CN113762329A - Method and system for constructing state prediction model of large rolling mill - Google Patents
Method and system for constructing state prediction model of large rolling mill Download PDFInfo
- Publication number
- CN113762329A CN113762329A CN202110764595.8A CN202110764595A CN113762329A CN 113762329 A CN113762329 A CN 113762329A CN 202110764595 A CN202110764595 A CN 202110764595A CN 113762329 A CN113762329 A CN 113762329A
- Authority
- CN
- China
- Prior art keywords
- data
- equipment
- model
- fault
- rolling mill
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Abstract
The invention provides a construction method and a construction system of a large-scale rolling mill state prediction model.A data acquisition sensor is arranged at a key part of rolling mill equipment to acquire equipment monitoring data, and the equipment monitoring data is transmitted to a data acquisition server through a wired and 5G wireless network; the data acquisition server stores the data in a database; constructing a machine learning type classification data neural network analysis model; a process data, fault symptoms and problem types are mined according to a classification data neural network analysis model, and a fault model based on process data change is constructed; and adding an equipment fault symptom early warning and grading pushing function module on the fault model, and checking by a user through a display terminal. The invention changes the limitation of the traditional equipment mainly based on a vibration monitoring method, so that the equipment can comprehensively reflect the problem of the running state of mechanical equipment under different working conditions, avoids the repetition of similar problems, and improves the equipment guarantee capability and the product quality stability.
Description
Technical Field
The invention relates to the technical field of rolling mills, in particular to a construction method and a construction system of a large rolling mill state prediction model.
Background
The intelligent state identification and high-precision fault diagnosis of large key equipment in the field of steel manufacturing are all the major and difficult points in the field of industry. Steel rolling mill equipment is under the bad operating mode such as thousands of tons of alternating impact loads, last thousand degrees of lasting high temperature environment, lead to system level trouble to relapse.
Factors in the system-level fault are complex, and include equipment faults, process faults and disturbance faults caused by raw material quality fluctuation. The existing equipment fault monitoring method generally only acquires and analyzes the running data of the equipment body, and basically does not integrate and analyze the running data with production process information and process parameters, so that the root cause of the equipment state analysis result is greatly different from the actual problem.
Moreover, due to the lack of flexible analysis tools, system-level faults can only depend on manual analysis and fault location according to the presented symptoms, so that the accident analysis period is long, and the qualitative difficulty of accident reasons is high. Meanwhile, some unknown risks and uncertain factors are often ignored, troubles are brought to production control and operation decision, and great uncertainty is brought to the quality of commodities.
Disclosure of Invention
In order to solve the problem of inaccurate system-level fault analysis in the existing steel large-scale rolling mill equipment state monitoring technology, the invention provides a construction method of a prediction model for performing state recognition and comprehensive guidance on large-scale rolling mill equipment and predicting the state of the large-scale rolling mill by fusing an analysis model of process data and equipment data;
the method comprises the following steps:
step 1: arranging a data acquisition sensor at a key part of rolling mill equipment, acquiring equipment monitoring data rolled piece process parameters and process control parameter data, and transmitting the acquired data to a data acquisition server through a wired and 5G wireless network;
step 2: the data acquisition server stores the data in a database;
and step 3: constructing a machine learning type classification data neural network analysis model;
and 4, step 4: a process data, fault symptoms and problem types are mined according to a classification data neural network analysis model, and a fault model based on process data change is constructed;
and 5: and adding an equipment fault symptom early warning and grading pushing function module on the fault model, and checking by a user through a display terminal.
Preferably, the data acquisition sensor is arranged at the rotating part of the equipment, and the interface is arranged on each equipment operation system and the information system.
Preferably, the database storage adopts an HBase distributed database, and the equipment operation monitoring data in the database correspond to the process operation data time points one by one;
the database data storage process comprises the following steps:
step 11: acquiring original data of the equipment by using the equipment data acquisition sensor, wherein the acquisition frequency of the original data is acquired for a preset number of times within a preset time period;
acquiring the technological parameters and the process control parameters of the rolled piece to a data acquisition server for storage, wherein the acquisition frequency is acquired at preset times per second;
step 12: cleaning, drying and redundancy removing are carried out on historical data and real-time data in a server, characteristic values of the data are obtained through Fourier transform and wavelet transform calculation, dimensionless processing is carried out on the characteristic data, and a normalization method is selected for improving the accuracy and speed of a BP neural network;
step 13: and establishing an HBase distributed database, and storing the equipment and process state characteristic data into the database.
