctuations in an amount of the melt in the tundish or casting speed, and clogging of the pouring nozzle. Once the balance is lost, there may immediately occur abnormalities, such as level fluctuations, boiling, biased flow and lack of powder, on the melt surface within the mold, such as, so called in this specification melt surface abnormalities. The boiling is a phenomenon that inert gas fed into the tundish, the upper nozzle or the dipped nozzle and then blown into the melt within the mold for the purposes of, e.g., preventing the dipped nozzle from being clogged, is boosted to such an extent as to hinder the melt from flowing from the tundish into the dipped nozzle, and the boosted gas is then blown out at a burst from discharge holes (for example, two) of the dipped nozzle. In the event of this phenomenon, the flames are flared up temporarily from the melt surface and the melt level is lowered. These melt surface abnormalities may directly cause problems including a break-out (hereinafter referred to simply as BO) which is most desirable in the continuous casting operation, and may also directly lead to defect of the surface quality attributable to capture of the powder onto the solidified shell. Thus, keeping the associated apparatus components and mechanisms including a mold stable at all times is the most important point not only in preventing the occurrence of BO, but also in carrying out continuous casting operation steadily while ensuring the quality of castings at the surface and thereabout. Conventionally, therefore, the melt surface conditions are being monitored within the mold and many detecting means have also been proposed for monitoring. For example, Japanese Patent Unexamined Publication No. 60-49846 discloses a method in which an infrared camera for scanning the melt surface is installed above a mold to measure a temperature distribution over the melt surface thereby to detect a thickness of and its distribution over a powder layer. Japanese Patent Unexamined Publication No. 54-71723 discloses a method of selecting two or more wavelength bands of light radiated from the melt surface within a mold and measuring a temperature of the melt surface from the ratio of light energy level between those wavelength bands, thereby to detect a condition of lack of powder. Japanese Patent Unexamined Publication No. 59-229267 discloses a method of arranging the tip end of an optical fiber to sense (or scan) the melt surface and measuring an electric signal in accordance with an amount of light emitted from the melt surface, thereby to detect a condition of lack of powder. However, those conventional methods do not allow precise detecting of the lack of the powder due to the fact that flames or the like caused upon the gas blown into a pouring nozzle flowing into a mold may erroneously be detected as lack of the powder, or because the melt surface is constantly moved wavily with, for example, oscillations of a mold, the detecting means may be affected by large fluctuations of a temperature condition at the melt surface, or light disturbance caused upon the molten part of the powder or by flames moving warily to appear or disappear with time. Another problem is in that since the temperature measuring device, the infrared camera or the optical fiber, requires a scanning time and a relatively long time for detecting a scattered condition of the powder throughout the melt surface within a mold, the melt surface condition may be changed during the detection process, making it impossible to take a necessary action with the proper timing. Furthermore, in the event of abrupt melt surface abnormalities such as biased flow and boiling mentioned above, the conventional methods can only perform the process of detecting lack of the powder and hence could not detect such abrupt melt surface abnormalities. Although there have also been proposed techniques of providing thermocouples embedded in a mold wall or mounting a magnetic sensor, an infrared camera or the like above a mold, thereby to a melt surface level, these detecting devices are intended to detect only a melt surface level for controlling constant level, and hence cannot directly detect the melt surface abnormalities such as biased flow, boiling and lack of powder. As described above, any of the conventional detecting methods had only a single function, and was difficult to quickly and precisely detect a condition of the melt surface in a stable manner. Thus, in actual circumstances various units and devices are arranged intricately in a narrow space near a mold and under high-temperature, dust-full environments, and there is no practical detecting device capable of detecting a condition of the melt surface with high reliability under such unfavorable conditions. To date, therefore, it has been customary for an expert skilled operator to monitor a condition of the melt surface and judge the occurrence of melt surface abnormalities based on his past experience and perception. As mentioned above, if the associated apparatus components and mechanisms including a mold are out of balance to cause melt surface fluctuations or the melt surface abnormalities such as biased flow, boiling and lack of powder, it has been usual in the past that the operator monitoring the melt surface visually detects such an abnormal condition of the melt surface and immediately takes a proper action based on the detected result. Because the action to balance and stabilize the associated apparatus components and mechanisms including a mold requires quickness and precision, the conventional methods have had to rely on an expert skilled operator, and this has been a big obstacle in an attempt of saving labor expenses. Notwithstanding such relying on expert skilled operators, there exists a large difference between the operators, and also frequently occurs a delay in detection or erroneous detection, which may cause variations in the quality of castings. In an extreme case, a delay in the proper action may lead to BO. B. Problems Related to Blow of Inert Gas In continuous casting of steel, it is common to once store a melt (molten steel) fed through a ladle in a tundish and then pour the melt from the tundish into a mold through a pouring nozzle. In this case, the melt includes impurities such as deoxidation products, for example, Al.sub.2 O.sub.3, or powder, slag and sulfides (hereinafter referred to collectively to "inclusions"). If any inclusions are captured and left in a casting, this would give rise to various drawbacks for example, causing internal defects called slag intrusion or surface flaws. Further, Al.sub.2 O.sub.3 and the like among the inclusions tend to adhere on the inner surface of the nozzle while passing therethrough and eventually accumulate to such an extent as to clogging the pouring nozzle, whereby the stable operation is hindered n many cases. It has been known to provide means for effectively separating inclusions from the melt and moving them to the melt surface. For example, Japanese Patent Examined Publication No. 49-28569 discloses a technique of blowing inert gas such as argon or nitrogen gas into a flow of the melt while being poured into a mold, thereby effectively move inclusions to the melt surface. This technique has been adopted widely in recent years. Also, Japanese Utility Model Unexamined Publication No. 62-142463 discloses a device for calculating an appropriate data to control a flow rate of gas based on, for example, the flow rate of the melt determined from the head size of the melt, the width and thickness of castings, and the casting speed. However, in either prior art where the flow rate of blown gas is visually adjusted by an operator, or where it is automatically controlled using the control device, it has been usual to measure a flow rate of the gas flowing through a pipe for being blown into the melt, to control a value of the gas flow rate. However, the reading on a flow gauge installed in the pipe does not always correspond to the flow rate of the gas actually flowing into the melt, because a portion of the gas may leak during the process before reaching the melt, for example, at refractories employed to form a flow passage, or the pressure loss in the flow passage may be changed. It may also happen that the gas blown into the melt flows into a mold along the wall surface of the pouring nozzle and then escapes above the tundish in vain without effecting the specific function. The ratio of an amount of the gas having been blown into the melt but leaked to an amount of the effective gas having been blown into the melt and reached the mold through the pouring nozzle changes dependent on operating conditions. Therefore, in the conventional techniques it has been very difficult to properly control an amount of the effective gas. If the amount of the effective gas is not properly controlled and the amount of the blown gas exceeds a required level, the flow rate of the gas would become unstable and the melt surface within the mold would be largely disturbed, thereby eventually causing a phenomenon in which the melt would not flow into the pouring nozzle, for example boiling. On the contrary, if the amount of the blown gas is reduced to be too small, the stable operation becomes hard to continue because of clogging of the pouring nozzle and other troubles. Thus, using the conventional techniques may cause boiling and clogging of the pouring nozzle, due to a difference between an amount of the gas flowing through the pipe and an mount of the effective gas as resulted from measuring a flow rate of the gas in the blowing pipe, or due to difficulties in quantitatively adjusting a flow rate of the gas in a stable manner as experienced even when the operator monitors a condition of the melt surface within the mold and adjusts the flow rate of the gas based on the monitored result. C. Problem of Slag Beard Formation As well known, powder is supplied (or scattered) over the melt surface within a mold in continuous casting for the purposes of heat-keeping and air shutdown of the melt, capture of nonmetallic inclusions, and lubrication between a solidified shell and the mold, for instance. The powder is melted upon being subjected to heat of the melt, and fluidized to move from the melt surface along the wall surface of the mold. While moving along the wall surface of the mold, the molten powder is cooled by the mold, but heated by the melt. Meanwhile, vertical oscillations are applied to the mold a solidified shell from sticking to the mold surface. Therefore, the powder, once melted, is caused to become solid again and adhere onto the wall surface of the mold in a region just above the solidifying interface of the melt. Such adhesion of the powder gradually increases an amount of the deposited powder with progress of casting, so that the raw powder projects out of the wall surface of the mold in the form of a terrace, thereby forming a slag beard. A slag beard is responsible for not only a detrimental effect on the quality of castings, but also more serious trouble in the continuous casting operations. For example, if the melt surface is abruptly raised up for some reason in the presence of a slag beard, the slag beard would be captured by a solidified shell and cause a serious defect in the surface quality of castings. In the worst case, the slag beard captured by the solidified shell just below the mold is forced to melt again upon being subjected to heat of the melt or the solidified shell, which may result in a BO (or break-out), which is more detrimental in the continuous casting operation. In view of the above, it is most preferable to carry out the operation in such a manner as to avoid the occurrence of a slag beard. However, because of complicated thermal conditions produced just above the solidifying interface, it is very difficult to suppress the occurrence of a slag beard itself. In the past, however, there have been no methods of detecting a condition of the slag beard formation with good precision. At most, only a method of measuring a temperature of the mold, estimating adhesion of the powder onto the wall surface of the mold based on temperature changes, and predicting formation of a slag beard has been proposed in some cases. Thus, in practice the detecting method using such indirect measuring means is poor in its precision and cannot be adopted in actual apparatus. For the reason, with a technique having been generally adopted to date, an operator directly visually monitors a condition in the mold to detect a condition of the slag beard formation. If a slag beard is detected being formed, the operator pushes or pokes the slag beard by a proper stick to peel it and remove from the wall surface of the mold. This imposes very great physical and mental burden on the operator, and also gives rise to a safety problem. In addition, it has frequently happened that the operator erroneously damages a solidified shell and hence causes a defect in the surface quality of castings. In an extreme case, a BO may be caused. Avoiding such problems requires expert skilled operators, which has been a big obstacle in an attempt of achieving the automated casting operation in