JP3579553B2 - Method and apparatus for diagnosing abnormal state of equipment and product process - Google Patents

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a method and an apparatus for diagnosing an abnormal state of equipment and a product process in a rolling equipment or the like, and in particular, to clarify a causal relationship between an abnormal cause and an abnormal state.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, various methods and apparatuses for diagnosing an abnormal state of equipment and product processes have been proposed.
[0003]
As a conventional method of diagnosing an abnormal state of a product process, a method of performing a linear multivariate analysis is known.
[0004]
JP-A-6-74876 discloses a facility diagnosis method and apparatus for diagnosing the state of a target facility online, by inputting observation information to a neural network, finding a candidate for a cause using the neural network, Use weights to narrow down probable causes and then use neural networks to determine the effects caused by potential causes. It is disclosed that the cause of the problem is determined, the presence or absence of the effect and the degree of the effect are confirmed in actual equipment, and the true cause is specified.
[0005]
Specifically, this equipment diagnosis method inputs a sensor or a driver's observation information that measures the state of the target equipment into a neural network representing a causal relationship between the state of the target equipment and a cause that causes the state, Following the corresponding path of the neural network, a candidate for causing the state is determined, and a state exerted by the candidate for the cause of the state is determined by tracing the corresponding path of the neural network in the reverse direction, and is determined by the neural network. It is designed to determine the true cause by checking the state that was given.
[0006]
In addition, an apparatus for performing such a facility diagnosis method uses a sensor or a driver's observation information that measures the state of a target facility and learning information from a learning device to cause the state of the target facility and the state thereof. An input device that inputs to the neural network representing the causal relationship with the cause, the neural network representing the causal relationship between the state of the target facility and the cause that causes the state, and the causal relationship between the target facility and the cause that causes the state Storage device that stores the neural network structure and neuron weights caused by the probability of occurrence of the condition and the cause of the state, and the probability of occurrence of the causal relationship stored in the storage device and the degree of the effect of the cause candidate Is used as the weight of the neuron, so that it follows the path of the neural network in the forward and reverse directions. Control device, an estimating device for estimating the true cause from the state obtained by the neural network, and a sensor or sensor for measuring the state of the target equipment or the observation information of the driver and the neural network by inputting the estimation result. It is composed of a learning device that learns the configuration of the network and the weights of the neurons, and a display device that displays the progress of the inference and its result.
[0007]
[Problems to be solved by the invention]
However, the above-described conventional equipment diagnosis method and equipment diagnosis apparatus have the following problems.
[0008]
In other words, the method for diagnosing an abnormal state of a product process by performing a linear multivariate analysis has a problem that it is difficult to handle a nonlinear process phenomenon, and the abnormal state cannot be quantitatively formulated. .
[0009]
Further, in the equipment diagnosis method and apparatus disclosed in Japanese Patent Application Laid-Open No. 6-74876 for diagnosing the state of a target equipment online, it is described that the cause of the abnormality is analyzed by using a neural network to analyze the cause of the abnormality. It does not show how to specifically derive a causal relationship with the state. In other words, just by identifying the cause of the abnormality using the neural network, the causal relationship is black-boxed, and the process of finding the explanatory variable, which is an unknown variable, from the answer, which is the objective variable, is unknown. is there.
[0010]
The present invention has been proposed in view of the above-mentioned problems of the related art, and has the following objects.
[0011]
The invention according to claim 1 and claim 4 has an algorithm that can handle nonlinear process phenomena, is easy to handle regardless of the shape of the function, is not easily affected by disturbance, and has an algorithm that avoids a local solution. It is an object of the present invention to provide equipment and a method and apparatus for diagnosing an abnormal state of a product process that can easily perform optimization.
[0012]
According to the invention of claim 2, in addition to the object of the invention of claim 1, the efficiency of abnormal cause analysis is improved, and a causal relationship is represented by an expression, thereby preventing the neural network from becoming a black box, thereby simplifying the process. Another object of the present invention is to provide a method of diagnosing an abnormal state of a product process, which can confirm the reliability of identifying an abnormal factor in light of a physical phenomenon.