Preferably, the classification data neural network analysis model is a BP neural network model designed based on a BP _ Adaboost strong classifier, and is used for analyzing and classifying database data, obtaining a data diagnosis result by adjusting the iteration times and the learning rate of the model, comparing the data diagnosis result with the current normal index value, and predicting the equipment state.
Preferably, the algorithm steps of classifying the data neural network analysis model are as follows:
step 21: randomly selecting m groups of training data from data sample space, and initializing distribution weight S of test datap(p) 1/m, determining a neural network structure according to the input and output dimensionality of the sample, and obtaining a weight and a threshold value of the initialized BP neural network by using an activation function sigmoid (x);
step 22: predicting a weak classifier; when training the nth weak classifier, training BP neural network with training data and predicting training data detection to obtain prediction error sum e of prediction sequence Z (n)n；
Step 23: calculating the weight of the prediction sequence;
step 24: adjusting the weight of the test data;
step 25: a strong classification function;
after N rounds of training, N groups of weak classification functions f (g) are usedn,an) From N groups of weak classification functions f (g)n,an) The combination yields a strong classification function h (x).
Preferably, the model prediction result is judged, and three indexes are adopted for judging, namely a loss function, a cost function and a target function;
the loss function is used for calculating the error of a single sample; the cost function is used for calculating the average error of the whole training sample set; the objective function is composed of a cost function and a regularization function, and a final optimization function is obtained.
Preferably, the model prediction result is evaluated, and the optimal model is obtained and stored by adopting a discrimination evaluation index AUC and a goodness-of-fit evaluation index AIC.
Preferably, a fault model based on process data change is constructed by analyzing a quantitative mapping corresponding rule of model mining process data, fault symptoms and problem types;
analyzing the data through a data analysis model to obtain the correlation relationship between the equipment state characteristics and the process parameters so as to generate a rule base corresponding to the quantitative mapping of fault symptoms and problem types caused by the change value of each parameter;
in the running process of the equipment, a real-time monitoring fault model is constructed by comparing and analyzing the real-time monitoring data with a rule base; the fault model is dynamically adjusted, and when a fault occurs, the data analysis model can automatically form fault information into a fault training data set, so that the subsequent fault symptoms can be predicted in advance and reference can be provided.
Preferably, an equipment fault symptom early warning and grading pushing function module is added to the fault model, and a user views the fault model through a display terminal;
and when the fluctuation of the technological data parameter value exceeds a threshold value, sending an alarm prompt, simultaneously acquiring instantaneous running data of the rolling mill equipment by using an equipment state analysis model for analysis, monitoring the equipment state if abnormal data continuously appear in a certain time, judging whether the current running state of the equipment is healthy or not, sending an alarm to the terminal equipment, making a judgment decision after a user knows the pushed information, and continuously adjusting and improving the technological parameter of the equipment according to the judgment decision.
The invention also provides a system for constructing the large-scale rolling mill state prediction model, which comprises the following steps: the system comprises a data acquisition sensor, a data acquisition server, a database and a data processing server;
arranging a data acquisition sensor at a key part of rolling mill equipment, acquiring equipment monitoring data rolled piece process parameters and process control parameter data, and transmitting the acquired data to a data acquisition server through a wired and 5G wireless network;
the data acquisition server stores the acquired data in a database;
the data processing server constructs a machine learning type classification data neural network analysis model;
a process data, fault symptoms and problem types are mined according to a classification data neural network analysis model, and a fault model based on process data change is constructed;
and adding an equipment fault symptom early warning and grading pushing function module on the fault model, and checking by a user through a display terminal.
According to the technical scheme, the invention has the following advantages:
the method for constructing the large-scale rolling mill state prediction model can find potential rules of production equipment problems in time, determine fault types, accurately position the reasons of the problems, improve the easy maintenance and the protection of the equipment under the conditions of fully exerting the manufacturing capacity of the equipment and adapting to high-load production, provide intelligent guarantee for avoiding major equipment process safety accidents, and reduce the maintenance cost of the equipment in the whole life cycle.