continuous casting. As one of means for removing a slag beard, for example, Japanese Patent Unexamined Publication No. 61-144249 proposes a method of irradiating heat flux such as an infrared ray or laser beam to the slag beard for melting and removing it. However, even such a method has to rely on human efforts to detect the formation of a slag beard, and also requires a great deal of additional energy to melt the slag beard. The case of using a laser beam results in a significant increase in the equipment cost. D. Problem of Deckel Formation A Deckel (German) is in the form of so-called "leather cover" which results from solidification of a surface layer of the molten steel or steel bath, when actual and latent heat of the steel bath are removed from the bath surface in a mold for continuous casting. If Deckels are formed, nonmetallic inclusions present in the bath, which should be properly captured by mold powder, would instead be captured by the Deckels to remain in a casting, thereby detrimentally affecting quality of the casting. Also, if a Deckel is formed all over the surface of the steel bath, this would cause more serious operational problems such as a BO and breakage of a dipped nozzle. For the reason, it is required to prevent the formation of a Deckel during continuous casting. However, when the molten steel poured into the mold has a temperature as low as that of the liquid line, or when the casting speed is low, Deckels may be formed because of the reduced temperature of the bath within the mold. Meanwhile, mold powder is supplied (or scattered) over the bath surface within the mold for the purposes of heat-keeping and thermal insulation of the bath, capture of non-metallic inclusions, and lubrication between a solidified shell and the mold, for instance. Therefore, a condition of the Deckels formation cannot be judged visually. Under such environmental situations, there have been neither devices for detecting the formation of Deckels nor devices for removing the formed Deckels in the past. Thus, it has conventionally been customary for an operator to thrust a proper stick of steel into the melt within the mold for sensing the formation of Deckels based on a feeling perceived by his hands. If Deckels are judged to be formed, the operator pushes the Deckels into the bath using that stick so that it is melted once again. The foregoing conventional method in which the operator detects and remelts Deckels is inefficient because of manual operation, and the operator is subjected directly to the melt or bath and hence safety is not ensured. In addition, unskilled operators may disturb the solidifying interface between the mold and the solidified shell, thereby causing surface flaws on castings. To improve reliability, therefore, the operation have had to rely on a few skilled operators. On the other hand, although technical development of achieving the automated casting operation in continuous casting has been actively performed in these years, the operation of detecting the formation of Deckels cannot be automated by utilizing optical detection with image processing, for example, because it is impossible to visually detect the Deckels as mentioned above, and hence has been a big obstacle in an attempt of realizing automated casting operation. E. Problems Related to Automation of Operations It can be considered to carry out by a robot operations associated with pouring of a molten steel into a mold. To this end, the above described problems A-D have to be solved. Solving of those problems requires to monitor and recognize process conditions changing from time to time as mentioned in A-D, determine necessary actions based on self-judgment dependent on the recognized conditions, and then implement a plurality of selected actions. However performance and operation control of robots have faced problems as follow. As to the contents of operation and control relating to utilization of industrial robots, the robots from the simplest ones to those capable of repetitively executing a series of operations, as represented by the teaching and playback system, are now within a practicable range and mainly used. In other words, present industrial robots are applied to only such operations as palletizing, painting, feeding, welding and simple assembling which can be executed through simple control of working positions, and hence can be regarded as robots having not perceptual and judgment abilities, but power of memory alone, because they are controlled only by performing reproductive operations of working timings, working contents and position data which have been taught beforehand. Changes of environments are hardly assumed in use of this type robots, and their operations are carried out essentially under the assumption that the objects to be handled are always in certain positions. Furthermore, being basically capable of repetitive operations as mentioned above, industrial use of those robots is restricted to relatively simple works. For enlarging an application range of industrial robots, therefore, it has been attempted to use various sensors and control operation of the robot based on information detected by the sensors. More specifically, the attempted method is to detect a condition of operating environments around the robot, and control operation of the robot while determining working timings, working contents and working positions of the robot based on the detected environmental condition. Sensors employed in robots include mainly touch, load (force), visual (image) and audible sensors. As to visual sensors among them, for example, some systems have already been developed to a practicable level through combination of a TV camera and an image processing unit. The contents of visual control are carried out by sequentially recognizing an environmental condition by a visual sensor to determine the process of work, and then issuing a command to a control system. Since the information obtained by the visual sensor ms related to only positions and attitudes, the control to be effected can be achieved through just position control. Stated otherwise, even when a visual function is added to a robot, the control system of robot operation can itself be used as it was. With widespread use of visual sensors using CCDs (see page 37), therefore, operation control of robots utilizing a visual sense is going ahead in an application level of other types sensors. As to load (force) and touch sensors, there exists simple load control of converting only a setting value for a position command based on a load or touch detected signal, and then controlling a position in accordance with the setting value. Although even such simple control can sufficiently execute some works, its application range is restricted. In most cases of using load sensors, it is demanded to continuously control the load so that the vector magnitude in the form of load is controlled to De constant or to provide a predetermined pattern. This type of control is more difficult than the simple load control, and hence problematic under the present state of art. As explained above, operation control of robots utilizing sensors ms now practiced mainly by making use of visual sensors in such a manner as to sequentially control positions, etc. based on detected signals in a large part of applications. Meanwhile, from the standpoint of multi-functions of robots, there has been used an automatic tool changer (hereinafter referred to as ATC) which has a function capable of easily attaching and detaching tools to the tip end of the robot. But, such multifunctional robots are comparable in level of operation and its control to that as mentioned above, and basically applied to simple works as represented by the teaching and playback system. Operation control of multi-functional robots using sensors (mainly visual sensors) to perform a plurality of works is also comparable in level to that as mentioned above, and limited in application to the field of sequential control. In short, the technology of utilizing robots remains at the level as mentioned above in the present stage of art. Notwithstanding intensive study for next stage robots, only few have reached a satisfactory level. Stated otherwise, changes of operating environments around robots are not taken into account, and much attention is not paid to the process of recognizing a varying condition of environments and determining the working program, for example, based on the recognized condition. Thus, there are mainly in use robots having no perceptual and judgment abilities necessary to program their own operation control depending on changes of operating environments. At the present stage of art, efficiency and reliability of works executed by robots are regarded as more important than perceptual and judgment abilities thereof. From the foregoing reasons, conventional robots have faced difficulties in their application to the working process where operating environments to be adopted are changed from time to time, and a plurality of working contents and/or working positions have to be determined by the robots themselves dependent on varying conditions of the operating environments detected by sensors, thereby to perform necessary works. Accordingly, that working process have had to be carried out in such situations through direct operations by plural skilled operators using jigs or the like, or indirect operations by operators using a plurality of manipulators, or introduction of multiple robots as mentioned above for relatively simple works. That working process therefore requires a great deal of equipment and man power costs, and also reduces working efficiency. In addition, a safety problem has been experienced because the operators have had to work in the midst of manipulators and robots. The present invention has been accomplished in view of the foregoing problems as stated in A-E. BRIEF SUMMARY OF THE INVENTION One object of the present invention is to provide a method and an apparatus for precisely and quickly detecting melt surface abnormalities in a mold, and effectively preventing their formation based on the detected result, to maintain stable continuous casting operation. The above-mentioned melt surface means the surface of contents, including molten steel, molten flux (or slag), non-molten flux and Deckels, within the mold. Another object of the present invention is to provide a method which can always ensure a required amount of effective gas precisely following fluctuations in operating conditions or fluctuations in an amount of gas leak through a nozzle and a pipe, if occur, thereby to produce castings excellent in quality without causing problems such as boiling and nozzle clogging. Still another object of the present invention is to provide a method of detecting a slag beard with good accuracy during operation, and an apparatus for effectively removing the slag beard based on the detected result, thereby making it possible to achieve the automated casting operation while maintaining the stable continuous casting operation. Still another object of the present invention is to detect and remove Deckels formed in the mold during continuous casting. Still another object of the present invention is to provide an automated control device or a robot suitable for fulfilling the above mentioned objects. More specifically, this object is to implement efficient multi-functional works, in such working process where operating environments to be adapted are changed from time to time, and hence where a plurality of working contents have to be determined and carried out dependent on varying conditions of the operating environments detected by sensors, through the procedures of using a robot to carry out plural works, controlling the preset priority of works dependent on information applied from the sensors and working conditions of the robot, and then scheduling and deciding the working program so as to perform necessary works with maximum efficiency. To achieve the above objects, in accordance with a first feature of the present invention, there is provided a method of detecting a melt surface abnormal condition in a mold for continuous casting, comprising the steps of installing one or more pairs of image sensors for detecting a melt surface condition at positions above the mold, which has a pouring nozzle being disposed in the center area of an upper opening portion of the mold, the image sensors being opposed to one another with the pouring nozzle located therebetween; converting input images of a melt surface condition, which are periodically applied from the image sensors to binary images using a reference threshold level for discriminating the image regions into light portions and dark portions; determining an area ratio Rn of the light portions to the entire melt surface region in the viewing fields of the image sensors; determining a change rate Dn of the light portion area ratio Rn per unit time; and detecting an abnormal condition of the melt surface based on preset correlation between the light portion area ratio Rn and/or the change rate Dn and a melt surface abnormal condition. According to one embodiment of the above method, the