JP2763254B2 - Press control system for refractory production - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control system for a refractory production press capable of efficiently producing a raw angle (semi-finished refractory product after molding and before firing) having a stable quality.

[0002]

2. Description of the Related Art In general, a raw angle of a refractory brick or the like is manufactured by sufficiently kneading kneaded soil as a raw material, measuring the weight or volume, and then press-molding. Conventionally, such control of the angle of the raw angle is performed by manually measuring and changing the molding conditions. However, this method requires labor and requires different controls depending on the person.Therefore, the control is limited to control using values obtained only within the press, such as the upper and lower plunger positions of the press and the pressing force. I was With such a control method, the dimensions and weight of the raw angle cannot be sufficiently controlled due to the influence of springback or the like. Feedback control is performed to adjust the amount of soil input manually or automatically.

[0003]

However, if the formed raw angle is inspected and the feedback control is performed based on the inspection result, a time lag occurs between the molding and the inspection, so that accurate control cannot always be performed. . As described above, since the conventional control device cannot know the nature of the raw angle until the end of the inspection, it forms one piece and waits for an inspection result to be fed back to the next molding. It is continuously formed and fed back using the inspection result. Once molding starts, it is molded regardless of whether it is defective or non-defective. However, in this case, the cycle time is long, and in the case, the inspection result of the raw angle formed immediately before cannot be used, so it is not possible to cope with changes such as changing batches of kneaded soil during that time, and defective products However, it is not possible to change the setting of the press during molding to obtain a good product. The present invention has been made to solve the above problems, and enables feedback control at an early stage without waiting for an inspection result, and feedforwards so that a raw angle during molding, which was conventionally impossible, is not made defective. It is an object of the present invention to provide a control system for a refractory manufacturing press capable of performing control.

[0004]

The control system for a refractory manufacturing press according to the present invention comprises a kneaded material input means and a plurality of pressurizing steps of applying the input kneaded material at different molding pressures. Forming means for pressure forming, inspection means for inspecting the raw angle formed by pressure forming, kneading means based on the inspection results and learning data, and forming means for obtaining the first pressurized data from the forming means. And a control means for performing feedback control on the control signal. Further, the present invention provides a kneaded material input means, a forming means for press-forming the injected kneaded material, an inspection means for inspecting the pressed and formed raw angle, and a kneaded material based on the inspection result and learning data. Control means for controlling the input means and the shaping means, and the control means for changing the weighting of the control based on the prediction and the weighting of the control based on the inspection result based on the learning progress obtained from the difference between the prediction result and the inspection result It is characterized by having. Further, the present invention is characterized in that the control means predicts a thickness, a head, a weight, and a bulk specific gravity at the time of completing the forming of the raw angle during the forming, and performs feedforward control. Further, the present invention is characterized in that the control means learns necessary control data from control parameters of the forming means and inspection results of raw angles.

[0005]

According to the present invention, artificial intelligence (AI) is applied to a press for manufacturing raw angle.
Applying a computer that can use the language for the application, predicting the thickness, bulk specific gravity, head, etc. of the final raw angle while forming the raw angle with the AI system, and performing predictive control on the raw angle, or at the latest at the next time By feeding back to the raw angle, it is not necessary to wait for the inspection result, it is possible to perform feedback control at an early stage, and since it is possible to control the raw angle during molding, which was not possible in the past, to make it a defective product, The processing yield can be greatly improved.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The system configuration of the present invention will be described with reference to FIG. The kneaded material supply device / press / inspection / loading device 1 in the present system is controlled by a sequencer 2 controlled by a line computer 3 and executes a kneaded material supply, pressure molding, inspection and the like in a predetermined sequence. The AI computer 4 is connected to the sequencer 2 by Ethernet.
, And compares the predicted value with the test result, learns and stores it, and performs prediction control and feedback control based on the learned data, as described later. The reason why the AI computer 4 is not connected to the line computer 3 is to prevent a problem in its responsiveness.

The AI computer used in the present system employs an engineering workstation (EWS) for the following reasons. -It has a multi-user multi-task type OS. -It has an X window that can manage and control a plurality of windows in the display.・ Abundant programming languages can be used.

[0008] For communication, Ethernet which satisfies the following conditions is adopted.・ High-speed communication is possible.・ Imozure connection is possible. -It is expected that the communication will become the mainstream in the future. The programming languages use OPS83 and C. The OPS83 expert system can be easily assembled.・ Good affinity with C language.・ High speed with compiler type. C. You can build a detailed program.・ High speed with compiler type.・ It is becoming the mainstream of programming languages.