The invention can early warn the hidden trouble of the equipment in advance and push and inform the relevant personnel, and simultaneously provides an improvement direction for continuously improving the equipment and the process after comprehensively analyzing the hidden trouble points, thereby increasing the equipment control accuracy, improving the product control level and providing reliable support for reducing the product quality risk.
The invention improves the utilization efficiency of low-value density data by fusing a new generation information technology, breaks through the bottleneck problem of large-scale key equipment state identification and system-level fault analysis in the industry, realizes data transparence, and truly ensures that the value added in the data-driven business process falls into the real place.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings used in the description will be briefly introduced, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained based on these drawings without creative efforts.
FIG. 1 is a flow chart of an embodiment of a method for constructing a state prediction model of a large rolling mill;
FIG. 2 is a flow chart of a method for constructing a state prediction model of a large rolling mill;
FIG. 3 is a schematic diagram of a data analysis model training of a construction method of a large-scale rolling mill state prediction model;
FIG. 4 is a schematic diagram of a system for constructing a large-scale rolling mill state prediction model.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The units and algorithm steps of each example described in the embodiment disclosed in the method for constructing the large rolling mill state prediction model provided by the invention can be realized by electronic hardware, computer software or a combination of the electronic hardware and the computer software, and in order to clearly illustrate the interchangeability of the hardware and the software, the components and the steps of each example are generally described according to functions in the above description. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The block diagram shown in the drawing of the construction method of the large-scale rolling mill state prediction model provided by the invention is only a functional entity and does not necessarily correspond to a physically independent entity. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.
In addition, the methods of constructing the large mill state prediction model provided by the present invention may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention may be practiced without one or more of the specific details, or with other methods, components, devices, steps, and so forth. In other instances, well-known methods, devices, implementations or operations have not been shown or described in detail to avoid obscuring aspects of the invention.
Referring to fig. 1 and fig. 2, the invention provides a flow chart diagram of a method for constructing a data-driven large rolling mill state prediction model. The method comprises the following steps:
s1, arranging a data acquisition sensor at a key part of the rolling mill equipment, acquiring equipment monitoring data, and transmitting the equipment monitoring data to a data acquisition server for storage through a wired and 5G wireless network; transmitting the process parameters and the process control parameter data of the rolled piece to a data acquisition server for storage;
and S2, the data acquisition server processes the stored data and establishes a database.
S3, constructing a machine learning type classification data neural network analysis model;
and S4, mining the quantitative mapping corresponding rule of the process data, the fault symptoms and the problem types through the classified data neural network analysis model, and constructing a fault model based on process data change.
And S5, adding a device fault symptom early warning and grading pushing function module on the fault model, wherein personnel at all levels can check the fault by different display terminals such as a computer and a mobile phone.
The data acquisition sensor is arranged at the rotating part of the equipment, and the interfaces are arranged on the running systems and the information systems of the equipment;
in the invention, the HBase distributed database is adopted for storing the database established after the system processes the acquired data, and the equipment operation monitoring data in the database correspond to the process operation data time points one by one;
the database data storage process comprises the following steps:
step 11: acquiring original data of the equipment by using the equipment data acquisition sensor, wherein the acquisition frequency of the original data is more than 10 times per second; collecting the technological parameters and the process control parameters of the rolled piece to a data collection server for storage, wherein the collection frequency is more than 10 times per second; .
Step 12: and cleaning, drying and redundancy removing are carried out on historical data and real-time data in the server. Firstly, data is extracted, in order to guarantee that the operation condition data of the unit under different loads and different environments are covered, the historical operation time of the extracted data is more than 1 year, and parameters of a device measuring point program for historical data acquisition, data acquisition start-stop time, intervals, data set names and the like are well set; and secondly, carrying out incidence relation comparison processing on data containing abnormal operation state data, noise data and equipment shutdown data of the equipment, excavating accurate and scientific effective data required by modeling to ensure the accuracy of the built model, carrying out dimensionless processing on the data after the data is cleaned, removing redundant measuring points of the equipment modeling, and not considering the measuring point data with incidence degree lower than 20%. The processed data is subjected to Fourier transform and wavelet transform to obtain a characteristic value of the data, the characteristic data is subjected to non-dimensionalization conversion, and in order to improve the accuracy and speed of the BP neural network, a normalization method is selected, wherein the formula is as follows:
y＝(x-min)/(max-min)
x and y are values before and after non-dimensionalization conversion, and max and min are the maximum value and the minimum value of the sample.