change rates Dn for the respective melt surface regions on the opposite sides of the pouring nozzle are compared with preset reference values of the change rate Dn for the respective melt surface regions; the occurrence of boiling is determined when both of the change rates Dn for the respective melt surface regions on the opposite sides of the pouring nozzle exceed the reference values; and the occurrence is determined when only either one of the change rates Dn exceeds the corresponding reference value, thereby detecting an abnormal condition of the melt surface. According to another embodiment of the above method, each of the respective melt surface regions in the viewing fields of the image sensors is subdivided into plural preset zones and a light portion area ratio Rnn and a change rate Dnn of the light portion area ratio Rnn per unit time for each of the divided zones are determined; the light portion area ratio Rnn and the change rate Dnn are compared with respective preset reference values for each of the divided zones; and a lack condition of powder is determined when the light portion area ration Rnn exceeds the threshold level, but the change rate Dnn does not exceed the reference value, thereby detecting lack of powder as one of melt surface abnormalities and positions where the powder is short. According to another feature of the present invention, there is provided a method of preventing a melt surface abnormality, which performs any one or more of casting speed control, flow rate control of gas blown into the pouring nozzle, flow rate control of melt poured into the mold, and powder scatter control dependent on a melt surface condition detected by the above method. According to still another feature of the present invention, there is provided, as a device for eliminating a lack of powder condition, a melt surface abnormality preventing device comprising a powder supply unit for storing a preset amount of powder and capable of being opened at the bottom thereof or turned; a feed unit for feeding powder to the powder supply unit; a multi-articulated support arm supporting at its distal end the powder supply unit and capable of rotating and moving up and down; a drive unit for driving the support arm; and a powder supply controller for controllably driving the powder supply unit and the drive unit in response to detected signals indicating lack of powder and it's position. According to still another feature of the present invention, there is provided a continuous casting method of pouring a melt stored in a tundish into a mold for continuous casting through a pouring nozzle, while blowing inert gas to the melt, comprising the steps of installing one or more image sensors above the mold for picking up images of the melt surface within the mold; processing image signals detected by the image sensors during continuous casting to detect the number of bubbles floating up to the melt surface and/or the size of flames flared up on the melt surface; comparing each of the detected values with allowable limit values previously determined from correlation between an amount of blown gas and the number of bubbles or the size of flames in the event of boiling or nozzle clogging; and immediately controlling the amount of blown gas when the number of bubbles and/or the size of flames exceeds the corresponding allowable limit values, thereby ensuring a proper mount of blown gas. According to still another feature of the present invention, there is provided a slag beard formation detecting method in continuous casting comprising the steps of disposing one or more pairs of image sensors for detecting a melt surface condition at positions above a mold, which has a pouring nozzle being disposed in the center area of an upper opening portion of the mold, the image sensors being opposed to one another with the pouring nozzle located therebetween; determining distribution of brightness along at least one check line intersecting the wall surface of the mold based on image input signals of a melt surface condition periodically applied from the image sensors; determining a distance between two positions where the distribution of brightness shows a maximum value and a minimum value; and estimating a condition of slag beard formation from the determined distance. According to still another feature of the present invention, there is provided a slag beard removing device comprising a multi-articulated arm capable of rotating and moving up and down and having at the distal end a hitting oscillator with an oscillation applying mechanism; a drive unit for driving the arm; and a controller which, by using one or more pairs of image sensors for detecting a melt surface condition installed at positions above a mold, which has a pouring nozzle being disposed in the center area of an upper opening portion of the mold, the image sensors being opposed to one another with the pouring nozzle located therebetween, performs the steps of determining distribution of brightness along at least one check line intersecting the wall surface of the mold based on image input signals of a melt surface condition applied from the image sensors from time to time; determining a distance between two positions where the distribution of brightness shows a maximum value and a minimum value; and issuing a command signal to start driving the hitting oscillator and a command signal to determine a position where the hitting oscillator is to be operated, based on a condition of slag beard formation estimated from the determined distance. According to still another feature of the present invention, there is provided a Deckel formation detecting device comprising a sensor rod dipped into the melt surface within a mold for continuous casting; a multi-articulated support arm supporting at the distal end the sensor rod and capable of rotating and moving up and down; a drive unit for driving the support arm; and a load sensor for detecting the load exerted on the sensor rod, thereby detecting a condition of Deckels formation from the load exerted on the sensor rod dipped into the melt surface within a mold. Also, there is provided Deckel removing device which further comprises a comparison operation controller for comparing a detected value of the load sensor with a preset Deckel formation reference value, and issuing a command to drive the drive unit when the detected value exceeds the reference value. Finally, according to still another feature of the present invention, there is provided a multi-functional robot which can be used as a universal industrial robot and is particularly suitable for use in continuous casting operation. This robot comprises a multi-articulated support arm having at the distal end an automatic tool changer mechanism (hereinafter referred to simply as ATC) and capable of freely moving to rotate and shift up and down, and a drive controller for driving the support arm. Specific features of the robot are as follows: i) an input section for receiving signals from, at least, one or more visual sensors installed near a working region and/or at the distal end of the support arm for picking up images of the working region of the robot, a load sensor installed near the automatic tool changer mechanism for detecting the load exerted on a tip end of a tool, and a process detecting sensor for detecting operating situations of a process to be handled by the robot; ii) a judging section for processing the input signals in accordance with the sequence previously established, and judging situations in the working contents of the robot and the working regions of the robot at the current time; iii) a setting section for previously setting priority of the working positions and the working contents based on the current operating conditions of the process to be handled by the robot and situations in the working region of the robot; iv) a comparing and deciding section for comparing respective signals from the judging section and the setting section to each other, and scheduling and deciding the highest-priority or most efficient working content dependent on the current situation; and v) a control section for issuing at least one of a tool change command signal to the automatic tool changer mechanism, a drive control signal to the support arm drive controller, and a control signal to the process based on decision of the working contents made by the comparing and deciding section. Other objects and features of the present invention will become apparent from the following description of respective embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS FIGS. 1 to 24 are a first group of drawings related to an embodiment I, FIG. 1 is a conceptual overall view showing one embodiment of the present invention explaining a continuous casting apparatus. FIGS. 2 and 3 are images picked up by a pair or cameras installed above a mold. FIGS. 4 and 5 are other examples of picked-up images. FIGS. 6 and 7 are images resulting from converting the images of FIGS. 4 and 5 to a binary representation. FIG. 8 is a graph showing time-dependent changes in the area of light portions. FIG. 9 is a graph result from eliminating disturbances in the graph of FIG. 8. FIG. 10 is a graph showing time-dependent changes in the light portion area ratio during the continuous casting operation. FIG. 11 is a graph showing change rates per unit time in the light portion area ratio during the continuous casting operation. FIGS. 12 and 13 are examples of dividing the respective images into multiple zones. FIGS. 14 and 15 are graphs, similar to FIG. 10, showing time-dependent changes in the light portion area ratios for the respective divided zones. FIGS. 16 and 17 are graphs, similar to FIG. 11, showing change rates per unit time in the light portion area ratios for the respective divided zones. FIG. 18 is a conceptual view showing one example of a powder supply unit used in the present invention. FIG. 19 is a view showing one example of a powder supply jig used in the present invention. FIG. 20 is a view showing one example of a powder supply reservoir used in the present invention with dimensions thereof; FIGS. 21, 22, 23 and 24 are views showing other example of the powder supply reservoir. FIGS. 25 to 29 are a second group of drawings related to an embodiment II; FIG. 25 is a view showing an embodiment of implementing the present invention in a general continuous casting equipment. FIGS. 26 and 27 are illustrations showing binary images taken by a pair of image sensing devices in which; FIG. 26 shows a state where bubbles are produced and FIG. 27 shows a state where flames are flared up. FIG. 28 is a graph showing one example of an examined result of the relationship between the number of bubbles and a flow rate of blown gas, FIG. 29 is a graph showing one example of an examined result of the relationship between the area of flames and a flow rate of blown gas. [FIGS. 30 to 35 are a third group of drawings related to an embodiment III] FIG. 30 is a sectional view and a graph showing a partial section illustrative of a melt surface condition in a mold, and brightness distribution obtained by picking up an image of the region near the mold corresponding to the melt surface condition shown in that partial section, respectively. FIG. 31 is a block diagram showing one example of a slag beard formation detecting device. FIG. 32 is a plan view showing one setting example of check lines used for detecting a slag beard thickness. FIG. 33 is a graph showing one example of detecting a condition of slag beard formation in thickness. FIG. 34 is a block diagram showing one example of a slag beard detecting and removing device. FIG. 35 is a graph showing time-series changes in the slag beard thickness resulted from operation of a slag beard detecting and removing device. FIGS. 36 to 40 are a fourth group of drawings related to an embodiment IV; FIG. 36 is a block diagram showing one example of a Deckel formation detecting device used in the present invention. FIG. 37 is an enlarged sectional view of a mold in which the Deckel is formed. FIG. 38 is a graph showing one example of an examined result of load changes due to formation of the Deckel. FIG. 39 is a block diagram showing an example of a Deckel removing device of the present invention. FIG. 40 is a graph showing one example of load data detected by a load sensor in a practical example. [FIGS. 41 to 48 are a fifth group of drawings related to an embodiment V] FIG. 41 is a block diagram showing an overall concept of a multi-functional robot of the present invention. FIG. 42 is a block diagram showing one example of the multi-functional robot of the present invention. FIG. 43 is a plan view of images picked up by a pair of cameras installed above a mold. FIG. 44 is a plan view of the images of FIG. 43 converted to a binary representation and divided into multiple zones. FIG. 45 is a front view for explaining the principles of measuring a slag beard. FIG. 46 is a plan view showing working positions for measuring and removing the slag beard in the mold. FIG. 47 is a graph showing one example of load data detected by a load sensor when a sensor rod is dipped into the melt surface. FIG. 48 is a plan view showing working positions for measuring and removing Deckels in the mold. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS(S) The following description relates to an embodiment solving the problem of melt surface abnormalities (see the above section A). FIG. 1 is a view showing one example of an overall concept of the present invention. In FIG. 1, designated by reference numeral 1 is a mold, 4 is a casting being cast into products, 5 is a melt (molten steel), 6 is a solidified shell produced from the melt 5 upon cooling thereof, and 7 is powder supplied (or scattered) in the mold 1 to mainly form a non-molten powder layer 8 and a molten powder layer 9, also in FIG. 1, numeral 12 is a pouring nozzle being of the SN (sliding nozzle) type in this embodiment, and 13 and 14 are flow rate controllers for the melt. Designated as 15 and 16 are flow rate controllers for gas blown into the pouring nozzle 17 and 18 are controllers for a casting speed of the casting 4, and 19 is a tundish. Further, designated as 22 is a melt surface abnormal condition detecting device and 41 is a melt surface abnormality preventing device. The melt surface abnormal condition detecting device 22 comprises a pair of image sensors 20 installed above the melt surface within the mold at positions opposite to each other with the pouring nozzle 12 located therebetween; an arithmetic processing unit 21 for processing input images of a melt surface condition applied from the image sensors 20 from time to time, as described later, and detecting a melt surface abnormal condition such as biased flow, boiling or lack of powder; and a controller 23 for instructing, based on detection of a melt surface abnormal condition, various actions (described later) to be carried out for stabilizing the melt surface abnormality, and issuing control commands to the respective devices and units. In this embodiment, although the image sensors 20 are installed in one pair at positions opposite to each other with the pouring nozzle 12 located therebetween, two or more pairs may be provided when the entire region of the melt surface within the mold cannot be covered by only one pair of image sensors in their viewing fields. The melt surface abnormality preventing device 41 comprises a powder supply unit 30, a multi-articulated support arm 25 supporting the powder supply unit 30 at the distal end thereof, a feed unit (or feeder) 35 for feeding powder to the powder supply unit 30, a drive unit 28 for driving the support arm 25, and a powder supply (or scatter) controller 24 for driving and controlling the powder supply unit 30 and the support arm 25 based on detected signals relating to the occurrence of lack of powder and the lack of powder position. The powder supply unit 30 has a cup-like reservoir 31 for storing a preset amount of the powder 7. The reservoir 31 is, as described later, arranged to be open at its bottom or turnable. The support arm 25 comprises a plurality of arms (two arms 26, 27 in this embodiments) interconnected through arm drivers 28a-28c each having a rotatable shaft built therein, so that the arms can be rotated about the respective rotatable shafts in an articulated manner. In addition, the support arm 25 is provided at the distal end thereof with a lift driver 28d and a lift frame 29 to provide a vertical movement ability. The powder supply unit 30, specifically, the reservoir 31 connected to the distal end of the support arm 25 through a connector 28, can thus be moved back and forth as well as up and down above the melt surface within the mold 1. Although the support arm 25 of the horizontally articulated type is used in this embodiment, any desired type arm may be employed so long as it allows the powder supply unit 30 to freely move within the mold 1. From the inventors experience, however, the above mentioned horizontally articulated arm has been found to be preferable because a space around the mold 1 and the tundish 19 is usually very small and narrow. Also, the combination and number of respective components of the support arm 25 and the drive unit 28 may be determined appropriately dependent on an installation space in the working site so as to ensure effective operation. In response to a command from the powder supply (or scatter) controller 24, the reservoir 31 can be moved to a predetermined location within the mold 1, and the powder 7 can then be supplied by opening the bottom of the reservoir 31 or turning (tilting) it. In this embodiment, the support arm 25 and the feed unit 35 of the powder 7 are mounted on a base 40 on which the tundish 19 is rested. But, a part or all of those system components, for example, the multi-articulated support arm 25 and the drive unit 28 for driving the support arm 25, may be constituted in self-propelled fashion, as required. Next, a method of detecting a melt surface abnormal condition will be described below. Brightness of the melt surface within the mold 1 appears in a very wide range from a dark state where the melt surface is entirely covered with the non-molten powder 8, to a light state as found in the event of boiling where the melt surface is covered with flames caused when the gas blown into the pouring nozzle 12 is discharged to the melt surface within the mold 1 in a large amount, or with the molten powder 9 exposed or emerged upon violent and wavy movement of the melt surface. Brightness of the melt surface is also constantly fluctuated due to oscillations of the mold, discharge of the blown gas, and emergence of the molten powder layer 9. Paying attention to such fluctuated condition of brightness of the melt surface, the inventors have studied correlation between that fluctuated condition and the melt surface abnormalities. First, to precisely detect brightness of the melt surface, an image sensor 20 was installed in positions above and facing the melt surface within the mold 1. Since the pouring nozzle 12 is disposed at the center of the mold 1, use of the single sensor 20 is difficult for viewing the whole of the mold 1 and cannot precisely detect brightness of the melt surface with such high accuracy as intended by the present invention. Therefore, the image sensor 20 is installed as a pair at positions opposite to each other with the pouring nozzle 12 located therebetween, or in two or more pairs when the entire melt surface region within the mold cannot be covered by a total viewing field of one pair of image sensors for the reason of size of castings to be cast. Although the image sensor 20 can be of any desired type, for example, a camera utilizing a vidicon generally used in video cameras, the inventors have employed, as a result of close study, CCD or C-MOS cameras using solid-state image sensors which can be easily reduced in size and are preferable to cope with phenomena of image burning and residual images. This is due to the fact that because brightness of the melt surface is constantly fluctuated in a fairly wide range as mentioned above, such cameras are required to be durable against phenomena of image burning and residual image, and also to be small from the standpoint of a space available above the mold. Incidentally, the CCD or C-MOS cameras are adjusted in irises such that they become substantially saturated in the event of boiling. FIGS. 2 and 3 are images of the melt surface picked up by the CCD or C-MOS cameras, each of the images representing a melt surface condition within the mold on either side of the pouring nozzle. It is assumed that FIG. 2 is an N-side image and FIG. 3 is an S-side image. The cameras are so disposed as to leave the pouring nozzle out of their viewing fields, because the pouring nozzle is always in a light state during casting operation. In FIGS. 2 and 3, designated by 2, 3 are wall surfaces of the mold 1 contacting with the casting 4 being cast into products and the powder 8.sub.1, and 8.sub.2 is non-molten powder layer of the powder 7 supplied (or scattered) over the melt surface. An original image 42 or 43 as shown in FIG. 2, 3 is input to the arithmetic processing unit 21 which processes only an image 42a or 43a in a frame corresponding to the melt surface region within the mold hatched in FIG. 2 or 3 for detection of the melt surface abnormality based on the input image, as described later. For practical use of the present invention, a profile of the image 42a or 43a corresponding to the melt surface region within the mold is not necessary, defined strictly along the boundary between the mold wall surface 2 or 3 and the melt surface region. The inventors have examined correlation between the above images and a melt surface abnormal condition from the various standpoints. At the outset, a reference threshold level was set between brightness of the non-molten powder layer 8 and brightness of the molten powder layer 9 to process the input image for conversion to binary, or light and dark representation in such a manner that respective pixels jointly constituting the input image are judged to be dark if pixel's brightness is comparable to that in a case where the melt surface is covered with the non-molten powder layer 8, and to be light if pixel's brightness is comparable to that in a case where the molten powder surface 9 is emerged to the melt surface. FIGS. 4 and 5 are other examples of the N-side and S-side images, showing states where the molten powder layers 9 are emerged in the non-molten powder layers 8 within the melt surface regions 42a, 43a, respectively, FIGS. 6 and 7 are binary images 42b, 43b resulting from converting the input images of a melt surface condition corresponding to FIGS. 4 and 5 to a binary representation, respectively, by the above mentioned method. In these binary images 42b, 43b, the emerged parts of the molten powder layers 9 over the melt surface are indicated as light portions 45, while the remaining non-molten powder layers 8 are indicated as dark portions 44. Next, the inventors determined the ratio of the area of the light portions in the binary images, such as, the emerged parts of the molten powder layer 9, (hereinafter referred to as light portion area) to the entire regions of the melt surface within the viewing fields of the cameras corresponding to the melt surface regions 42a and 43a (that is, total area of the emerged parts of the molten powder layer 9 and the remaining non-molten powder layer 8), that is, the ratio of the light portion area to the melt surface area within the viewing fields of the cameras (that ratio is called the light portion area ratio in this specification and hereinafter referred to simply as area ratio Rn), and then also examined time-dependent changes in the area ratio Rn. The reason for using the area ratio here is because various examinations have revealed that the regions 42a, 43a corresponding to the entire melt surface to be observed do not necessarily become the same area at all times due to varying size of castings to be cast into products, for example, and also varied in area for different melt surface abnormal phenomena or even the same melt surface abnormal phenomenon, thereby leading to difficulties in steadily detecting the respective melt surface abnormal phenomena in a quantitative manner. In view of the above, it has been decided to calculate the area ratio of the light portion area in the melt surface to the regions 42a, 43a corresponding to the entire melt surface to be observed. It should be understood that in a case where the melt surface abnormal condition can be detected and processed constantly under the same conditions for the regions 42a, 43a corresponding to the entire melt surface to be observed, an absolute value of the area may be used rather than the area ratio in relation to the detecting procedures explained below. FIG. 8 is a graph showing one example of an examined result during the casting operation, in which the vertical axis represents the area ratio Rn and the horizontal axis represents the lapse of time. In this example that there occurred lack of powder as one of the melt surface abnormalities, parallel to measurement of the area ratio, a skilled operator monitored a melt surface condition and scattered powder from his judgment at the time point indicated by an arrow A in FIG. 8. It will thus be seen that the area ratio Rn is increased with the lapse of time and then abruptly reduced at the time of scatter of the powder. In FIG. 8, fine fluctuations of the area ratio Rn as indicated by arrows P, Q are caused by oscillations of the mold 1. Specifically, when the melt surface is relatively risen up upon a descent of the mold 1, the molten powder layer 9 is marged in a large part to provide the maximum as indicated by P. Conversely, when the mold 1 is ascended, the emerged part of the molten powder layer 9 is hidden and the non-molten powder layer 8 occupies a larger part, thereby providing the minimum as indicated by Q. In addition, the gas blown into the pouring nozzle is periodically discharged to and disappears from the melt surface corresponding to oscillations of the mold 1, and this can also produce the fine fluctuations including extreme. Stated otherwise, the area ratio Rn is finely fluctuated due to wavy movement