FIG. 2 is a diagram for explaining the control of the present invention. First, after the mixed soil is charged and deaerated, the molding process is started a plurality of times. The molding step usually comprises a plurality of pressure treatments. The pressure-formed raw angle is inspected for dimensions, weight and the like in an inspection process. In the present invention, inspection results and learning data based on technical discoveries, experiences, etc. are accumulated, and based on the results, the size, weight, etc., at the end of forming of the raw angle, are determined based on the results of the pressurizing step 1. Then, based on the result of the prediction, prediction control for the subsequent pressurization is performed, and feedback control is performed for the kneading of the kneaded soil so that the raw angle during the forming is prevented from becoming defective as much as possible.

Next, an apparatus configuration used in the control system of FIG. 2 will be described with reference to FIG. FIG. 3 is a diagram illustrating the dough charging in the case of the weight weighing method. The kneaded material supplied from the kneaded material replenishing device 10 is conveyed by a conveyor 11, weighed in a lightweight container 12, and a predetermined amount of kneaded material is put into a hopper 13. The hopper 13 is moved in the horizontal direction by the charger 15 which is moved up and down by the elevating table 16, and the clay is filled in the metal frame 20. In addition,
A stirrer 14 is provided in the hopper 13, and the kneaded material is sufficiently stirred before being filled in the metal frame.

FIG. 4 is a diagram for explaining the volumetric dosing of the volumetric weighing method. In the case of the volume weighing method, an amount corresponding to several moldings is charged from the container 12 to the charger box 17. The bottom surface of the charger box 17 can be opened and is moved in the horizontal direction by the charger 15. Therefore, when the charger box 17 faces the depression of the metal frame 20, the clay is put into the metal frame 20, and the relative movement between the charger box 17 and the metal frame 20 causes the kneading of a fixed volume corresponding to the volume of the metal frame. The soil is filled.

FIG. 5 is a view for explaining a press apparatus for forming the kneaded material filled in the metal frame. The press machine 30 lowers the upper plunger 34 fixed to the pressure block 33 by the hydraulic cylinder 32 pressurized from above the metal frame 20 through the pressure load cell 31, and the lower plunger 36 facing the metal frame 20. The compact is formed by pressurizing the kneaded material. The position of the upper plunger 34 is detected by the rotary encoder 40, and the surface pressure distribution is measured by the pressure sensor 41. The position of the lower plunger 36 is adjusted by the floating pressure load cell 35 so that the kneading depth of the metal frame can be adjusted. The position of the lower plunger is detected by the rotary encoder 42.

FIG. 6 is a diagram for explaining a live angle test. The brick formed after the pressing step is set in the inspection device 50 by the robot 54. The inspection device 50 includes a dimension measuring station 53, a weight measuring station 52, and a frame stacking station 51.
The weight is measured, loaded on an iron pallet (palletizing), and the process proceeds to the next drying step.

Next, a situation in which a defective product is likely to occur during the molding period will be described. At the start of molding, the correct molding result was not obtained due to unstable period or human error such as incorrect setting of technical data or defective frame changing work due to changes in the properties of the clay and the setting of the equipment itself. It may not be obtained, and for example, as shown in FIG. 7A, the brick thickness varies depending on the brick. Further, even after the molding result is stabilized, if the molding is stopped for a while, for example, by a break of the operator, the molding result is likely to be out of order when the molding is resumed. For example, as shown in FIG. This is considered to be due to the fact that the temperature of the clay decreased during the stop and the properties of the clay changed. Also, in the case of the final consolidation, the level of the consolidation in the charger box changes, unlike the usual case. For this reason, even if the kneaded material is supplied so as to have the same kneaded material depth, the packing density is different, and consequently the supply amount of the kneaded material fluctuates. Assuming that the packing density when the level is reduced is D ', D>D'. In addition, during continuous molding, the properties of the kneaded material may change suddenly due to a change in the batch of the kneaded material, or the occurrence rate of defective products may increase suddenly due to other causes. The change of the batch of the kneaded soil means that when kneading the necessary kneaded clay with a mixer, the amount that can be kneaded at a time is determined by the capacity of the mixer. When it is necessary, the properties of the two clays are not exactly the same, so that the properties of the clay may change suddenly at the time of this switching, and the thickness of the brick varies as shown in FIG. However, the time of occurrence cannot be predicted exactly.