Step 13: and establishing an HBase distributed database, and storing the equipment and process state characteristic data into the database.
As an embodiment of the invention, the data analysis model is a BP neural network model designed based on a BP _ Adaboost strong classifier, and is used for analyzing and classifying database data, obtaining a data diagnosis result by adjusting the iteration times and the learning rate of the model, comparing the data diagnosis result with the current normal index value, and predicting the state of equipment;
as shown in fig. 3, an algorithm of the data analysis model training diagram provided by the present invention has the following steps.
Step 21: randomly selecting m groups of training data from data sample space, and initializing distribution weight S of test datapAnd (p) 1/m, determining the neural network structure according to the input and output dimensionality of the sample, and obtaining the weight and the threshold of the initialized BP neural network by using an activation function sigmoid (x). The activation function is formulated as:
deriving a weight value and a threshold value updating formula of the BP neural network by an activation function sigmoid (x);
by way of illustration: the input layer is set to be described by s attributes, the hidden layer is set to be p hidden layer neurons, the output layer is a multilayer feedforward network structure formed by m-dimensional real value vectors, and the mathematical formula is expressed as follows: training set T { (x)1，y1)，…，(xn，yn)}，
Threshold for h neuron of output layer neuron is represented by θhThe connection weight between the jth neuron of the input layer and the ith neuron of the hidden layer is rhojiThe connection weight between the ith neuron of the hidden layer and the ith neuron of the output layer is sigmaih(ii) a Then the input received by the i-th neuron of the hidden layer is
The h-th neuron of the output layer receives as input
For a certain case (x) of training samplej,yj) The output of which neural network is
The neural network is in (x)j,yj) Mean square error of
Given a learning rate lambda, the updating formula of the connection weight between the ith neuron of the hidden layer and the ith neuron of the output layer is as follows:
threshold value Delta theta of j-th neuron of output layer neuronsjThe update formula of (2) is:
the updating formula of the threshold value gamma of the ith neuron of the hidden layer of the neural network is as follows:
connection weight value delta rho from output layer to hidden layer of neural networkjiThe update formula of (2) is:
step 22: and predicting by a weak classifier. Training the nth personWhen the classifier is used, training the BP neural network by using the training data and predicting the training data to obtain the prediction error of the prediction sequence Z (n) and enError sum enIs calculated by the formula
Wherein Z (n) is the predicted classification result and y is the expected classification result.
Step 23: and calculating the weight of the prediction sequence. Prediction error sum e from prediction sequence Z (n)nCalculating the weight d of the sequencenThe weight is calculated as
Step 24: and adjusting the weight of the test data. According to predicted sequence weight dnAdjusting the weight of the next round of training samples according to the formula
In the formula, BnIs a normalization factor in order to make the sum of the distribution weights 1 without changing the weight ratio.
Step 25: a strong classification function.
Train N rounds of rear use NComponent (B) ofClass function f (g)n，an)，
From N groups of weak classification functions f (g)n，an) The combination yields a strong classification function h (x) of the formula:
and when the prediction model is trained, the historical operating data is used as a training set, so that the model learns the characteristics of different stages. For example, 10000 sets of equipment process data at a certain stage are shared, the input of each set of data is 10 dimensions, represents 8 indexes of process parameters, the output is 1 dimension, represents the equipment condition, represents that the equipment condition is good when the input is 1, and represents that the equipment condition has a problem when the input is-1. 8000 groups of data are randomly selected from the data as training data, 2000 groups of data are selected as test data, a BP neural network structure is 8-4-1 according to the data dimension, 8 BP neural network weak classifiers are generated through co-training, and finally, the strong classifiers formed by the 8 weak classifiers are used for classifying the equipment conditions. And adjusting the number of layers of the model and the number of the neurons according to the classification result until the predicted value is consistent with the actual value, and then saving the model.
The model prediction result is judged, and three indexes are adopted for judging, namely a loss function, a cost function and a target function.