of the melt surface caused by the mold oscillations and the like. In the method of precisely detecting a melt surface condition, the fluctuations like P, Q caused by, for example, the mold oscillations become disturbances which are objectionable for the arithmetic processing to detect the melt surface abnormality. Accordingly, the effect of such disturbances has to be eliminated. As one eliminating method, it can be considered to calculate a time-series average of instantaneous values of the light portion area detected at respective preset time points. While depending on a length of the time for the time-series average, processing becomes relatively intricate in this method. Besides, in order to precisely eliminate the disturbances, a length of the time for calculating the time-series average must be prolonged to some extent (at least longer than an oscillating period of the mold). Accordingly, the resulting increased processing time is unfavorable for the purpose of detecting a melt surface abnormal condition precisely and quickly. In this embodiment, therefore, the logical products of the binary images 42b and 43b are taken during a duration corresponding to the oscillating period of the mold 1, thereby to eliminate periodic or abrupt fluctuations of the light portion area in the melt surface images. FIG. 9 shows a processed result obtained by taking the logical products of the binary images in the region corresponding to that of FIG. 8. It will be seen that there scarcely appear fine fluctuations in the processed result of FIG. 8, and the time-dependent changes in the area ratio Rn is clarified to ensure more stable detection, as a result of satisfactory elimination of the disturbances. The time-dependent changes in the area ratio Rn for the mold during the casting operation were measured while eliminating the disturbances by the above method. FIG. 10 shows one example of a measured result of the area ratio Rn. FIG. 10 comprises a set of upper and lower graphs which correspond to the results obtained by processing, through the above described manner, the input images from the N- and S-sides, that is, from the two cameras installed in positions opposite to each other with the pouring nozzle located therebetween, respectively. Thus, the upper graph represents the processed result of the N-side image, while the lower graph represents that of the S-side image. In FIG. 10, arrows X, Y and Z indicate the timings at which the melt surface abnormalities such as lack of powder (powder scatter), boiling and biased flow occurred, respectively, Subscripts indicate the side, N-side or S-side, and the number of occurred times from the start of measurement. In this embodiment, parallel to the above measurement, a skilled operator monitored a melt surface condition and, when some melt surface abnormality occurred, he took an action to stabilize that melt surface abnormality. The arrow X represents an example where lack of powder occurred and the powder was scattered at the arrow x from judgment of the operator. It will be seen that the area ratio Rn is increased with the lapse of time and then abruptly reduced at the same time as scatter of the powder. Likewise, at the arrows, Y, Z representing the occurrence of boiling and biased flow, respectively, any one or more of the following, casting speed control, flow rate control of the gas blown into the pouring nozzle, flow rate control of the melt poured into the mold, and scatter of the powder, were carried out based on the judgment of the operator to solve the melt surface abnormalities. In any cases, the timings selected to eliminate the melt surface abnormalities match those timings at which the area ratio Rn is abruptly reduced. From FIG. 10, correlation between the abnormal phenomena occurred in the melt surface and the time-dependent changes in the area ratio Rn is apparent. But, while FIG. 10 makes clear a qualitative tendency that leads to the respective abnormal phenomena, the area ratios Rn at the timings to take actions are somewhat different from one another and not enough stable in level to ensure precise detection of the abnormal phenomena. This means the necessity of further quantative evaluation of the data. Therefore, change rates Dn of the area ratio Rn per unit time were examined. FIG. 11 is a graph showing the change rates Dn with the lapse of time, in which the horizontal axis represents a time base on the same scale as that of FIG. 10. Note that the change rates after the timings to eliminate the melt surface abnormalities, that is , negative change rates, are not shown in FIG. 11, because those change rates correspond to a decrease in the area ratio Rn incidental to, for example, scatter of the powder and hence are meaningless for detection of an abnormal condition. From such evaluation, it is first found that the occurrence of biased flow and boiling can be precisely detected based on the area ratio Rn and the change ratio Dn thereof shown in FIGS. 10 and 11. More specifically, the area ratio Rn is relatively varied and not stable in level as indicated by Y.sub.N1, Y.sub.S1 and Y.sub.N2, Y.sub.S2, such that it may sometimes take a level comparable to that in a case of lack of powder. However, the change rate of the area ratio Rn is at most 150%/min at the occurrence of lack of powder, while it shows a much larger value beyond 300%/min at the occurrence of biased flow and boiling. Further, such large positive change rates are observed on only either one side with respect to the pouring nozzle Like Z.sub.S1, Z.sub.S2 at the occurrence of biased flow, but are observed substantially at the same time on both the opposite sides of the pouring nozzle at the occurrence of boiling. It was thus found that the occurrence of biased flow and boiling could be detected distinctively from comparison of the change rates Dn of the area ratios Rn obtained for the opposite sides. Here, am the occurrence of boiling, the area ratios for the corresponding sides or zones are relatively varied due to flames of the blown gas and the like. In other words, the area ratios Rn are saturated up to 100% in some cases like Y.sub.N1, Y.sub.S1, but not saturated up to 100% in other cases like Y.sub.N2, Y.sub.S2. The latter cases where the area ratios Rn are not saturated up to 100% correspond to a phenomenon that the melt surface is forced to warily move due to a light degree of boiling, so-called rippling phenomenon. Since the rippling phenomenon can be eliminated or prevented in the similar manner as that in the event of boiling, there is usually no need to further discriminate rippling from boiling. On demand, however, such discrimination can be made by detecting whether or not the value of the area ratio Rn is saturated at the occurrence of abnormality. Then, for detection of an abnormality related to the powder, that is, lack of powder, it was found that both the area ratio Rn and the change rate Dn were required to be taken into consideration. This is because, in FIGS. 10 and 11, some area ratios Rn as indicated by arrows X.sub.N8, X.sub.S6, for example, may have a level comparable to that in a case of boiling (rippling) as indicated by arrows Y.sub.N2, Y.sub.S2, while the changes rates Dn of the area ratios Rn per unit time are always significantly small as high as 150%/min, as mentioned above. However, also as stated above, the area ratio Rn and the change rate Dn observed at the occurrence of lack of powder are both relatively varied and not stable in level. This is attributable to the fact that those zones in the melt surface which are subjected to lack of powder are not uniformly distributed over the whole area, but localized into some parts, that is the zones subjected to lack of powder are changed in the mold between the pouring nozzle and the short side wall opposite thereto on each of the N- and S-sides, namely, in a direction of width of the mold, whereby the light portions are also changed corresponding to the varying area of the zones subjected to lack of powder. Accordingly, it is required for the detection of lack of powder to monitor the melt surface within the mold in a quantitative manner for both positions and conditions of the lack of powder thereat. Therefore, the images 42a and 43a of the melt surface within the mold were each subdivided into multiple zones. Here, the division into multiple zones is made in accordance with the mode or state how lack of powder occurs. This point will be taken into further consideration below. Since the powder is melted and then consumed while flowing into between the mold walls and a casting, lack of powder is frequently caused along the mold walls as a phenomenon appeared on the melt surface. In general, however, the powder is in the form of fine particles or granules which have a rest angle of about 30.degree.-60.degree. in either case, so that the scattered powder will not be kept at one location. Therefore, the powder at the center in a direction of thickness of the mold, in both non-molten and molten states, tends to move in all directions on the plane of the melt surface with wavy movement of the melt surface due to oscillations of the mold. Here, the movement mode of the powder on the melt surface within the mold is dependent on fluctuations of the melt surface, entire mass balance of the powder over the melt surface, thermal balance of the powder in a molten state, and the like. Thus, lack of powder may happen at the central region at a direction of thickness of the mold. Specifically, it has been found from observation of the melt surface that, while undergoing some variations, the powder is usually moved in a direction of thickness of the mold at a speed comparable to or about four times that in a direction of width the mold. Further, although the powder can be scattered in various manners, it is in any case followed that, the powder is supplied not at a single point, but over some supply zone. From the foregoing, the division of the image into too many zones is apparently ineffective and meaningless, because it reduces a processing speed for recognition of a melt surface abnormality, which in turn leads to a delay in detection and hence beginning of a necessary action to be taken, thereby increasing the probability of an outbreak of any operational trouble such as sticking type BO (break-out), for instance. Accordingly, in view of the above mentioned movement mode of the powder, it is sufficient to divide the image into three or less zones in a direction of thickness of the mold, and into the zones in number from the nearly same number as the above to four in a direction of width of the mold. Taking into account the foregoing, in this embodiment, the melt surface regions within the mold corresponding to the images 42a, 43a in FIGS. 2 and 3 are divided into five zones 42a1-42a5 and 43a1-43a5 as shown in FIGS. 12 and 13 for the opposite N- and S-sides with respect to the pouring nozzle, respectively. The divided zones have all the same size in both sides. The total size of all the zones on each side is 500 mm in the direction of thickness and 250 mm in the direction of width. Each divided zone has a length of 100 mm in the direction of width. For the respective zones which are individually subjected to similar examination as that in the above case, the measured results of area ratios Rnn are shown in FIGS. 14, 15 and the measured results of changes rates Dnn are shown in FIGS. 16, 17, in which drawings the horizontal axis as a time base has the same scale as that of FIG. 10. FIGS. 14 to 17 each comprises five graphs which correspond to the measured results of the respective divided zones in FIGS. 12 and 13. Thus, those five graphs correspond to the zones 42a1-42a5 from above for each of FIGS. 14 and 16, and to the zones 43a1-43a5 from above for each of FIGS. 15 and 17, respectively. As with the above case, correlation between the phenomena occurred in the melt surface and time-dependent changes in both the area ratios Rnn and the change rates Dnn is clear. It is also apparent that the division of the image enables to monitor the respective divided zones in a stable and quantitative manner for detection of lack of powder. More specifically, in any event of lack of powder, biased flow and boiling, the area ratios Rnn reach a high level of 80-90% or more. But, as described above, it is found also, in case of dividing the image, that the change rates Dnn show a large variation of 300%/min or more at the occurrence of biased flow and boiling, while they become small at most 150%/min at the occurrence of lack of powder, thereby making it possible to discriminate lack of powder from biased flow and boiling. Furthermore, in contrast with a variable and unstable state in level of the area ratios Rn and the change rates Dn at the occurrence of lack of powder as appeared in FIGS. 10 and 11, the division of the image enables to detect a condition of lack of powder for the respective divided zones in a still more clear and stable manner. More specifically, when lack of powder has occurred in any one zone, the light portion area ratio Rnn is increased almost in that zone alone as seen at the lapsed time points 3, 15, 20 and 25 minutes in FIG. 14 and