Examples of data used in the present invention are shown in, for example, Table 1.
It is as shown in. In the example of Table 1, the positions of the upper and lower plungers are used for predicting the thickness, the head, and the bulk ratio, and the amount of the clay input is used for predicting the weight and the bulk ratio. Then, by accumulating the inspection results (relationship between the upper and lower plunger positions and the thickness, the drop, relationship between the amount of the kneaded material input and the weight, and the bulk specific gravity) as learning data, the raw angle of the currently formed raw angle is formed. It is used as data for predicting the thickness, head, weight, and bulk ratio at the end of molding.

First, the upper and lower plunger positions will be described. FIG. 8 is a diagram for explaining the relationship between the upper and lower plunger positions and the brick thickness. FIG. 8A shows the case where the distance between the upper and lower plungers is large and the brick is thick, and FIG. 8B shows the case where the distance between the upper and lower plungers is small and the brick is thin. Yes, the thickness of the brick depends on the distance between the upper and lower plungers. Therefore, the relationship between the gap between the upper plunger and the lower plunger and the thickness of the brick is accumulated as data each time each raw angle is inspected.

FIG. 9 is a diagram showing the relationship between the inclination of the plunger and the drop of the brick. As shown in FIG. 9 (a), when the upper and lower plungers are tilted to the left in the figure, and the thickness of the left end of the brick is A, and the thickness of the right end of the brick is B, the drop is AB, and FIG. 9 (b) When the brick is tilted in the opposite direction, the direction of the drop of the brick changes. Since the head changes depending on the degree of the inclination, data on this relationship is accumulated. When the amount of the clay to be put into the metal frame increases, both the weight and the thickness increase. The bulk specific gravity is represented by the following formula: bulk ratio = brick weight / brick volume. Therefore, the bulk ratio increases as the weight increases, and decreases as the thickness increases. In the weight weighing method, the amount of the kneaded soil is controlled by directly inputting the weight measured by a weighing device into a metal frame, and increasing or decreasing the weight by changing a weighing set value. In the volume weighing method, when changing the amount of the clay, as shown in FIG. 10, the depth of the metal frame is changed by moving the lower plunger 36 up and down. The volume increases, and if the depth of the metal frame is reduced, the mass will decrease. The plunger position may be fixed in order to keep the amount of kneaded clay constant.

As shown in FIG. 11, the flow of kneaded soil in volumetric weighing using a charger box is such that the surface of the lower plunger is flush with the metal frame (FIG. 11A). (Figure 11)
(B)) After lowering the lower plunger by a predetermined amount to obtain a predetermined kneading depth (FIG. 11 (c)), the charger box is slid and returned to the original position (FIG. 11 (d)). ),
Lower the lower plunger by a predetermined distance (FIG. 11).
(E)). Although not shown in Table 1, the thickness is determined by the depth of the kneaded soil, the number of times of molding, and the molding pressure. The bulk ratio can be controlled by using this together with the control of the amount of kneaded soil. The prediction is obtained by deriving a relational expression using the least squares method based on the inspection result of the raw angle formed in the past and the data obtained up to the first pressurization, and substituting the data currently being formed into the expression. Further, a relational expression is derived from the output result and the inspection result using the least squares method, and an optimum output amount is determined. Now, taking the thickness as an example, the output result is the amount (x) of putting the clay into the metal frame, the inspection result is the thickness (t) of the brick measured by the inspection device, and a number indicating the number of the brick. (N), Δt _n = t _n
Assuming that a straight line At = ax + b is obtained by the least-squares method when the data of each point shown in FIG. 12 is obtained when −t _n−1 is set, a metal frame is obtained by substituting the thickness t. The amount x of the clay to be put in is obtained.

When a relational expression is derived by using the least squares method based on the inspection result of the raw angle formed in the past and the data obtained up to the first pressurization, the intermediate results are classified into each classification item as shown in Table 2. It is classified and made into a database to make learning data. The number of moldings and the molding pressure will be described.
As shown in FIG. 3, the pressure is changed as in molding pressures 1 to 3 and pressurization is performed several times at each molding pressure. The thickness changes depending on the molding frequency and molding pressure. In FIG. 13, A
Since all the data necessary for the prediction are prepared at the point of, the calculation is started from here. In order to perform feedback control, the molding pressure is not reflected on the molding pressure unless an instruction is given to the press by the time point C. So when we start the calculation from A,
With a margin, the process ends with B, so that there is enough time for the update instruction of the set value.