The loss function is used for calculating the error of a single sample, and an exponential loss function is used for the classification model;
L(y,f(x))＝e-yf(x)
the cost function is used for calculating the average error of the whole training sample set; the average cross entropy of all loss function values is calculated, the cross entropy is used for evaluating the difference condition of probability distribution obtained by current training in real distribution, and the smaller the cross entropy is, the higher the accuracy of prediction of the model is. The classification problem function formula is as follows:
where p (x) is the probability of the true distribution and q (x) is the probability estimate calculated by the model from the data.
The objective function is a cost function plus a regularization function to obtain a final optimization function.
Wherein the first part is a cost function, and L represents a loss function; the second part is a regularization function, λ represents the regularization term coefficients, regularization prevents model overfitting, and the entire function expression is the best value to find for the objective function.
According to the method, the model prediction result is evaluated, and the optimal model is obtained and stored by adopting a discrimination evaluation index AUC and a goodness of fit evaluation index AIC.
(1) AUC is a performance measure for machine learning models, with which models can be found that can best interpret data but contain the fewest free parameters.
(2) AIC is a criterion to measure the goodness of a statistical model fit.
AIC＝2k-21n(L)
Where k is the number of model parameters (model complexity) and L is the likelihood function. The smaller the AIC value the better when selecting the best model among a set of alternative models.
In the invention, a fault model based on process data change is constructed by analyzing a quantitative mapping corresponding rule of model mining process data, fault symptoms and problem types. Analyzing the data through a data analysis model to obtain the correlation relationship between the equipment state characteristics and the process parameters so as to generate a rule base corresponding to the quantitative mapping of fault symptoms and problem types caused by the change value of each parameter; in the running process of the equipment, a real-time monitoring fault model is constructed by comparing and analyzing the real-time monitoring data with a rule base; the fault model is dynamically adjusted, and when a fault occurs, the data analysis model can automatically form fault information into a fault training data set, so that the subsequent fault symptoms can be predicted in advance and reference can be provided.
Based on the method, the invention also provides a system for constructing the large-scale rolling mill state prediction model, which is shown in figure 4: the method comprises the following steps: the system comprises a data acquisition sensor 1, a data acquisition server 2, a database 3 and a data processing server 4;
arranging a data acquisition sensor 1 at a key part of rolling mill equipment, acquiring equipment monitoring data rolled piece process parameters and process control parameter data, and transmitting the acquired data to a data acquisition server 2 through a wired and 5G wireless network;
the data acquisition server 2 stores the data in the database 3;
the data processing server 4 constructs a machine learning type classification data neural network analysis model;
a process data, fault symptoms and problem types are mined according to a classification data neural network analysis model, and a fault model based on process data change is constructed;
and adding an equipment fault symptom early warning and grading pushing function module on the fault model, and checking by a user through a display terminal 5.
In the system, an equipment fault symptom early warning and grading pushing function module is added on a fault model, and personnel at all levels can check the fault model through different display terminals such as a computer and a mobile phone. When the fluctuation of the technological data parameter value exceeds a threshold value, an alarm prompt is sent out, meanwhile, an equipment state analysis model collects instantaneous running data of rolling mill equipment for analysis, if abnormal data continuously appear in a certain time, the equipment state is monitored, whether the current running state of the equipment is healthy or not is judged, an alarm is sent out and pushed to equipment managers at all levels, the equipment managers at all levels can know the pushed information through terminal equipment such as a computer and a mobile phone and then make judgment decision, and the technological parameters of the equipment are continuously adjusted and improved.
The invention aims to solve the problem that the system-level fault analysis of the large-scale rolling mill is inaccurate, the method for realizing the state recognition of the large-scale rolling mill equipment and the predictive maintenance of the comprehensive guidance equipment by fusing the analysis model of the process data and the equipment data and fusing the new generation information technology improves the utilization efficiency of low-value density data, breaks through the bottleneck problem of the state recognition of the large-scale key equipment and the system-level fault analysis in the industry, realizes the data transparence and truly ensures that the value added in the data-driven service process falls into the real place.