[0013]
A third object of the present invention is to provide a method of diagnosing an abnormal state of a product process which can more quickly find a solution, in addition to the object of the first or second embodiment of the present invention.
[0014]
According to a fifth aspect of the present invention, in addition to the object of the fourth aspect of the present invention, there is provided a diagnostic apparatus for a product process abnormal state capable of issuing an alarm and notifying the abnormal state when an abnormal state is found. The purpose is to do.
[0015]
A sixth aspect of the present invention, in addition to the object of the fourth or fifth aspect of the present invention, aims to provide an apparatus for diagnosing an abnormal state of a product process which can more quickly find a solution.
[0016]
[Means for Solving the Problems]
The present invention has been made to achieve the above-mentioned object, and its contents will be described below.
[0017]
The invention according to claim 1 isHandling nonlinear process phenomenaIn the method of diagnosing the abnormal state of the target equipment and the product process online or offline, first, when the measured abnormal state exceeds a predetermined threshold set to a level that does not cause a problem of the equipment abnormality or the quality abnormality, the method is performed. A state explanatory variable related to the abnormality and an abnormal target variable are obtained.
Next, one or more abnormal factor analysis is performed by the neural network to identify the influence of the state explanatory variable on the state objective variable.
Then, the state explanatory variable to be corrected to keep the abnormal objective variable within a predetermined threshold valuepluralThe solution is automatically derived by a combinatorial optimization method.
[0018]
Specifically, using the present invention, for example, an abnormal state of meandering of a steel sheet in a winding process is diagnosed for a winding facility of a hot rolling plant as a target facility.
In addition, the above-mentioned state explanatory variables are variables relating to an abnormality factor of the equipment, and specifically, are variables relating to a gap of a pinch roll, a load of the pinch roll in the winding equipment, and the like.
The above-mentioned state objective variable is a variable relating to the operating state of the equipment, and specifically, is a variable relating to the meandering amount of the steel sheet in the winding equipment. Among the state objective variables, a variable indicating an abnormal state is an abnormal objective variable.
[0019]
According to a second aspect of the present invention, in addition to the configuration of the first aspect of the present invention, in the method of expressing a causal relationship in the abnormal factor analysis step, the factor corresponding to the neuron in the neural network is equal to or less than a predetermined weight. In, the factors corresponding to the neuron are excluded from the state explanatory variables, and only the factors necessary for the judgment are selected.
Also, the evaluation function DI is
DI = 1 / Σ | (Neural network output)-(Self-organizing output) |²
Formulation by self-organization is performed by the genetic algorithm expressed by
[0020]
According to a third aspect of the present invention, in addition to the configuration of the first or second aspect of the present invention, the combination optimization method uses a genetic algorithm.
[0021]
The invention according to claim 4 is the invention according to claims 1 to 3.Handling nonlinear process phenomenaA diagnostic apparatus for implementing a method for diagnosing equipment and product process abnormalities, comprising: a data collecting means for collecting data necessary for the diagnosis; and converting the collected data into a state explanatory variable and an abnormal target variable, respectively. Calculating means for comparing with a predetermined threshold value, storing parameters, and, when an abnormal sign appears, creating a neural network from the abnormal sign parameter and the abnormal cause parameter to identify the phenomenon, and based on the identified neural network. A formulating means for selecting a necessary number of abnormal factor parameters and formulating by self-organization using a genetic algorithm, and an abnormal factor necessary for suppressing an abnormal sign within a predetermined threshold.Multiple solutionsAn arithmetic processing means for determining the selection by combinational optimization, a storage means for storing data in the data collection means, the arithmetic means, the formulating means, the arithmetic processing means, and a display means for displaying a calculation process and a diagnosis result. Have. Here, the reason why the neural network is used to identify the phenomenon and the abnormal factor parameters are selected is to increase the efficiency of the calculation, and it is also possible to apply the genetic algorithm directly and simultaneously execute the formulating and identification. It is possible.
[0022]
Specifically, the data used for the diagnosis is, for example, input from a keyboard by an operator or input by a sensor provided at a predetermined location of the facility.
Further, the data collection device, the arithmetic means, the formulating means, and the arithmetic processing means are constituted by, for example, a computer having a CPU and an auxiliary device, and the computer operates according to a preset program, whereby the computer Function as
Further, the storage means is constituted by, for example, a RAM, a hard disk storage device, or another magnetic or optical storage device.