at the lapsed time points 3 and 6 minutes in FIG. 15. Such a condition of lack of powder can be detected by setting reference levels (or values) of the area ratio Rnn and the change rate Dnn for each of the divided zones, and then determining whether the detected value exceeds the reference level of the area ratio Rnn, but not the reference level of the change rate Dnn. Also, when lack of powder has occurred over a plurality of zones, the area ratios Rnn are increased in those respective zones as seen at the lapsed time points 2, 5, 15 and 23 minutes in FIG. 14 and at the lapsed time points 2, 5, 21 and 25 minutes in FIG. 15, similarly to the above. In this case, too, lack of powder can be detected by setting reference levels of the area ratios Rnn and the change rates dnn for the respective zones, and then determining whether the detected values exceed the reference levels of the area ratios Rnn, but not the reference levels of the change rates Dnn. Since each reference level has to be set in accordance with actual operating situations, the respective reference levels are required to be set to match individual cases dependent on, for example, size of a casting to be cast into products and environments around the mold. In addition, at least some of the divided zones may have the same reference level, or all the zones may have reference levels different from one another. In this embodiment, biased flow, boiling and lack of powder could be detected by respectively setting a reference level of the area ratio at 70% uniformly and that of the change rate at 150%/min uniformly. Note that the reference levels are not limited to those values as see in this embodiment, because they should be set dependent on such conditions as size of the pouring nozzle, length of the pouring nozzle projecting above the melt surface, and cross-sectional size of a casting to be cast into products, as required. In this way, it is possible to steadily and quantitatively detect conditions and positions of lack of powder. Also, it is needless to say that boiling and biased flow can be detected by obtaining the change rates Dnn of the area ratios Rnn for the respective zones and comparing the change rates Dnn of the corresponding zones on the opposite sides of the pouring nozzle to each other, in a case of image subdivision as well. However, since biased flow and boiling have to be detected in a shortest time, they are usually detected by the above mentioned method using the entire regions on both sides of the pouring nozzle without dividing their images. It should be understood that such a technique is not necessarily needed in a case where the division of the image will not arise problems, for example, significant delay in detection, in view of the relation between the operating situations and the detection time. Thus, the melt surface abnormal phenomena such as lack of powder, biased flow and boiling can be automatically detected in a stable and quantitative manner. In the above description, automatic detection of the melt surface abnormal phenomena is made by processing the image information obtained by the image sensors installed above the melt surface within the mold. In addition to that, it is also possible to detect the melt surface abnormal phenomena for the purpose of stable casting operation, by incorporating temperature information from thermocouples built in the mold wall. More specifically, thermocouples are embedded in the mold wall at a certain depth from its surface in plural series along a direction of casting and in plural number along a circumferential direction of the mold to measure temperatures at the respective embedded points, so that heat flux at the respective built-in points as well as an amount of heat drawn out of the solidified shell of the casting being cast into products are calculated based on the embedded situations, and a heat drawn condition, such as, a condition of the intruding powder, is monitored in a circumferential direction of the mold, thereby detecting an abnormal condition in the mold. For example, when the amounts of drawn heat are abnormally varied in a circumferential direction of the mold, this is judged as a state where the powder does not flow into between the mold and the solidified shell with satisfactorily uniform distribution. Such an abnormality can be eliminated by, for example, supplying (or scattering) the powder to a region corresponding to the circumferential location at which the powder was found short, modifying a taper of the mold, or changing a casting speed. Next, there will be described a melt surface abnormality preventing method which is effective in stabilizing a melt surface abnormal condition immediately after it has been detected. To begin with, a preventing method to be effected when the occurrence of biased flow or boiling has been detected by the melt surface abnormal condition detecting method as mentioned above will be described. When boiling occurs, the volume of the blown gas in a flow passage of the melt is increased and the flow rate of the melt is hence relatively reduced. Accordingly, it is required to adjust a casting speed to match the flow rate of the melt at that time. At the same time, the flow rate of the melt is also adjusted in order to suppress fluctuations of the melt surface level within the mold. Further, the flow rate of the blown gas is then adjusted to match the occurred situations, thereby to eliminate a boiling phenomenon. If lack of powder is found simultaneously, the powder is scattered to the lack position thereof. At any stage during the preventing procedures, the above adjustments may be stopped at that stage if the stage has been stabilized. After stabilization, the operating conditions are gradually restored to those immediately prior to the occurrence of the abnormality. Next, in the event of biased flow, it is observed in many cases that the flow passage of the melt is disturbed by deposits such as Al.sub.2 O.sub.3 precipitated or separated out in the melt flow passage. Therefore, a biased flow state is eliminated through a slightly varied adjustment of the flow rate of the melt by, for example, moving a sliding nozzle (SN) vibratingly with small strokes, when it is used for adjusting the flow rate of the melt, thereby to change a condition of the deposits, or regulating the flow rate of the blown gas. At this time, the casting speed may additionally be adjusted in order to suppress fluctuations of the melt surface level. Further, if lack of powder is found simultaneously, the powder is supplied to the position of lack thereof. The preventing method of the present invention, to be effected when some melt surface abnormality has been automatically detected by the melt surface abnormal condition detecting method, can be performed by the controller 23 which instructs and controls actions needed to stabilize the melt surface abnormality in response to detection of an abnormal condition. Thus, the controller 23 is operated to issue an alarm, allowing an operator to take a proper action in response to the alarm, or to directly issue a control signal to the casting speed controller 18, the blown gas flow rate controller 16, the melt flow rate controller 14, and/or the powder supply unit 30 for automatic control to eliminate the abnormality. In a case of the latter automatic control, the casting speed, the flow rate of the blown gas or the flow rate of the melt is first lowered down to setting values which are determined dependent on the operating conditions, for example, width of a casting and casting speed, as established at the time of the occurrence of an abnormality, thereby to automatically detect the melt surface abnormality. If the occurred abnormality has been prevented or eliminated, the operating conditions are then gradually restored to those established when the abnormality occurred. If not prevented or eliminated, the above variables are further lowered down to other setting values which are determined dependent on the operating conditions as that time, for repeating similar control. When the occurrence of biased flow or boiling is detected by the melt surface abnormal condition method in this way, a proper action is taken to stabilize the melt surface abnormal condition. Next, a preventing device to be operated when lack of powder has been detected by the melt surface abnormal condition method will be described below. Since lack of powder is detected for each of the divided zones of the melt surface as mentioned above, the powder is scattered to the area corresponding to the divided zone where the powder was found short. Conventional powder supply units have been accompanied by not problems such as clogging of a pipe, as explained above. Therefore, the present invention has employed a powder supply unit of the type that is separated into a powder supply section and a powder scattering section, as shown in FIG. 18. Designated by 30 is the powder scattering unit which comprises a cup-like reservoir 31 for storing a preset amount of the powder 7 corresponding to each of the divided zones, and a support frame 32 supporting the reservoir 31 and mounted to the distal end of the multi-articulated support arm 25 (comprising two support arms 26, 27 in this embodiment which are referred to collectively as support arm 25). The support arm 25 is constituted, as explained above, such that a plurality of support arms 26, 27 are interconnected through arm drivers 28a-28c each having a rotatable shaft built therein, whereby the support arms can be rotated about the respective rotatable shafts in an articulated manner. In addition, the support arm 25 is provided at the distal end thereof with a lift driver 28d and a lift frame 29 to provide a vertically moving ability. The support frame 32 supported on the distal end of the support arm 25 through a connector 28 can thus be moved back and forth as well as up and down above the melt surface within the mold 1 by controllably driving the arm drivers 28a-28c and the lift driver 28d. The powder supply unit 30, specifically the reservoir 31, is supplied with the preset amount of powder 7 by a feeder 35 which comprises a hopper 39, valves 36, 37, and a feed nozzle 38. In response to a control signal from the controller 23 which issues a command for a position of lack of the powder, and the like, a controller 24 drives and controls a later-described drive mechanism of the powder scattering unit 30 and the arm drivers 28a-28c (referred to collectively as drive unit 28) of the support arm 25, thereby supplying the powder to the zone corresponding to a position of lack of the powder detected by the melt surface abnormal condition detecting method explained above. When a condition of lack of powder has been detected over a plurality of zones simultaneously as appeared at the elapsed time points 5 and 25 minutes in FIGS. 14 and 15, for example, the controller 23 decides the priority of scatter for those zones where lack of powder has been detected, thereby permitting to carry out efficient powder scatter. FIG. 19 is a partial structural view showing one example of the powder supplying section including a reservoir 31 used in this embodiment for storing a preset amount of the powder 7. This example comprises the reservoir 31 for storing the powder, a support frame 32 for supporting the reservoir 31, a bottom lid 33 to be opened at the time of supplying the powder 7, a drive motor 46 for driving the bottom lid 33 to open and close, and a rotatable shaft 47. Here, the volume of the reservoir 31 for storing the powder 7 is selected as follows. First, the plane area of the reservoir 31 is set to be less than the size of each divided zone, taking into account downward spread of the powder when supplied. Then, the height of the reservoir 31 is a factor determining the total volume thereof, and hence required to be set so as to make a powder supply rate, which is determined dependent on both an operation speed of the powder supply unit 30 and the volume of the reservoir 31, not less than a maximum consumed rate of the powder during the casting operation. In this embodiment, therefore, since the required volume of the reservoir 31 becomes 1 Q under assumption that the maximum consumed rate of the powder during the casting operation is 2.4 kg/min, the powder supply time (period) per stroke is 20 sec and specific gravity of the powder is 0.8, the dimensions of the reservoir 31 were set to have length, width and height of 100 mm.times.100 mm.times.120 mm as shown in FIG. 20 based on the area of each divided area, respectively, in view of a rest angle found in supply of the powder from the feed nozzle 38 to the reservoir 31 as well. In a case where the reservoir 31 cannot have the sufficient dimensions due to restrictions in equipment layout such that the maximum consumed rate of the powder largely exceeds the supply rate of the powder supply unit 30, the powder may be supplied by providing plurality of the powder supply units 30 for example two, which are each allocated to cover one of the melt surface regions on the opposite sides with respect to the pouring nozzle 12. The powder feed unit 35 is selected to provide a slope angle allowing the powder to drop by gravity or dead load from the hopper 39 to a feed port 38 capable of facing the reservoir 31. A fixed feed amount of the powder from the hopper 39 to the reservoir 31 can be adjusted by changing a diameter or length of a pipe connecting the upper valve 36 and the lower valve 37. The powder can be fed to the reservoir 31 which has been preparatorily moved below the feed port 38, on demand, by following the procedures of closing the lower valve 37, opening the upper valve 36, closing the upper valve 36 and then opening the lower valve 37. In a case where various types of powder are used during the casting operation it is possible to adapt for such a case by providing a plurality of hoppers corresponding to the powder types used, sharing the pipe downstream of the upper valve 36 or the feeder nozzle 38, and arranging recovery lines led back to the respective hoppers. The powder can be scattered into the mold in various manners. While this embodiment employs scatter means of the type in which the single bottom lid 33 is opened, the bottom lid may be comprised of plural members. Also, as shown in FIG. 21, the reservoir 31 may itself be turned or tilted in a direction of an arrow R1 or R2 for scattering the powder. Alternatively, as shown in FIGS. 22 and 23, the powder may be scattered by pivoting a side plate 34 and turning the reservoir 31 in a direction of an arrow R1. As an alternative, as shown in FIG. 24, the bottom lid 33 may be turned or slid in a direction of an arrow R3 or R4 for scattering the powder. In this way, the powder is supplied by the powder supply unit 30 to the zone corresponding to a position of lack of the powder that is detected by the melt surface abnormal condition detecting method. As a result of implementing the casting operation by using the melt surface abnormal condition detecting method, the melt surface abnormality preventing method and the powder lack preventing device, as explained above, under the casting conditions that the cross-sectional size of castings is 250.times.1250 mm and the casting speed is 1.6 m/min, that casting operation could be stably implemented with no intervention of operators. In other words, even when the melt surface abnormality such as boiling or biased flow occurred upon changes in the weight of the melt within the tundish due to replacement of ladles during the continuous casting process in which a plurality of different ladles are used, it was possible to quickly detect and stabilize an abnormal condition and also to promptly detect lack of powder for supplying the powder. Furthermore, the castings thus cast into products were completely free from any surface flaws and other defects and hence very excellent in quality. In short, the method and apparatus of this embodiment enables to automatically detect and prevent an abnormal condition of the melt surface within the mold. As a result, it becomes feasible to cut down an amount of manual work required around the mold, which has been so far an obstacle in saving human labors, and to provide the stable continuous casting operation with less variations otherwise possibly caused by operators, as well as the good quality of castings. The second embodiment deals with the problems related to blow of inert gas into a melt (see the above section B). FIG. 25 is a view showing an example of implementing this embodiment in a general continuous casting equipment. A melt (molten steel) 5 is once stored from a ladle 48 into a tundish 19 and then poured into a mold 1 through a nozzle 49. The nozzle 49 of this embodiment comprises an upper nozzle 50 mounted to a bottom wall of the tundish 19, a sliding nozzle 51 mounted at the bottom of the tundish 19 in contact with the upper nozzle 50, and a pouring nozzle 52 attached integrally to a movable plane of the sliding nozzle 51. A gas supply line 53 is connected at its fore end to the upper nozzle 50 so that gas is blown into a flow of the melt in the upper nozzle 50 through its wall. In this embodiment, a small-type CCD camera is used as an image sensing device (or image sensor) 56. The image sensing device 56 is installed above the mold including a pair of sensors, each locating on either side of the pouring nozzle 52. The image sensing devices 56 pick up images of the surface of the molten steel, that is, melt surface y, within the mold during continuous casting, and apply detected image signals to an image processor 57. For recognizing bubbles produced on the melt surface y within the mold, the image processor 57 first converts the image signals to a binary representation, that is, bilevel signal. The melt surface y is usually covered with powder and hence indicated as a dark portion in the image. If bubbles are produced on the melt surface y, the melt is exposed or emerged as a light portion with the bubbles. Therefore, only the bubbles can be recognized as a light portion by making a binary representation with a threshold level set so as to discriminate between the dark powder portion and the light melt emerged portion. Then, noises included in binary images caused by, for example, flames flared up from the melt surface y are removed by subjecting the binary images to the AND-ing process plural times in a time-series manner, and then superposing them. Flames are momentarily changed in their positions and sized, while bubbles remain as a light portion at the same position for a longer time than flames. Accordingly, by taking the binary images plural times for a short period of time and subjecting them to the AND-ing process, the noises due to flames can be removed. Then, the number of light portion islands caused by bubbles in the binary images, which has been processed to extract only the bubble data, is measured by counting them on the images, thereby detecting the number of bubbles floating over the melt surface y. Further, the size (area) of flames flared up from the melt surface y within the mold by burning of powder components are measured. While flames flared up from the melt surface y are thus caused by burning of powder components, when a larger amount of gas is supplied into the melt, an amount of burning gas generated from the powder is also increased incidental to movement of the gas up to the melt surface, causing the flames to be flared up in larger size. Accordingly, a flow rate (amount) of gas supplied into the mold can be grasped by measuring the size of the flames. As procedures of measuring the size of the flames, the image is first converted to a binary representation with a threshold level appropriately preset, because the flames are indicated as a light portion in the image. Only the light portion fluctuating in a short time is then extracted contrarily to the method of detecting the bubbles. This extraction can be achieved by OR-processing the binary images which have been taken several times in a time-series manner. Afterward, the size of the extracted flames is measured from the resulting image. The number of bubbles and the size of flames thus detected are input to a comparator 58. The comparator 58 stores therein correlation between a flow rate of the blown gas and the number of bubbles within a range up to the occurrence of boiling and nozzle clogging, and upper and lower limit values (hereinafter referred to collectively as "allowable limit values") of a flow rate of the blown gas previously derived from the correlation. The detected values for the number of bubbles and the size of flames obtained through the above mentioned processing carried out by the image processor 57 are compared with the allowable limit values in the comparator 59, thereby to perform flow rate control so that the gas is blown within the allowable limit values. More specifically, when the detected values exceed the allowable upper limit values, the comparator 58 issues a control signal to a valve 54 installed in the gas supply line 53, thereby making control to reduce a flow rate of the blown gas. On the contrary, when the detected values fall below the allowable lower limit values, a control signal is issued to the valve 54 to increase a flow rate of the blown gas, thereby ensuring that a proper flow rate of the blown gas within the allowable limit values is always supplied during continuous casting. Adjustment of the gas supply valve 54 is not necessarily limited to an automatic manner effected using the comparator 58. Alternatively, in accordance with an indication of the proper gas flow rate shown by the comparator 58, an operator may adjust the valve 54 while looking at a flow meter 55. Further, like a control device as disclosed in Japanese Utility Model Laid-Open No. 62-142463, it is also possible to obtain a required flow rate of gas to match an amount of the poured melt which is calculated based on the head size of the tundish, the width and thickness of castings, and the casting speed. A difference between that required flow rate of gas and the flow rate of gas detected as described above within the allowable limit values is then determined, and the resulting differential flow rate is employed as a modification value for a calculated gas flow rate setting value to be used by the comparator. With this technique, calculation of the proper gas flow rate is not necessarily required to be carried out from time to time, and may be performed only at start-up of the casting operation, or at the time of changes in the operating conditions or the conditions of hardware such as the nozzle. As a result, the burden imposed on the image processing and the like is alleviated. FIGS. 26 and 27 are each a binary representation of an image picked up by either one of the image sensing devices 56 directed to the melt surface within the mold. FIG. 26 shows a state where the gas blown into the melt within the mold is moved up to the melt surface and bubbles k are produced on the melt surface. The number of those bubbles is detected to determine a proper value for the flow rate of the blown gas. FIG. 28 is a graph showing one example of an examined result of the relationship between the number of bubbles and a flow rate of blown gas. It will be seen that the number of bubbles is increased with an increase in the flow rate of gas and, when the flow rate of gas exceeds a certain value, there occurs a boiling phenomenon. Further, the flow rate of gas is decreased with a decrease in the number of bubbles, eventually leading to clogging of the nozzle. In short, it was confirmed that there was clear correlation between the flow rate of gas and the number of bubbles. The control can be practiced by, after determining the above correlation dependent on the equipment conditions and the operating conditions in advance, setting a maximum value of the number of bubbles at lower than which the stable casting operation can be continued without causing the boiling phenomenon, that is, upper limit value, and a minimum value of the number of bubbles at higher than which the stable casting operation can be continued without causing clogging of the nozzle, that is, lower limit value. It has also been found that the upper and lower limit values may be set for safety while leaving an allowance of about 20% with respective to each of a boiling risk line a and a clogging risk lien line b shown in FIG. 28. FIG. 27 shows a result of detecting flames i flared up on the melt surface when the gas blown into the melt within the mold is moved up to the melt surface and combustible components of the powder are burnt. The result of FIG. 27 is obtained by converting an original image to a binary representation and extracting only a light portion of the flames. The size of the flames i can be detected by obtaining a binary image through the above conversion to clarify images of the flames i and then measuring the area of the light portion. FIG. 29 is a graph showing one example of an examined result of the relationship between the size of flames and a flow rate of blown gas. The flow rate of gas is increased with an increase in the size or area of flames, while it is decreased with a decrease in the area of flames. For the size of flames as well, upper and lower limit values are set with a boiling risk line a1 and a clogging risk line b1 being taken as references, respectively, similarly to the case of FIG. 28. The stable continuous casting operation can thus be continued by ensuring a flow rate of the blown gas within the allowable limit values. As will be apparent from FIGS. 28 and 29, however, the correlation between the size of flames and the flow rate of gas is considerably varies as compared with the correlation between the number of bubbles and the flow raze of gas. In a case of desiring to improve controllability by setting the allowable limit values closer to the boiling risk line a and the clogging risk line b, for example, it is preferable to use the detected value for the number of bubbles. On the other hand, the size of flames is advantageous in prompt response to changes in the flow rate of gas. Accordingly, whether to use the detected value for the number of bubbles or the detected value for the size of flames may be determined dependent on, for example, the equipment and operating conditions as well