The system for constructing a large rolling mill state prediction model provided by the invention is the units and algorithm steps of each example described in connection with the embodiments disclosed herein, and can be realized by electronic hardware, computer software or a combination of the two, and in order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functions in the above description. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A method for constructing a large-scale rolling mill state prediction model is characterized by comprising the following steps:
step 1: arranging a data acquisition sensor at a key part of rolling mill equipment, acquiring equipment monitoring data rolled piece process parameters and process control parameter data, and transmitting the acquired data to a data acquisition server through a wired and 5G wireless network;
step 2: the data acquisition server stores the data in a database;
and step 3: constructing a machine learning type classification data neural network analysis model;
and 4, step 4: a process data, fault symptoms and problem types are mined according to a classification data neural network analysis model, and a fault model based on process data change is constructed;
and 5: and adding an equipment fault symptom early warning and grading pushing function module on the fault model, and checking by a user through a display terminal.
2. The method for constructing the large rolling mill state prediction model according to claim 1, characterized in that:
the data acquisition sensor is arranged at the rotating part of the equipment, and the interfaces are arranged on the running system and the information system of each equipment.
3. The method for constructing the large rolling mill state prediction model according to claim 1, characterized in that:
the database storage adopts an HBase distributed database, and equipment operation monitoring data in the database correspond to process operation data time points one by one;
the database data storage process comprises the following steps:
step 11: acquiring original data of the equipment by using the equipment data acquisition sensor, wherein the acquisition frequency of the original data is acquired for a preset number of times within a preset time period;
acquiring the technological parameters and the process control parameters of the rolled piece to a data acquisition server for storage, wherein the acquisition frequency is acquired at preset times per second;
step 12: cleaning, drying and redundancy removing are carried out on historical data and real-time data in a server, characteristic values of the data are obtained through Fourier transform and wavelet transform calculation, dimensionless processing is carried out on the characteristic data, and a normalization method is selected for improving the accuracy and speed of a BP neural network;
step 13: and establishing an HBase distributed database, and storing the equipment and process state characteristic data into the database.
4. The method for constructing the large rolling mill state prediction model according to claim 1, characterized in that:
the classification data neural network analysis model is a BP neural network model designed based on a BP _ Adaboost strong classifier and used for analyzing and classifying database data, obtaining a data diagnosis result by adjusting the iteration times and the learning rate of the model, comparing the data diagnosis result with the current normal index value and predicting the equipment state.
5. The method for constructing the large rolling mill state prediction model according to claim 4, characterized in that:
the algorithm steps of the classification data neural network analysis model are as follows:
step 21: randomly selecting m groups of training data from data sample space, and initializing distribution weight S of test datap(p) 1/m, determining a neural network structure according to the input and output dimensionality of the sample, and obtaining a weight and a threshold value of the initialized BP neural network by using an activation function sigmoid (x);
step 22: predicting a weak classifier; when training the nth weak classifier, training the BP neural network by using the training data and detecting the prediction training data to obtain the prediction error of the prediction sequence Z (n) and en；
Step 23: calculating the weight of the prediction sequence;
step 24: adjusting the weight of the test data;
step 25: a strong classification function;
after N rounds of training, N groups of weak classification functions f (g) are usedn，an) From N groups of weak classification functions f (g)n，an) The combination yields a strong classification function h (x).
6. The method for constructing the large rolling mill state prediction model according to claim 4, characterized in that:
judging the model prediction result by adopting three indexes, namely a loss function, a cost function and a target function;
the loss function is used for calculating the error of a single sample; the cost function is used for calculating the average error of the whole training sample set; the objective function is composed of a cost function and a regularization function, and a final optimization function is obtained.
7. The method for constructing the large rolling mill state prediction model according to claim 4, characterized in that:
and evaluating the model prediction result, and storing after obtaining the optimal model by adopting a discrimination evaluation index AUC and a goodness of fit evaluation index AIC.
8. The method for constructing the large rolling mill state prediction model according to claim 1, characterized in that:
a fault model based on process data change is constructed by analyzing a quantitative mapping corresponding rule of model mining process data, fault symptoms and problem types;
analyzing the data through a data analysis model to obtain the correlation relationship between the equipment state characteristics and the process parameters so as to generate a rule base corresponding to the quantitative mapping of fault symptoms and problem types caused by the change value of each parameter;
in the running process of the equipment, a real-time monitoring fault model is constructed by comparing and analyzing the real-time monitoring data with a rule base; the fault model is dynamically adjusted, and when a fault occurs, the data analysis model can automatically form fault information into a fault training data set, so that the subsequent fault symptoms can be predicted in advance and reference can be provided.