The inclination of the hopper at the head and the forward / reverse rotation time of the stirrer will be described. For example, when weighing, FIG.
When the hopper is tilted as shown in (a), the level of the clay in the hopper changes in an inclined manner (FIG. 14 (b)). Since the level is inclined, a drop can be given to the brick. When the stirrer in the hopper is rotated forward with the hopper being horizontal as shown in FIG.
As shown in FIG. 4 (d), the level of the clay is inclined upward to the right, and when the stirrer is reversed, as shown in FIG. 14 (e),
Since the level of the clay is inclined to the upper left, a head can be similarly provided.

Since there is no stirrer at the time of volume weighing,
The head in the charger box is leveled by a leveling device so as to give a head by inclining the ground level. FIG. 15 (b) shows the level of the clay as shown by the broken line in FIG. 15 (a), and FIG. 15 (c) shows the level of the clay as shown by the solid line in FIG. 15 (a). is there. In the above description, the thickness, the head, the weight, and the bulk specific gravity are made into a database for each classification item and learned. However, other parameters such as a control constant of the press may be made into a database and learned. .
Also, the learning result is compared with the inspection result, and the learning progress is determined based on the error.When the learning progress is not so advanced, the conventional feedback control is performed, and when it is determined that the learning progress is advanced. Alternatively, the control may be switched to the control of the present invention. The weighting may be changed according to the learning progress. For example, assuming that the prediction result is A and the inspection result is B, if the error C = | AB− is calculated, it can be determined that the smaller C is more accurate, that is, that learning is progressing. Here, a function f (C) that gives a value in the range of 0 to 1 (however, when C1 <C0, f (C
1) Determine the output using <f (C0)). That is, assuming that the output amount obtained from the predicted value is D and the output amount obtained from the inspection result is E, the actual output amount = {1−f (C)} × D + f (C) × E.

Next, a description will be given of a control for controlling the thickness as a control amount and the kneading depth, forming pressure and the number of times of forming as an operation amount. The control method shown in FIG. 16 divides the kneading material input in FIG. 2 into the kneading material replenishment and the kneading material filling, and ff indicates feedforward and fb indicates feedback.

[Ff1] ff1 is a control used at the time of the first molding at the time of restart after stopping, in order to avoid the inconvenience described with reference to FIG. And the temperature of the compacting soil) are used to correct the compacting depth using the data obtained by learning the effect on the compression ratio. At this time, the compression ratio = the kneaded soil depth / thickness. In this case:-If the stop time is long, it becomes difficult to tighten, so the compression ratio becomes small. -If the temperature difference is large, it becomes difficult to tighten, so the compression ratio becomes small. A quadratic curve (coefficients f0, f1, f) as shown in FIG. 17 shows the relationship between (the product of the stop time and the temperature difference) and (the reduction amount of the compression ratio).
Approximate in 2). f0, f1, and f2 are coefficients (regression coefficients) of a relational expression derived from the previously formed raw angle inspection data by using the least squares method, and are updated as the inspection data is accumulated. When the product of the stop time and the temperature difference is Tt, the compression ratio decrease amount Δm is estimated using the values of the regression coefficients f0, f1, and f2 in the learning data. Δm = f0 + f1 · Tt + f2 · Tt ² At this time, the correction amount ΔD of the moulding depth is ΔD = Δm × set thickness. However, the value of the set thickness uses a value of a weighted average based on the number of outputs during continuous molding. The original set thickness is used only when molding has not begun immediately after the start of molding. When the inspection result of the brick struck at this time is obtained, the compression ratio decrease amount Δm ′ obtained from the actually measured value is added to the learning data together with Tt, and the value of the regression coefficient is updated.

[Ff2] ff2 is used at the time of molding with the final clay to avoid the inconvenience shown in FIG. 7 (c).
As shown in FIG. 18, the moulding depth is corrected using data obtained by learning the relationship between the moulding level in the charger box and the compression ratio. Compression ratio = pulverized soil depth / thickness • When the level increases, the compression ratio increases because the overtightening occurs. The relationship between the level and the increase in the compression ratio is 2 as shown in FIG.
Approximate by the following curve. When the level Lv is greater than 0, the compression ratio reduction amount Δm is calculated using the values of the regression coefficients f0, f1, and f2 in the learning data.
Is estimated. Δm = f0 + f1 · Lv + f2 · Lv 2 · Nerido depth of the correction amount [Delta] D is, [Delta] D = Delta] m × set thickness, however, the value of the set thickness using a value of the average weighted by the number of output times of the continuous molding. When the inspection result of the brick struck at this time is obtained, the compression ratio reduction amount Δ obtained from the actually measured value
m ′ is added to the learning data together with Lv, and the value of the regression coefficient is updated.

[Fb1] fb1 is used in the single action mode.
The single action mode is a mode in which the next brick is not struck until the inspection result of the struck brick is obtained. As shown in FIG. 20, when the inspection result of the molded brick is obtained,
Based on the difference from the set value, the correction amount is determined by the learning method or the on-site method, and is reflected on the next brick to be struck. Since the correction is performed for each line, the occurrence of defective products is small, but the cycle time is long. The correction amount is obtained by one of a field method and a learning method described below. If the learning data is stored in a certain amount (the number of bricks) or more, the learning method is used, and in the other cases, the learning method is used in the field method.

(A) On-site method Kneaded soil depth correction amount = thickness error × cc c is a correction constant and is selected, for example, as follows. When the error is on the positive side 2 When the error is on the negative side 3 (b) Learning method Data of a regression line f (x) = ax as shown in FIG. 21 in which the relationship between the kneaded soil depth and the thickness is learned is used. Correction amount of kneaded soil depth = thickness error × a × c a: slope of regression line c: correction constant, for example, selected as follows. When the error is on the positive side 0.67 When the error is on the negative side 1.0 Add the kneading depth (molding result) and thickness (inspection result) of the brick struck at this time to the learning data and update the regression coefficient a. I do.

[Fb2] fb2 is used in the continuous mode.
The continuous mode is a mode in which the bricks are struck one after another without waiting for the completion of the inspection of the struck bricks. As shown in FIG. 22, when the prediction result or the inspection result of the struck brick is obtained, the presence or absence of the correction is determined from the result of accumulating the error with the set value. When it is determined that there is a correction, a correction amount is obtained by a learning method or an on-site method, and is reflected on bricks to be hit later. The on-site method or the learning method is determined by the learning method if learning data is stored in a certain amount (the number of bricks) or more, and the other method is determined by the on-site method. (A) Field method It is obtained by the same method as fb1. Correction amount of soil depth = thickness error × c When the error is on the positive side 2 When the error is on the negative side 3 Thickness error: error obtained when determining the presence or absence of correction (b) Learning method The relationship is obtained by the same method as fb1, and the learning data is shared with fb1. Amount of kneaded soil correction = thickness error × a × ca a: slope of regression line c: correction constant (shared with fb1) When error is + side 0.67 When error is − side 1.0 Thickness error : Error obtained at the time of judging the presence or absence of correction The kneading depth (molding result) and thickness (inspection result) of the brick struck at this time are added to the learning data, and the regression coefficient is updated.

[Ff3] Pressing is performed in a process as shown in FIG. 13, and the molding pressure and the number of times of pressing are corrected using data obtained by learning the relationship between the thickness at the time of initial pressing and the final thickness of the brick. I do.・ When the initial pressurization result is obtained, the correction amount is obtained and output. The output correction amount is to be reflected in the stage after the molding pressure 2 of the brick being struck. The correction amount is determined by determining the molding pressure 2 and the number of additional pressurizations according to the following rules. If the estimated thickness is large, increase the molding pressure 2 by 10% and re-estimate the thickness If the estimated thickness is good (no correction), if the estimated thickness is large, increase the molding pressure 2 by 10% Increase the number of times of pressurization by +1. Molding pressure 2 if the estimated thickness is good
By 10%. If the estimated thickness is small, the molding pressure 2 is corrected by obtaining the pressure by interpolation.

In obtaining the estimated thickness, the relationship between the initial pressurized thickness and the final thickness is approximated by a straight line. When the initial pressure thickness is h,
Using the regression coefficients f0 and f1 in the learning data, the final thickness H
Is estimated. H = f0 + f1 · h The method of obtaining the thickness re-estimated value estimates the final thickness using learning data under the same conditions as the pressure when the molding pressure is increased by 10%. When there is no learning data of the same condition, learning data of a similar condition is used, and when there is no learning data, calculation is performed using a default constant. The learning data is classified according to the product name, the quality name, the molding pressure, and the number of times of pressurization. When the inspection result of the brick struck at this time is obtained, the actually measured final thickness value H ′ is obtained.
Is added to the learning data together with h, and the value of the regression coefficient is updated. When the molding pressure and the number of times of pressurization are corrected, the learning data under the condition equal to the actual output (obtained as a molding result) is updated. However, in order to prevent the learning data from becoming unnecessarily large, the correction of the molding pressure is performed only when the pressure is 10% up.