as other environmental conditions. It is also possible to use both of them. As described above, this embodiment makes it possible to improve the quality of castings and significantly reduce the probability of nozzle clogging, by determining a proper value for the flow rate of gas blown into the melt for control off the blown gas. The third embodiment deals with the problem of slag beard formation (see the above section C) . As a result of repeatedly observing a melt surface condition, the inventors found that there is a slight or delicate difference between powder, a slag beard and a wall surface within the mold. Therefore, the inventors installed an image sensor utilizing a TV camera or the like above the mold in a location facing a melt surface within the mold and examined lightness or brightness of the melt surface including the mold wall surface. FIG. 30 shows one example of a result of the above examination in contrast with a partial sectional view illustrative of a melt surface condition within the mold. In FIG. 30, designated by 1 is a mold, 5 is a melt (molten steel), 6 is a solidified shell produced by being contacted with the mold 1 and cooled, and 7 is powder comprising non-molten powder 8 and molten powder 9. The reference 60 is a slag beard formed of the molten powder 9 which has adhered or deposited and grown on the mold wall surface 2 by being cooled by the mold 1. An amount of the slag beard 60 projecting into the mold, that is, thickness of the slag beard 60, is indicated by t. Brightness of the melt surface was measured by installing a pair of image sensors 30 at the bottom of a tundish 19 in locations opposite to each other, as shown in later-described FIG. 31, detecting the melt surface vicinity (which hereinafter collectively implies the melt surface within the mold, the mold wall surface, etc.) by the image sensors 20 in the form of an image signal or a brightness signal, and then determining distribution of brightness along a line L intersecting the mold wall surface 2 and shown in later-described FIG. 32. The resulting distribution of brightness is indicated by a solid line X in FIG. 30. As will be seen from FIG. 30, an upper surface 2a of the mold 1 exhibits relatively high brightness, because it reflects light therefrom. On the other hand, a surface portion of the non-molten powder 8 exhibits a lower level of brightness, because it comprises a layer of just scattered powder at a lower temperature and is in the form of powder having lower light reflectance. However, if the slag beard 60 is formed, an adhering portion (A) of the slag beard 60 to the mold wall surface 2 exhibits extremely low brightness, because it is cooled by the mold 1 down to a low temperature and has characteristics less effective to reflect light. With increasing a distance from the mold 1, a temperature of the slag beard 60 is raised by being subjected to heat of the molten powder 9 so that brightness is also raised correspondingly (from A to B). At a distal end (B) of the slag beard 60 most remote from the mold wall surface 2, brightness becomes maximum and indicates a sharp peak in the distribution of brightness, because the molten powder 9 is caused to appear and disappear due to oscillations of the mold 1 and the slag beard 60 is itself red-heated by heat of the molten powder 9. Meanwhile, if the slag beard 60 is not formed, the melt surface is uniformly covered with the non-molten powder 8, whereby the distribution of brightness provides a flat profile substantially free of a peak, as shown in a dotted line Y in FIG. 30. Accordingly, by measuring the distribution of brightness in the melt surface vicinity along a line intersecting the mold, preferably a line substantially perpendicular to the mold wall surface 2, it becomes possible to detect a melt surface condition, that is, a position of the mold wall surface 2, and recognition of the slag beard 60, the non-molten powder 8 and the molten powder 9, particularly, a condition of the formed slag beard 60 such as the presence or absence of the slag beard 60 and size (thickness) thereof. In other words, by measuring the distribution of brightness along the line L, and determining the presence or absence of the maximum value B above a predetermined level and the minimum value A below a predetermined level, as well as a distance between the positions where the maximum value B and the minimum value A are detected (hereinafter referred to simply as maximum to-minimum distance) t1, it becomes possible to detect the presence (corresponding to the presence of the maximum value B and the minimum value A) or absence of the slag beard 60, and a thickness t thereof (for example, size of the slag beard slag 60; in a horizontal direction) . Thus, a condition of the formed slag beard 60 can be detected through procedures of previously determining an indication pattern of the maximum value B and the minimum value A, as well as correlation between the detected maximum-to-minimum distance t1 and the thickness t of the actually formed slag beard 60, dependent on the equipment conditions around the mold, the operating conditions, characteristics of the installed image sensors and a signal processing system, and then measuring the actual distribution of brightness representing a melt surface condition from time to time during the casting operation. FIG. 31 is a block diagram showing one example of a device for detecting a condition of slag beard formation in accordance with the present invention. Designated by 62 is a slag beard formation detecting device, 52 is a pouring nozzle for the melt 5, 51 is a sliding nozzle, and 19 is a tundish. Designated by 20 is an image sensor for detecting the melt surface vicinity within the mold 1. In this embodiment, the image sensor 20, including one pair of sensors is installed above the mold 1 in opposite relation to the pouring nozzle 52, so that at least one image sensor can view the entire melt surface vicinity on either side about the pouring nozzle 52. Designated by 61 is a cooling and dustproof box for protecting the image sensor 20. Each image sensor 20 has its bottom surface formed of heat-resistant and transparent glass, and is constituted so as to allow cooling air to circulate the interior thereof. The cooling and dustproof boxes 61 have a practically valuable function of effectively protecting the image sensors 20 even under such severe environments where the image sensors 20 are installed near the melt 5 and subjected to heat from the melt 5 and dust such as the powder 7. Designated by 63 is an image processor for processing images of the melt surface vicinity picked up by the image sensors 20 in a digital manner, 64 is an arithmetic unit for arithmetically and logically processing the image data from the image processing unit 63 and determining a condition of the formed slag beard 60, and 65 is a display unit for displaying a detected result for the condition of slag rim formation. The image sensor 20 may be of any desired type, for example, a camera utilizing a vidicon usually employed in a video camera, as long as it can pick up the melt surface vicinity in the form of an image, that is, a brightness signal. Based on experience and various studies, however, the inventors have confirmed that those cameras utilizing coupled charged devices (CCD), such as, CCD cameras, are excellent in responsity and less affected by residual images, and hence that they are optimum in effectively presenting functions necessary for the present invention. Further, since the pouring nozzle (or dipped nozzle) 52 is disposed at the center of the mold 1, and the tundish 19, the sliding nozzle 51 and so forth are disposed above the mold 1, a space around the image sensors 20 is very small and narrow. It is therefore usual that the image sensors 20 have difficulties in their arrangement remote from the mold 1, and must be installed at positions near the melt surface. This sometimes restricts the total viewing field of the image sensors 20 dependent on, for example, their types and characteristics. In such cases, the image sensor 20 may be installed in two or more pairs in locations opposite to each other with the pouring nozzle 52 located therebetween. In some cases, however, only one pair of image sensors 20 can be installed due to restrictive conditions such as the equipment positions around the mold, as well as structure and size of a slag beard removing device (later described). In these cases, a required viewing field can be obtained by using a wide-angle optical lens for the image sensor 20. A melt surface condition detected by the image sensors 20 is periodically input as image signals to the image processor 63. The image processor 63 subjects the image input signals from the image sensors 20 to digital processing described below, thereby periodically determining distribution of brightness along the check line L set so as to intersect the mold wall surface 2. More specifically, the image processor 63 first determines the profile of the mold, 1, i.e., the mold wall surface 2 appeared in the images represented by the image signals, and then detects (derives) the position of the mold wall surface 2 on the images. Taking into account that brightness is high in the upper surface 2a of the mold and abruptly reduced at the mold wall surface 2 contacting the slag beard 60 or the non-molten powder 8 as shown in FIG. 30, the mold wall surface 2 is determined by detecting the minimum value A or a bent point C in the distribution of brightness. Since the position of the mold wall surface 2 is of course constant under the same equipment conditions, that position detected or set through the first processing can be used continuously. There is thus no need for determining the position of the mold wall surface many times. After detecting or setting the position of the mold wall surface 2, lines intersecting the mold wall surface 2 (hereinafter referred to as check lines L) are then defined as shown in FIG. 32. The check lines L are set such that they intersect the mold wall surface 2 around the periphery of the mold with a predetermined spacing therebetween, and each have a length extending to a position nearer to the center of the mold 1 than the position at which the slag beard 60 is expected to form. The check lines L are preferably set perpendicular to the mold wall surface 2, but may be deviated to some degree from perpendicular relation in a strict sense. This is because when the mold 1 having a rectangular shape is picked up from above by the image sensors 20 utilizing wide-angle lenses, the resulting image of the mold is distorted as shown in FIG. 32. Specifically, the region near the center of each image sensor 20 is picked up in the form relatively analogous to the real image, while the picked-up image is distorted in the region remote from the center such that linear lines are shifted or swollen outwardly. Accordingly, it is difficult to draw the check lines L precisely perpendicular to the mold wall surface 2 based on the image signal from the image sensor 20. However, such distortion of the image can be corrected through geometrical calculations. In this example, therefore, the check lines L are drawn on the image screen picked up by the image sensors 20, as shown in FIG. 32, and an actual condition of slag rim formation is determined after being corrected through geometrical calculations. An example of FIG. 32 shows a case of casting of 250 mm.times.1350 mm in size, in which check lines L are set for each of the image sensors 20. In total, 44 check lines L are set for both the image sensors 20 and illustrated in a single composite image. Here, each of the check lines L comprises 55 pixels. In addition, no check lines L are set in the vicinity of the pouring nozzle 52 because the slag rim 60 is less liable to form. Depending on the width or thickness of the mold 1 changed on demand, the number of the check lines L may be increased or decreased. Then, the image signals representing a melt surface condition and input from time to time are converted to distribution of brightness along each of the check lines L. The distribution of brightness thus taken at some moment along one of the check lines L is the solid lines X shown in FIG. 30, which corresponds to the distribution of brightness along the check line L1 in FIG. 32. The time-dependent distributions of brightness determined by the image processor 63 along the respective check lines L are input to the arithmetic unit 64 which arithmetically and logically processes the input signals to judge a condition of the formed slag beard 64. More specifically, the arithmetic unit 64 first detects the minimum value A, the bent point C and the maximum value B mentioned above. In other words, it determines the presence or absence of the minimum value A and the maximum value B and, if present, then calculates the maximum-to-minimum distance therebetween. The position where the slag beard 60 is adhering to the mold wall surface 2, that is, the minimum value A, can be determined by, for example, scanning the distribution of brightness shown in FIG. 30 from the left, that is, the side adjacent the mold upper surface 2a, to the right (or |