9. The method for constructing the large rolling mill state prediction model according to claim 1, characterized in that:
adding an equipment fault symptom early warning and grading pushing function module on the fault model, and checking by a user through a display terminal;
and when the fluctuation of the technological data parameter value exceeds a threshold value, sending an alarm prompt, simultaneously acquiring instantaneous running data of the rolling mill equipment by using an equipment state analysis model for analysis, monitoring the equipment state if abnormal data continuously appear in a certain time, judging whether the current running state of the equipment is healthy or not, sending an alarm to the terminal equipment, making a judgment decision after a user knows the pushed information, and continuously adjusting and improving the technological parameter of the equipment according to the judgment decision.
10. A system for constructing a large-scale rolling mill state prediction model is characterized in that: the method comprises the following steps: the system comprises a data acquisition sensor, a data acquisition server, a database and a data processing server;
arranging a data acquisition sensor at a key part of rolling mill equipment, acquiring equipment monitoring data rolled piece process parameters and process control parameter data, and transmitting the acquired data to a data acquisition server through a wired and 5G wireless network;
the data acquisition server stores the acquired data in a database;
the data processing server constructs a machine learning type classification data neural network analysis model;
a process data, fault symptoms and problem types are mined according to a classification data neural network analysis model, and a fault model based on process data change is constructed;
and adding an equipment fault symptom early warning and grading pushing function module on the fault model, and checking by a user through a display terminal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110764595.8A CN113762329A (en) | 2021-07-06 | 2021-07-06 | Method and system for constructing state prediction model of large rolling mill |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110764595.8A CN113762329A (en) | 2021-07-06 | 2021-07-06 | Method and system for constructing state prediction model of large rolling mill |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113762329A true CN113762329A (en) | 2021-12-07 |
Family
ID=78787571
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110764595.8A Pending CN113762329A (en) | 2021-07-06 | 2021-07-06 | Method and system for constructing state prediction model of large rolling mill |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113762329A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114253242A (en) * | 2021-12-21 | 2022-03-29 | 上海纽酷信息科技有限公司 | VPN-based Internet of things cloud equipment data acquisition system |
| CN114971252A (en) * | 2022-05-19 | 2022-08-30 | 浙江理工大学 | Operation and maintenance and fault pre-diagnosis system for textile equipment |
| CN115331155A (en) * | 2022-10-14 | 2022-11-11 | 智慧齐鲁(山东)大数据科技有限公司 | Mass video monitoring point location graph state detection method and system |
| CN116586441A (en) * | 2023-07-10 | 2023-08-15 | 江苏甬金金属科技有限公司 | Intelligent control method and system for metal rolling mill |
| CN116702030A (en) * | 2023-05-31 | 2023-09-05 | 浙江大学 | Blast furnace state monitoring method and device based on sensor reliability analysis |
| CN116931530A (en) * | 2023-09-15 | 2023-10-24 | 山东星科智能科技股份有限公司 | Dynamic prediction system and dynamic prediction method for iron-making steel rolling production process |
| CN117019889A (en) * | 2023-07-18 | 2023-11-10 | 常州润来科技有限公司 | Method and system for intelligently detecting faults of three-roller rotary rolling equipment |
| CN117086117A (en) * | 2023-10-18 | 2023-11-21 | 江苏甬金金属科技有限公司 | Temperature control method and system for titanium belt hot rolling process |
| CN114861421B (en) * | 2022-04-22 | 2024-04-26 | 兮然科技(江苏)有限公司 | Intelligent optimization control system and method for air flow mill classifying wheel |
-
2021
- 2021-07-06 CN CN202110764595.8A patent/CN113762329A/en active Pending
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114253242B (en) * | 2021-12-21 | 2023-12-26 | 上海纽酷信息科技有限公司 | VPN-based cloud equipment data acquisition system for Internet of things |
| CN114253242A (en) * | 2021-12-21 | 2022-03-29 | 上海纽酷信息科技有限公司 | VPN-based Internet of things cloud equipment data acquisition system |
| CN114861421B (en) * | 2022-04-22 | 2024-04-26 | 兮然科技(江苏)有限公司 | Intelligent optimization control system and method for air flow mill classifying wheel |