[0030]

As described above, according to the present invention, by predicting the inspection result, it is possible to grasp what raw angle is formed without waiting for the inspection result.
The following effects are achieved. There is no need to wait for test results, and feedback can be provided at an earlier stage. Feedforward control can be performed so that the raw angle during molding, which was not possible in the past, is not made defective. As a result, although the processing yield was about 80%, 9
The percentage of non-defective products fell from 90% to 95%.

[Brief description of the drawings]

FIG. 1 is a diagram showing a system configuration of the present invention.

FIG. 2 is a diagram for explaining control according to the present invention.

FIG. 3 is a diagram for explaining the injection of kneaded soil in the case of a weighing method.

FIG. 4 is a view for explaining the volumetric dosing of the dough.

FIG. 5 is a view for explaining a press device for forming clay into a metal frame.

FIG. 6 is a diagram illustrating product inspection.

FIG. 7 is a diagram illustrating a molding phase in which a defective product occurs.

FIG. 8 is a diagram illustrating the relationship between the upper and lower plunger positions and thickness.

FIG. 9 is a diagram showing the relationship between the inclination of the plunger and the drop of the brick. FIG.

FIG. 10 is a diagram for explaining a method of changing the amount of clay in the volume weighing method.

FIG. 11 is a diagram for explaining a flow of pulverizing soil input in a volume weighing method.

FIG. 12 is a diagram illustrating a method for obtaining a relational expression by the least square method.

FIG. 13 is a diagram illustrating a molding step.

FIG. 14 is a diagram illustrating how to give a head when weighing.

FIG. 15 is a diagram for explaining how to give a head during volume weighing.

FIG. 16 is a diagram showing a control method using thickness as an example.

FIG. 17 is a diagram illustrating control at the time of restart after stopping.

FIG. 18 is a diagram illustrating control during final kneading.

FIG. 19 is a diagram showing the relationship between the level in the charger box and the amount of increase in the compression ratio.

FIG. 20 is a diagram illustrating control in a single action mode.

FIG. 21 is a diagram showing the relationship between thickness and kneading soil depth.

FIG. 22 is a diagram illustrating control in a continuous mode.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Matter supply device, press and inspection loading device, 2 ... Sequencer, 3 ... Line computer, 4 ... AI computer, 10 ... Mould material supply device, 11 ... Conveyor, 12 ... Measurement container, 13 ... Hopper, 14 ... Agitator , 15 ... charger, 16 ... elevator, 17 ... charger box, 2
0: gold frame, 30: press machine, 34: upper plunger, 3
6 lower plunger, 50 inspection device.

──────────────────────────────────────────────────続 き Continued on front page (58) Field surveyed (Int.Cl. ⁶ , DB name) B28B 3/02

Claims (4)

(57) [Claims] 1. A method for feeding a kneaded material , comprising:
Having a plurality of pressurizing steps with a forming pressure of the following, forming means for press forming, inspecting means for inspecting the raw angle formed by pressing, and clay injection means based on the inspection result and learning data, Pressing the molding means to obtain the first data
Control system for a refractory manufacturing press, comprising a control means for performing feedback control by means of a press. 2. A kneading material input means, a forming means for press-forming the injected kneaded material, an inspection means for inspecting the pressed and formed raw angle, and a kneading material input based on the inspection result and learning data. Controlling the means, the forming means and the control means
Prediction based on the learning progress obtained from the difference between
The weight of control based on inspection and control based on inspection results
A control system for a refractory manufacturing press, comprising the control means described above . 3. An apparatus according to claim 1 and / or 2, wherein the system, said control means, forming at the end of the thickness of the raw angular, drop, weight, controlling feedforward predict the bulk density during the molding A press control system for refractory manufacturing, characterized by the following. 4. The method of claim 1 and / or 2, wherein the system, said control means, the control parameter of the forming means,
A control system for a press for manufacturing refractories, characterized by learning necessary control data from inspection results of raw angles.