The display means is constituted by, for example, a CRT display device.
[0023]
According to a fifth aspect of the present invention, in addition to the configuration of the fourth aspect of the present invention, there is provided an alarm means for generating an alarm when an abnormal state occurs.
Specifically, the alarm means is constituted by, for example, an amplifier and a speaker that generate an alarm sound, and a lamp that generates an alarm by emitting light.
[0024]
According to a sixth aspect of the present invention, in addition to the configuration of the fourth or fifth aspect, the arithmetic processing means selects an abnormal factor by combination optimization using a genetic algorithm.
[0025]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an example of an embodiment of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic configuration of a diagnostic device according to the present invention, and FIG. 2 is an explanatory diagram showing a schematic configuration of a winding facility of a hot rolling plant to which the diagnostic device is applied.
[0026]
The diagnostic apparatus 1 according to the present invention is used, for example, by attaching it to a winding facility 2 of a hot rolling mill as shown in FIG. The winding equipment 2 detects a meandering of the steel sheet 3 generated in the step of winding the rolled steel sheet 3 into a coil shape as an abnormal state, and displays a necessary state description and a correction amount thereof. .
[0027]
As shown in FIG. 2, a winding apparatus 2 to which the diagnostic device 1 is applied includes a guide for guiding the steel sheet 3 on both sides in the width direction of the steel sheet 3 rolled and sent out by a finishing mill (not shown). 4 and 4 are provided, and two pairs of upper and lower pinch rolls 5 are provided downstream of the guides 4 and 4 for sandwiching and sending out the steel plate 3 from above and below, and a winding core is provided downstream of the pinch rolls 5. A mandrel 6 is provided, and four wrapper rolls 7 serving as auxiliary rolls for winding up the steel plate 3 on the mandrel 6 are arranged on the outer periphery of the mandrel 6.
[0028]
The above-mentioned pinch rolls 5 are connected to a pinch roll driving device 8 such as a motor and are driven to rotate.
The above-described mandrel 6 is connected to a mandrel driving device 10 such as a motor via a reduction gear 9 and is driven to rotate.
[0029]
As shown in FIG. 1, the diagnostic device 1 includes an input device 11 such as a keyboard operated by an operator, a sensor 12 attached to the pinch rolls 5 to measure the amount of meandering of the steel plate 3, and a diagnostic device necessary for diagnosis. A data collection device 13 for collecting data, a data processing device 14 for processing and processing the collected data, an identification calculation device 15 for creating a neural network and performing an identification calculation of an abnormal factor by a genetic algorithm, and a data collection device 13, a data processing operation device 14, a storage device 16 for storing data handled by the identification operation device 15, a display device 17 for displaying a calculation process and a diagnosis result, and an alarm for generating an alarm when an abnormal state occurs. Device 18.
[0030]
The input device 11 composed of a keyboard or the like operated by the operator and the sensor 12 function as an input device for data collection means. In this case, the input from the operator is performed offline, and the input from the sensor 12 is performed online.
[0031]
Further, the data collection device 13, the data processing operation device 14, and the identification operation device 15 are constituted by, for example, a computer having a CPU and an accessory device, and the computer operates according to a preset program. Functions as a device.
[0032]
Among these devices, the data collection device 13 functions as data collection means for collecting data necessary for diagnosis. Further, the data processing arithmetic unit 14 and the identification arithmetic unit 15 convert the collected data into a state explanatory variable and an abnormal target variable, respectively, and compare the abnormal sign with a predetermined threshold value, and store the parameters to store the abnormal sign. Appears, a neural network is created from the abnormal sign parameter and the abnormal factor parameter to identify the phenomenon, and the required number of abnormal factor parameters are selected based on the identified neural network, and the genetic algorithm is executed. The formulating means for formulating by self-organization using the algorithm and the calculation processing means for selecting the cause of the abnormality necessary to suppress the sign of abnormality within a predetermined threshold by combining and optimizing using a genetic algorithm. Function.
[0033]
The storage device 16 includes, for example, a RAM, a hard disk storage device, and other magnetic or optical storage devices, and functions as a storage unit that stores a series of data.
[0034]
The display means is constituted by, for example, a CRT display device or the like, and functions as a display means for displaying a calculation process and a diagnosis result.
[0035]
The alarm device 18 includes, for example, an amplifier and a speaker that generate an alarm sound, and a lamp that generates an alarm by emitting light, and functions as an alarm unit that generates an alarm when an abnormal state occurs.
[0036]
The neural network used in the above-described diagnostic device 1 is a system in which a plurality of relatively simple elements are interconnected to perform information processing by exchanging simple signals. A technology that mimics the structure and function of a computer with hardware and software, and aims to realize advanced information processing similar to that performed by the human brain. This neural network has a learning function, and even when it is difficult to find a clear relationship between input data and output data, if the system learns the combination of input data and output data, the system Rules can be created.
[0037]
The genetic algorithm used in the above-described diagnostic apparatus 1 is one of optimization techniques that mathematically imitates that an organism is evolving to increase its adaptability while repeating selection, gene crossover or mutation. is there. Conventional techniques for solving an optimization problem incorporating a genetic algorithm are described in, for example, "Genetic Algorithms Search Optimization, and Machine Learning" (David E. Goldberg Addison Wesley) or "Japan Society of Research Management". , Papers "(June 7, 1995, published by the Japan Institute of Equipment Management).
[0038]
The outline is to set an initial value for a variable that determines a system to be subjected to mathematical optimization, and to improve the variable while judging the quality of an evaluation function to be optimized. In the genetic algorithm, variables that determine a system to be optimized are represented by a sequence (a sequence is a gene, and a part of the sequence is a chromosome), and the system determined by this is a group of individuals ( Generation), evaluate fitness from the evaluation function, reduce individuals with low fitness, increase individuals with high fitness by that amount (selection), and then establish several sets of individuals from the population. And then perform an operation (crossover) to replace a part of the sequence corresponding to each of the two individuals (crossover), or select some individuals stochastically from the population and then select the sequence corresponding to each individual Perform an operation (mutation) to replace a part with another numerical value.
[0039]
As these operations are continued to update the population (change of generations), the proportion of individuals with high fitness in the population gradually increases. Therefore, an optimal solution can be obtained using the case where the ratio of individuals with high fitness exceeds a preset limit ratio as a condition for determining the end of calculation. This method can discretely define variables and evaluation functions to be optimized, eliminates the need for sensitivity analysis, and increases the possibility of reaching global optimal solutions for problems with many local optimal solutions. The reliability of optimization is high.
[0040]
Next, a procedure of a diagnostic method using the diagnostic device 1 described above will be described with reference to FIG. FIG. 3 is a flowchart illustrating the outline of the procedure of the diagnostic method according to the present invention.
[0041]
As shown in FIG. 3, when the diagnosis process is started, first, the state objective variable is measured offline by an input from an operator or online by an input from a sensor 12 (S1). Specifically, the state objective variable is a meandering amount of the steel plate 3 representing an abnormal state. That is, the meandering of the steel sheet 3 leads to an abnormal state of defective winding shape, and an abnormal product process that reduces the yield.
[0042]
Next, a determination is made by comparing the threshold value with the state objective variable (the meandering amount) (S2). Here, if the state objective variable (the meandering amount) does not exceed the threshold value, it is determined that the state is the normal state, and the process ends.
[0043]
On the other hand, when the state objective variable (the meandering amount) exceeds the threshold, the state explanatory variable of the state objective variable (the meandering amount) is measured offline or online (S3). Specifically, the state explanatory variables include the pressing force of the pinch roll 5, the gap difference between the left and right of the pinch roll 5, the number of rotations of the pinch roll 5, the lead rate of the pinch roll 5, the pressing force of each wrapper roll 7, The difference between the left and right gaps of each wrapper roll 7, the number of rotations of each wrapper roll 7, the lead rate of each wrapper roll 7, the gap of the guide 4, the plate width, the plate thickness, and the elapsed time of each roll biting.
[0044]
Next, an abnormality factor analysis and state variables are narrowed down by a neural network (S4).
Specifically, a threshold value is set in the neural network by narrowing down the state explanatory variables, and for those having a low influence coefficient of the input layer, the pressing force of the pinch roll 5 and the left and right of the pinch roll 5 are deleted as a result of deletion. , Except for the gap difference, plate width and elapsed time of roll biting.
[0045]
Next, formulating by a genetic algorithm (DI = 1 / Σ | (neural network output) − (self-organizing formula output) |²Is performed (S5).
When the identification and formula are directly executed, DI = 1 / Σ | (actual output) − (self-organized formula output) |²And
[0046]
Next, the result of the formula and the qualitative evaluation of the physical phenomenon are performed (S6), the cause of the abnormal state is specified (S7), and the state to be corrected necessary to suppress the state objective variable within the threshold is explained. Variables are derived by combination optimization using a genetic algorithm (S8).
[0047]
Next, the analysis result and the calculation process are displayed on the display device 17 (S9), and a warning of an abnormal state is issued by the alarm device 18 (S10), and the process is terminated.
[0048]
In addition, when the results of the above formulating and the qualitative evaluation of the physical phenomena are performed (S6), if the results of the formulating are not mature enough to identify the cause, the measurement of the state explanatory variables is performed. (S3 to S6) are repeated.
[0049]
The details of the above processing will be described in more detail.
First, based on FIG. 4, the result of studying the effectiveness of the abnormal factor analysis (identification) (S4) using the neural network will be described using the data measured in the measurement of the state explanatory variables (S3).
FIG. 4 is a graph showing the results of studying the effectiveness of process identification by a neural network. The vertical axis represents the amount of meandering of the output of the neural network, and the horizontal axis represents the measured value of the meandering angle of the steel plate 3.
[0050]
No. Using the data measured from the pinch roll 5 of No. 1 and comparing the abnormal factor analysis (identification) result by the neural network and the actual measurement value of the meandering angle of the steel plate 3, the correlation coefficient was 99 as shown in FIG. %, Showing a very good correlation, indicating that the identification by the neural network is effective.
[0051]
Next, the procedure of formulating (S5) using the above-described genetic algorithm will be described in more detail.
Formulation by a genetic algorithm aims to express unknown phenomena (abnormalities) in formulas and to generalize causality.
The procedure of formulating by the genetic algorithm will be described step by step.
[0052]
1) Input the target variables and explanatory variables on the spreadsheet software, and create a text file by using a macro program. Here, the basic statistic of each variable is calculated, and data other than the parameters of each Pn average value ± 2.5 × standard deviation> Pn are extracted, and the data number and the data content are displayed on the display device 17. Judgment of adoption or rejection.
[0053]
2) Input various specifications of the operation of the genetic algorithm using an editor. The various specifications to be input are, for example, one or more intersections, the initial population number, the mutation rate, forbidden items (restrictions, termination conditions), and the like.
[0054]
3) Execute the operation of the genetic algorithm.
i) Generate an initial population. In the creation of a tree in the generation of the initial group, enter the four arithmetic operators or the n-th multiplier at the root, perform the operation repeatedly according to the grammar according to the contents of the input specifications, up to the limit of the tree, and until the condition is met by random numbers, Create a structure.
ii) The explanatory variable and the objective variable of the text file created in advance are respectively substituted into the expression represented by the tree, and the reciprocal of the sum of the square difference between the result (prediction) output by the expression and the objective variable is calculated as the fitness. (Evaluation function) and multiply by the genetic algorithm operator.
iii) Proliferation is performed following selection.
iv) The intersection method selects two sets of trees with high fitness with a high probability of fitness, and performs an intersection with basically the same rank of a node based on one of them. However, intersection is possible even at the lower rank unless the maximum tree (node) restriction is imposed.
v) Mutation may be performed if an expression represented by a tree is satisfied. If not, repeat until satisfied. However, when a mutation occurs at the explanatory variable coefficient K, only the numerical value of the explanatory variable coefficient K is changed.
vi) By the correlation coefficient and the F-value test, calculation is performed until the expression of the maximum fitness exceeds a predetermined number or until a predetermined time is exceeded.
vii) Repeat ii) to vi) described above.
[0055]
4) Output the calculation result of the genetic algorithm including the progress.
The calculation result to be output is displayed on the display device 17, and the display contents are as follows.
i) Distribution diagram of predicted and actual values by the formula of maximum fitness, F value, correlation function
ii) Maximum fitness and average fitness with the number of generations on the horizontal axis
iii) Formula of the selected average and maximum fitness
As a result of formulating using the genetic algorithm as described above, the following formula was obtained.
[0056]
Next, identification by a neural network in the entire winding process from the pinch rolls 5 to the mandrel 6 described above is performed, and the necessary state variables for suppressing the meandering amount, which is the state objective variable, within a predetermined threshold and The procedure for obtaining the correction amount (S8) will be described in more detail.
The purpose of this combination optimization is to extract an explanatory variable and a correction amount that cause an unknown phenomenon (abnormality) within a threshold value.
The procedure of this combination optimization will be described step by step.
[0057]
1) An abnormal target variable and an abnormal explanatory variable are input by a neural network having a plurality of intermediate layers, and an abnormal cause (explanatory variable) is identified using the abnormal phenomenon as an objective variable.
[0058]
2) Input the number of input layers, intermediate layers, and output layers of the neural network, objective variables, explanatory variables, and the like on spreadsheet software, and create a text file using a macro program. Here, the basic statistic of each variable is calculated, and data other than the parameters of each Pn average value ± 2.5 × standard deviation> Pn are extracted, and the data number and the data content are displayed on the display device 17. Judgment of adoption or rejection.
[0059]
3) Input various specifications of the operation of the genetic algorithm using an editor. The various specifications to be input are, for example, one or more intersections, the initial number of populations, the mutation rate, forbidden items (restrictions, termination conditions), and the like.
i) Here, it is also possible to specify an explanatory variable whose correction amount of the explanatory variable cannot be corrected. The specified explanatory variable correction amount is excluded from the target of the genetic algorithm operation.
ii) The explanatory variable with the smallest explanatory variable correction amount is determined to be the main cause.
iii) If the main cause cannot be corrected, the correction explanatory variable and the correction amount other than the main cause are obtained again.
iv) If the correction amount is too large to be corrected, a plurality of explanatory variables are extracted in the order of 1 → 2 → 3 →.
[0060]
4) Execute the operation of the genetic algorithm.
i) Generate an initial population. (Here, the target of the genetic algorithm is a filter α coefficient expressed by a numerical value.) A real number of 0 to 1 is represented by 8 bits, a bit string of the same number as the explanatory variable is created, and determined by various specifications. Generated as many as The operation of the genetic algorithm is performed on a bit string. Further, a value obtained by converting the bit string into a numerical value is defined as (correction amount) = 1− (numeric value).
ii) The obtained correction amount is set as a filter on the input side of the input layer in the neural network, and the calculation by the neural network is performed.
iii) Repeat the above steps i) to iii) in the order of selection → proliferation → crossing → mutation up to a predetermined fitness or time.
[0061]
5) Output the calculation result of the genetic algorithm including the progress.
The calculation result to be output is displayed on the display device 17, and the display contents are as follows.
i) Maximum fitness and average fitness with the number of generations on the horizontal axis
ii) selected mean, maximum fitness correction factor
[0062]
As a result of performing the combination optimization in this manner, it has been derived that the meandering amount falls within the threshold value by reducing the pressing force of the pinch rolls 5 by 10%. According to this, when the operation setting was actually changed, it was confirmed that the meandering amount was within the threshold value.
[0063]
In the above-described embodiment, a genetic algorithm is used as a combination optimization method. However, another combination method such as SA (simulated annealing method) or LP (linear programming) is used. Method) can also be used.
[0064]
【The invention's effect】
Since the present invention has the above-described configuration, the following effects can be obtained.
[0065]
According to the first and fourth aspects of the present invention, by using the neural network, it is possible to handle a non-linear process phenomenon, and it is easy to handle irrespective of the shape of the function, and is hardly affected by disturbance. Further, by using the genetic algorithm, optimization can be easily performed with an algorithm that avoids a local solution.
[0066]
According to the second aspect of the present invention, in addition to the effect of the first aspect of the present invention, the efficiency of the abnormal factor analysis is improved, and the causal relationship is expressed by an equation, thereby preventing the neural network from being black-boxed, thereby simplifying the process. The reliability of abnormal factor identification can be confirmed in light of physical phenomena.
[0067]
According to the third aspect of the present invention, in addition to the effects of the first or second aspect of the present invention, a solution can be obtained more quickly by using a genetic algorithm as a combination optimization technique.
[0068]
According to the fifth aspect of the present invention, in addition to the effect of the fourth aspect, when an abnormal state is found, an alarm can be issued to notify the abnormal state.
[0069]
According to the sixth aspect of the present invention, in addition to the effect of the fourth or fifth aspect, a solution can be obtained more quickly by using a genetic algorithm as a combination optimization technique.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a schematic configuration of a diagnostic device according to the present invention.
FIG. 2 is an explanatory diagram showing a schematic configuration of a winding facility of a hot rolling plant to which a diagnostic device is applied.
FIG. 3 is a flowchart illustrating an outline of a procedure of a diagnostic method according to the present invention.
FIG. 4 is a graph showing the results of studying the effectiveness of process identification using a neural network.
[Explanation of symbols]
1 Diagnostic device
2 Winding equipment
3 steel plate
4 Guide
5 Pinch roll
6 Mandrel
7 Wrapper roll
8 Pinch roll drive
9 Speed reducer
10 Mandrel drive
11 Input device
12 sensors
13 Data collection device
14 Data processing arithmetic unit
15 Identification arithmetic unit
16 Storage device
17 Display device
18 Alarm device

Claims (6)

In a method for diagnosing an abnormal state of a target facility and a product process that handles a non-linear process phenomenon online or offline,
When the measured abnormal state exceeds a predetermined threshold set at a level that does not cause a problem of equipment abnormality or quality abnormality, determine a state explanatory variable and an abnormal purpose variable related to the abnormality,
The neural network performs one or more anomaly analysis to identify the effect of the state explanatory variable on the state objective variable,
A method for diagnosing equipment and product process abnormal conditions, wherein a plurality of solutions of state explanatory variables to be corrected in order to keep abnormal target variables within a predetermined threshold value are automatically derived by a combination optimization method. In the method of expressing a causal relationship in the abnormal factor analysis process,
When the factor corresponding to the neuron in the neural network is equal to or less than the predetermined weight, the factor corresponding to the neuron is excluded from the state explanatory variables, and only the factors necessary for the determination are selected.
The evaluation function DI is expressed as DI = 1 / Σ | (neural network output) − (self-organizing expression output) | ²
The method for diagnosing an abnormal state of equipment and a product process according to claim 1, wherein the formalization by self-organization is performed by a genetic algorithm represented by: In the combination optimization method,
The method for diagnosing an abnormal state of equipment and product processes according to claim 1 or 2, wherein a genetic algorithm is used. In equipment and product process abnormal state diagnostic equipment that diagnoses abnormal conditions of target equipment and product processes that handle nonlinear process phenomena online or offline, and outputs the diagnostic results,
Data collection means for collecting data required for diagnosis;
Calculating means for converting the collected data into a state explanatory variable and an abnormal objective variable, respectively, and comparing the abnormal sign with a predetermined threshold value;
When parameters are stored and an abnormal sign appears, a neural network is created from the abnormal sign parameter and the abnormal cause parameter to identify the phenomenon, and the necessary number of abnormal cause parameters are selected based on the identified neural network. Formulating means for formulating by self-organization using a genetic algorithm,
An arithmetic processing means for selecting a plurality of solutions of the anomaly factors necessary for suppressing the anomalous symptom within a predetermined threshold value by combinational optimization;
Storage means for storing data in the data collection means, calculation means, formulating means, calculation processing means,
Display means for displaying a calculation process and a diagnosis result;
An apparatus for diagnosing an abnormal state of equipment and a product process, comprising: 5. The diagnostic apparatus according to claim 4, further comprising an alarm unit for generating an alarm when an abnormal state occurs. In the arithmetic processing means,
6. The apparatus for diagnosing an abnormal state of a facility and a product process according to claim 4, wherein an abnormal factor is selected by combination optimization using a genetic algorithm.

Images