| CN114971252A (en) * | 2022-05-19 | 2022-08-30 | 浙江理工大学 | Operation and maintenance and fault pre-diagnosis system for textile equipment |
| CN115331155A (en) * | 2022-10-14 | 2022-11-11 | 智慧齐鲁(山东)大数据科技有限公司 | Mass video monitoring point location graph state detection method and system |
| CN115331155B (en) * | 2022-10-14 | 2023-02-03 | 智慧齐鲁(山东)大数据科技有限公司 | Mass video monitoring point location graph state detection method and system |
| CN116702030A (en) * | 2023-05-31 | 2023-09-05 | 浙江大学 | Blast furnace state monitoring method and device based on sensor reliability analysis |
| CN116702030B (en) * | 2023-05-31 | 2024-01-23 | 浙江大学 | Blast furnace state monitoring method and device based on sensor reliability analysis |
| CN116586441A (en) * | 2023-07-10 | 2023-08-15 | 江苏甬金金属科技有限公司 | Intelligent control method and system for metal rolling mill |
| CN116586441B (en) * | 2023-07-10 | 2023-11-14 | 江苏甬金金属科技有限公司 | Intelligent control method and system for metal rolling mill |
| CN117019889A (en) * | 2023-07-18 | 2023-11-10 | 常州润来科技有限公司 | Method and system for intelligently detecting faults of three-roller rotary rolling equipment |
| CN116931530A (en) * | 2023-09-15 | 2023-10-24 | 山东星科智能科技股份有限公司 | Dynamic prediction system and dynamic prediction method for iron-making steel rolling production process |
| CN116931530B (en) * | 2023-09-15 | 2023-12-26 | 山东星科智能科技股份有限公司 | Dynamic prediction system and dynamic prediction method for iron-making steel rolling production process |
| CN117086117B (en) * | 2023-10-18 | 2023-12-29 | 江苏甬金金属科技有限公司 | Temperature control method and system for titanium belt hot rolling process |
| CN117086117A (en) * | 2023-10-18 | 2023-11-21 | 江苏甬金金属科技有限公司 | Temperature control method and system for titanium belt hot rolling process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113762329A (en) | Method and system for constructing state prediction model of large rolling mill | |
| CN108805202B (en) | Machine learning method for electrolytic bath fault early warning and application thereof | |
| CN113255848B (en) | Water turbine cavitation sound signal identification method based on big data learning | |
| US8090676B2 (en) | Systems and methods for real time classification and performance monitoring of batch processes | |
| CN114358152A (en) | Intelligent power data anomaly detection method and system | |
| CN112085947A (en) | Traffic jam prediction method based on deep learning and fuzzy clustering | |
| CN113156917A (en) | Power grid equipment fault diagnosis method and system based on artificial intelligence | |
| CN113378990B (en) | Flow data anomaly detection method based on deep learning | |
| CN112904810B (en) | Process industry nonlinear process monitoring method based on effective feature selection | |
| CN113111591B (en) | Automatic diagnosis method, device and equipment based on internal fault of modular power distribution terminal | |
| CN110757510A (en) | Method and system for predicting remaining life of robot | |
| CN112906764A (en) | Communication safety equipment intelligent diagnosis method and system based on improved BP neural network | |
| CN113627735A (en) | Early warning method and system for safety risk of engineering construction project | |
| CN115096627A (en) | Method and system for fault diagnosis and operation and maintenance in manufacturing process of hydraulic forming intelligent equipment | |
| CN117193222A (en) | Intelligent quality control system based on industrial Internet of things and big data and control method thereof | |
| CN112434390A (en) | PCA-LSTM bearing residual life prediction method based on multi-layer grid search | |
| CN110689140A (en) | Method for intelligently managing rail transit alarm data through big data | |
| CN117076869A (en) | Time-frequency domain fusion fault diagnosis method and system for rotary machine | |
| CN113987905A (en) | Escalator braking force intelligent diagnosis system based on deep belief network | |
| CN116956702A (en) | Electricity safety early warning method, medium and system | |
| CN116109039A (en) | Data-driven anomaly detection and early warning system | |
| CN116108371A (en) | Cloud service abnormity diagnosis method and system based on cascade abnormity generation network | |
| CN116204825A (en) | Production line equipment fault detection method based on data driving | |
| CN116061690A (en) | Safety early warning method and device in electric automobile charging process | |
| CN116311739A (en) | Multi-sensor fire detection method based on long-short-term memory network and environment information fusion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |