Tightening tool parameter optimization system and method based on deep learningTechnical Field
The invention relates to the technical field of tool parameter optimization, in particular to a tightening tool parameter optimization system and method based on deep learning.
Background
In modern industrial production, the tightening operation is used as a key step of an assembly link, the parameter optimization of the tightening operation has an important influence on the quality and the production efficiency of a product, and the parameter optimization method of the traditional tightening tool mostly adopts fixed process parameters, so that the method has certain applicability under standardized production conditions, but in actual production, due to dynamic changes of environment and working conditions, such as differences of material characteristics, external vibration interference, fluctuation of operation conditions and the like, the fixed parameters are difficult to meet actual demands, and the applicability is poor.
Aiming at the problem, a part of schemes are preset for combining process parameters under different scenes to meet production requirements, but the mode is difficult to flexibly respond to real-time changes in a production line, in multi-station cooperative work or complex assembly tasks, unified communication and cooperative mechanisms are lacked among tools, so that parameters among devices are not matched, the consistency of operation is influenced, and finally the assembly quality is negatively influenced;
Along with popularization of industrial Internet of things IIoT, multi-equipment coordination and intellectualization become industry trends, and traditional tightening tools cannot realize seamless integration with an industrial Internet of things platform, lack of uniform data acquisition and analysis means, limit further improvement of tool performance and also cannot meet the demand of intelligent production, so that a tightening tool parameter optimization system and method based on deep learning are needed to solve the problems.
Disclosure of Invention
The present invention has been made in view of the above-described problems occurring in the prior art.
The invention provides a tightening tool parameter optimization system and method based on deep learning, which solve the problems that the traditional scheme test core logic is still optimized by a single tool, is difficult to cope with the requirement of a complex production environment, and has obvious defects in real-time performance, multi-equipment coordination, dynamic adjustment and intelligent application.
In order to solve the technical problems, the invention provides the following technical scheme:
in a first aspect, embodiments of the present invention provide a deep learning based tightening tool parameter optimization system, comprising,
The data acquisition module is used for acquiring operation data and environmental parameters of the tightening tool in real time;
The edge calculation module is used for receiving the real-time data transmitted by the data acquisition module, performing preliminary processing on the data, rapidly analyzing the real-time data by using the deep learning model RNN, predicting the optimal tightening parameters of a single tool in real time, and transmitting the result to the process parameter optimization module;
the analysis module is used for predicting the cooperative operation parameters among multiple tools by utilizing a deep learning transducer algorithm based on the data transmitted by the edge calculation module, outputting an optimization strategy and updating the model parameters of the process parameter optimization module and the edge calculation module;
The process parameter optimization module is used for receiving the optimization parameters output by the analysis module and the edge calculation module, adjusting the tightening parameters in combination with real-time working conditions, wherein the real-time working conditions comprise the current tool state and the environmental change dynamics;
the feedback module is used for collecting the operation performance and the adjusted operation parameters in real time after the tightening operation is completed, uploading the operation parameters to the cloud end and transmitting the operation parameters to the analysis module through the industrial Internet of things IIoT platform;
The industrial Internet of things IIoT platform is used as a central point for connecting all modules, is responsible for real-time transmission and sharing of data, receives the data uploaded by the feedback module and transmits the data to the analysis module.
The tightening tool parameter optimization system and the tightening tool parameter optimization method based on deep learning are used for optimizing the tightening tool parameters, wherein the operation data comprise tightening torque, tightening speed and tightening times, and the environment parameters comprise material hardness and roughness, and temperature, humidity and vibration in a production environment.
The tightening tool parameter optimization system and the tightening tool parameter optimization method based on deep learning are used as an optimal scheme, wherein the operation parameters comprise operation effects and execution parameters, the operation effects comprise whether tightening quality reaches a standard, and the execution parameters comprise torque values and tightening angle deviations.
The optimal scheme of the tightening tool parameter optimization system and the tightening tool parameter optimization method based on deep learning is that the industrial Internet of things IIoT platform is used for optimizing parameter sharing among multiple tools, and adjusting operation sequence, load balancing and task distribution.
In a second aspect, the present invention provides a method for optimizing parameters of a tightening tool based on deep learning, comprising,
Step S1, a data acquisition module acquires operation parameters of a tightening tool, including moment, speed, tightening material hardness and temperature and humidity, and the operation parameters are cleaned, standardized and feature extracted in an edge calculation module to generate real-time data;
S2, the edge calculation module sends real-time data to the analysis module through network connection, the analysis module adopts a deep learning RNN model to predict optimal tightening parameters, including moment and speed, and the process parameter optimization module automatically adjusts the tightening parameters based on the optimal tightening parameters and in combination with real-time working conditions;
Step S3, the tightening tool performs tightening operation according to the adjusted tightening parameters, the data acquisition module acquires the adjusted operation parameters and operation performance in real time, and the operation parameters and operation performance are uploaded to the cloud end through an industrial Internet of things IIoT platform;
And S4, the cloud receives the adjusted operation parameters and the adjusted working performance through an industrial Internet of things IIoT platform, analyzes load balancing and operation sequences among tightening tools by using a deep learning transducer model, updates an optimization algorithm based on an analysis result, and transmits the optimized model to an edge calculation module.
As a preferable scheme of the tightening tool parameter optimization system and method based on deep learning, the invention comprises the steps of cleaning, normalizing and extracting features in an edge calculation module, generating real-time data,
Detecting abnormal values based on distribution analysis, and detecting whether the collected data points xi belong to a reasonable interval or not, wherein a monitoring formula is xi∈[Q1-1.5·IQR,Q3 +1.5·iqr, wherein Q1 is the 1 st quartile of data, Q3 is the 3 rd quartile of data, and iqr=q3-Q1 is a quartile distance;
and (3) carrying out outlier processing on the detected outlier by adopting a linear interpolation method, wherein the processing formula is as follows:
Wherein xi is an outlier, and xi-1,xi+1 is a normal data point before and after the outlier;
ZScore normalization was performed for each data point, and the normalization formula was:
Wherein zi is normalized data, xi is cleaned original data, mux is the mean value of the data, and sigmax is the standard deviation of the data;
Feature extraction is performed in the normalized data, and the following features are extracted for the moment T (T) and the speed v (T), including:
Average value of
Maximum tmax:
Tmax＝max(Ti),
root mean square value Trms:
wherein Ti is the moment value of the ith time point, and N is the total number of sampling points;
for the dynamic change rate of temperature and humidity calculation, the calculation formula is as follows:
Wherein ht,Tt is the humidity and temperature at the current time point, Rh,RT is the rate of change of the humidity and temperature, and Δt is the sampling time interval;
and calculating the Pearson correlation coefficient of the moment and the speed, wherein the calculation formula is as follows:
Wherein Ti,vi is the moment and the speed at the ith time point respectively,As an average of the moment and the speed,
RT,v is a linear dependence.
As a preferable scheme of the tightening tool parameter optimization system and method based on deep learning, the invention comprises the steps that the data acquisition module acquires the adjusted operation parameters and operation performance in real time and uploads the operation parameters and operation performance to the cloud through an industrial Internet of things IIoT platform,
The real-time collected adjusted operation parameters comprise a moment value Tact,t and a tightening angle thetaact,t, and the collected operation parameter data set is set as Prun,Prun＝{(Tact,t,θact,t) to t=1, 2.
The collected operation performance comprises:
The operation effect Qtight,t is that whether the tightening quality reaches the standard,
If Tact,t∈[Tmin,Tmax and θact,t∈[θmin,θmax ], then Qtight,t =1, otherwise Qtight,t =0, where Tmin,Tmax is the allowed torque range, θmin,θmax is the allowed tightening angle range,
Execution parameters including actual torque value deviation Δtt and angle deviation Δθt:
ΔTt＝Tact,t-T*,Δθt＝θact,t-θ*, wherein T*,θ* is a target torque value and a target angle;
The collected working performance data set is Pperf:
Pperf＝{(Qtight,t,ΔTt,Δθt)∣t=1,2,...,N};
Integrating the operation parameters and the job performance data into a data set D, d= { (Prun,Pperf) }, wherein D is a complete data packet after the single job is completed;
The integrated data is uploaded to the cloud through an industrial Internet of things IIoT platform, the uploading frequency is controlled by adopting a time window method, and the data is uploaded to the cloud through a IIoT platform.
As a preferable scheme of the tightening tool parameter optimization system and method based on deep learning, the invention comprises the steps of automatically adjusting the tightening parameters by combining real-time working conditions based on the optimal tightening parameters by the process parameter optimization module,
Assuming that the real-time data set after cleaning, normalization and feature extraction is X= { X1,x2,…,xN }, wherein Xi represents the data vector of the ith sampling point, Xi＝[Ti,vi,hi,RT,Rh,rT,v ], wherein Ti is a moment value, vi is a speed value, hi is material hardness, RT,Rh is a temperature and humidity change rate, and RT,v is the correlation of moment and speed;
Converting X to batch format Xbatch：Xbatch＝{X1,X2,…,XM through a network connection, where M is the number of batches, each batch Xm containing a fixed amount of time step data;
Analyzing input time series data through a deep learning RNN model, predicting optimal tightening parameters, receiving input batch data Xbatch by the RNN, and calculating characteristics of a time series through recursion propagation of a hidden state, wherein a calculation formula is ht＝f(Wh·ht-1+Wx·xt+bh), ht is a hidden state vector of a time step t, Wh is a weight matrix of the hidden state, Wx is a weight matrix of the input data, Xt is an input data vector of a current time step, bh is a bias term, and f (·) is an activation function;
The final output of RNN predicts the optimal tightening parameters through the full connection layer, the prediction formula is:
wherein,For the predicted tightening parameter of the time step t, Wy is a weight matrix of the output layer, and by is a bias term of the output layer;
According to the output tightening parameters of the analysis module, the tightening parameters of the real-time working condition dynamic adjustment tool are combined, and the adjustment parameters are corrected according to the working condition function, wherein the correction formula is as follows:
wherein T*,v* is the adjusted torque and speed,For the moment and the speed predicted by the analysis module, deltaTenv is the current temperature deviation, deltahmat is the deviation of the current material hardness, alpha and beta are sensitivity coefficients of the environment and the material working condition to moment and speed adjustment, and finally the adjusted optimal tightening parameters are used as output and transmitted to the tool execution module.
As a preferable scheme of the tightening tool parameter optimization system and method based on deep learning, the method comprises the following steps of analyzing load balance and work sequence among tightening tools by using a deep learning transducer model,
The integrated operation parameters and operation performance data are received through an industrial internet of things IIoT platform, and uploading data Dcloud of all tightening tools are received from a IIoT platform:
Wherein Dcloud is all tool data integrated in the cloud, K is the number of tightening tools, and decoding (-) is a decoding function;
extracting the operation parameters and performance data of each tool, wherein the extraction formula is as follows:
Xk＝{(Tact,t,θact,t,Qtight,t,ΔTt,Δθt)∣t＝1,2,…,Nk},
Wherein Xk is the data set of the kth tool, and Nk is the total number of sampling points of the kth tool;
Sorting the tool data into a time series form Sk＝Sequence(Xk, wherein Sk is time series data of a kth tool, sequence (-) represents formatting the multidimensional data in time steps;
analyzing the multi-tool data by using a transducer model, mining rules of load unbalance and operation sequence optimization, converting the time series data Sk into an embedded representation, and converting the embedded representation into a conversion formula:
wherein,For initial embedding matrix, the embedded (·) is a feature embedding function, the multidimensional feature is mapped into a fixed dimension vector, posEnc (·) is a position code;
the transducer captures the correlation between tools through a multi-head attention mechanism, the capture process being expressed as:
wherein q=zkWQ,K＝ZkWK,V＝zkWV is a query, key and value matrix, obtained by projection of a weight matrix WQ,Wk,WV, dk is the dimension of the key, and Attention (·) is the Attention score;
After L-layer transform coding, the final output is obtained
The data for all tools are combined, expressed as:
The output is used for representing the Z prediction inter-tool load balance state Bk and the recommended job sequence Qk, and the prediction formula is as follows:
Wherein, Bk is the load score of the kth tool, Ok is the recommended order of the kth tool, and MLPB,MLPO is the multi-layer perceptron.
The invention relates to a tightening tool parameter optimization system and a tightening tool parameter optimization method based on deep learning, wherein the steps of updating an optimization algorithm based on an analysis result and issuing an optimized model to an edge calculation module are as follows,
According to the load balancing and operation sequence analysis results, updating parameters and strategies of an edge calculation model, and expressing the optimized tightening task allocation strategy as follows:
wherein Wk is the task weight of the kth tool, task allocation is adjusted according to Wk, and Bk is the load score of the kth tool;
The tool operation order is reordered, and the ordering formula is:
O=Sort(O1,O2,…,OK),
Wherein O is the ordered tool operation sequence,
And updating model parameters of the edge computing module by combining load optimization and sequence adjustment results, wherein an updating formula is as follows:
Wherein, thetanew is the updated model parameter, thetaold is the old model parameter, eta is the learning rate,And issuing the updated model to an edge calculation module as a loss function.
The invention has the beneficial effects that:
The method and the system combine a deep learning technology and an industrial Internet of things IIoT platform, solve the problems of poor real-time performance, insufficient coordination, lack of data closed loop and the like in the traditional method, remarkably improve production efficiency and tightening quality, firstly, clean, normalize and extract features of real-time data through an edge calculation module, predict optimal tightening parameters through an RNN model, dynamically adjust parameters by combining real-time working conditions, quickly adapt to complex environment changes, secondly, a cloud end analyzes load balancing and operation sequences among multiple tools through a transducer model, optimizes task allocation strategies, meanwhile, the IIoT platform achieves optimal parameter sharing among multiple tools, remarkably improves coordination efficiency of a production line, and in addition, a feedback module collects tightening operation performance and operation parameters, uploads the data to the cloud end, continuously analyzes the coordination and historical data among the tools through the deep learning model, dynamically updates an optimization algorithm, and improves adaptability and continuous improvement capability of the system to the dynamic working conditions.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic diagram of a framework of a tightening tool parameter optimization system in example 1.
FIG. 2 is an optimization flow chart of the adaptive tightening tool process parameter optimization system of example 1;
fig. 3 is a flowchart showing the operation of the process parameter optimization module in embodiment 1.
Detailed Description
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, but the present invention may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present invention is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the invention. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Embodiment 1, referring to fig. 1, 2 and 3, this embodiment provides a tightening tool parameter optimization system based on deep learning, comprising:
the data acquisition module is used for acquiring operation data and environmental parameters of the tightening tool in real time;
the operation data comprises tightening torque, tightening speed and tightening times,
The environmental parameters include hardness and roughness of the material, temperature and humidity and vibration in the production environment,
The edge calculation module is used for receiving the real-time data transmitted by the data acquisition module, performing preliminary processing on the data, rapidly analyzing the real-time data by using the deep learning model RNN, predicting the optimal tightening parameters of a single tool in real time, and transmitting the result to the process parameter optimization module;
the analysis module is used for predicting the cooperative operation parameters among multiple tools by utilizing a deep learning transducer algorithm based on the data transmitted by the edge calculation module, outputting an optimization strategy and updating the model parameters of the process parameter optimization module and the edge calculation module;
The process parameter optimization module is used for receiving the optimization parameters output by the analysis module and the edge calculation module, adjusting the tightening parameters in combination with real-time working conditions, wherein the real-time working conditions comprise the current tool state and the environmental change dynamics;
the feedback module is used for collecting the operation performance and the adjusted operation parameters in real time after the tightening operation is completed, uploading the operation parameters to the cloud end and transmitting the operation parameters to the analysis module through the industrial Internet of things IIoT platform;
the operation parameters comprise operation effects and execution parameters;
The operation effect comprises whether the tightening quality reaches the standard, and the execution parameters comprise a torque value and a tightening angle deviation;
The industrial Internet of things IIoT platform is used as a central point for connecting all modules, is responsible for real-time transmission and sharing of data, receives the data uploaded by the feedback module and transmits the data to the analysis module,
The industrial Internet of things IIoT platform is used for optimizing parameter sharing among multiple tools, and adjusting the operation sequence, load balancing and task distribution.
The embodiment also provides a tightening tool parameter optimization method based on deep learning, which comprises the following steps:
Step S1, a data acquisition module acquires operation parameters of a tightening tool, including moment, speed, tightening material hardness and temperature and humidity, and the operation parameters are cleaned, standardized and feature extracted in an edge calculation module to generate real-time data;
cleaning, normalizing and extracting features in an edge computing module, generating real-time data,
Detecting abnormal values based on distribution analysis, and detecting whether the collected data points xi belong to a reasonable interval or not, wherein a monitoring formula is xi∈[Q1-1.5·IQR,Q3 +1.5·iqr, wherein Q1 is the 1 st quartile of data, Q3 is the 3 rd quartile of data, and iqr=q3-Q1 is a quartile distance;
and (3) carrying out outlier processing on the detected outlier by adopting a linear interpolation method, wherein the processing formula is as follows:
Wherein xi is an outlier, and xi-1,xi+1 is a normal data point before and after the outlier;
ZScore normalization was performed for each data point, and the normalization formula was:
Wherein zi is normalized data, xi is cleaned original data, mux is the mean value of the data, and sigmax is the standard deviation of the data;
Feature extraction is performed in the normalized data, and the following features are extracted for the moment T (T) and the speed v (T), including:
Average value of
Maximum tmax:
Tmax＝max(Ti),
root mean square value Trms:
wherein Ti is the moment value of the ith time point, and N is the total number of sampling points;
for the dynamic change rate of temperature and humidity calculation, the calculation formula is as follows:
Wherein ht,Tt is the humidity and temperature at the current time point, Rh,RT is the rate of change of the humidity and temperature, and Δt is the sampling time interval;
and calculating the Pearson correlation coefficient of the moment and the speed, wherein the calculation formula is as follows:
Wherein Ti,vi is the moment and the speed at the ith time point respectively,As an average of torque and speed, rT,v is a linear dependence,
Specifically, the cleaned data is normalized by ZScore to eliminate unit influence, and time series characteristics, dynamic environment change rate and inter-variable correlation are extracted.
S2, the edge calculation module sends real-time data to the analysis module through network connection, the analysis module adopts a deep learning RNN model to predict optimal tightening parameters, including moment and speed, and the process parameter optimization module automatically adjusts the tightening parameters based on the optimal tightening parameters and in combination with real-time working conditions;
the process parameter optimization module automatically adjusts the tightening parameters based on the optimal tightening parameters in combination with real-time working conditions,
Assuming that the real-time data set after cleaning, normalization and feature extraction is X= { X1,x2,…,xN }, wherein Xi represents the data vector of the ith sampling point, Xi＝[Ti,vi,hi,RT,Rh,rT,v ], wherein Ti is a moment value, vi is a speed value, hi is material hardness, RT,Rh is a temperature and humidity change rate, and RT,v is the correlation of moment and speed;
Converting X to batch format Xbatch：Xbatch＝{X1,X2,…,XM through a network connection, where M is the number of batches, each batch Xm containing a fixed amount of time step data;
Analyzing input time series data through a deep learning RNN model, predicting optimal tightening parameters, receiving input batch data Xbatch by the RNN, and calculating characteristics of a time series through recursion propagation of a hidden state, wherein a calculation formula is ht＝f(Wh·ht-1+Wx·xt+bh), ht is a hidden state vector of a time step t, Wh is a weight matrix of the hidden state, Wx is a weight matrix of the input data, Xt is an input data vector of a current time step, bh is a bias term, and f (·) is an activation function;
The final output of RNN predicts the optimal tightening parameters through the full connection layer, the prediction formula is:
wherein,For the predicted tightening parameter of the time step t, Wy is a weight matrix of the output layer, and by is a bias term of the output layer;
According to the output tightening parameters of the analysis module, the tightening parameters of the real-time working condition dynamic adjustment tool are combined, and the adjustment parameters are corrected according to the working condition function, wherein the correction formula is as follows:
wherein T*,v* is the adjusted torque and speed,For the moment and the speed predicted by the analysis module, deltaTenv is the current temperature deviation, deltahmat is the deviation of the current material hardness, and alpha and beta are sensitivity coefficients of the environment and the material working condition to moment and speed adjustment;
specifically, the edge calculation module sends data to the analysis module, the optimal tightening parameters are predicted by utilizing the time sequence feature extraction capacity of the RNN model, and the tightening parameters output by the dynamic adjustment of the real-time working conditions are combined, so that the real-time requirements can be met, and the complex working conditions can be dynamically adapted.
Step S3, the tightening tool performs tightening operation according to the adjusted tightening parameters, the data acquisition module acquires the adjusted operation parameters and operation performance in real time, and the operation parameters and operation performance are uploaded to the cloud end through an industrial Internet of things IIoT platform;
The data acquisition module acquires the adjusted operation parameters and operation performance in real time and uploads the operation parameters and operation performance to the cloud through an industrial Internet of things IIoT platform,
The real-time collected adjusted operation parameters comprise a moment value Tact,t and a tightening angle thetaact,t, and the collected operation parameter data set is set as Prun,Prun＝{(Tact,t,θact,t) to t=1, 2.
The collected operation performance comprises:
The operation effect Qtight,t is that whether the tightening quality reaches the standard,
If Tact,t∈[Tmin,Tmax and θact,t∈[θmin,θmax ], then Qtight,t =1, otherwise Qtight,t =0, where Tmin,Tmax is the allowed torque range, θmin,θmax is the allowed tightening angle range,
Execution parameters including actual torque value deviation Δtt and angle deviation Δθt:
ΔTt＝Tact,t-T*,Δθt＝θact,t-θ*, wherein T*,θ* is a target torque value and a target angle;
The collected working performance data set is Pperf:
Pperf＝{(Qtight,t,ΔTt,Δθt)∣t=1,2,...,N};
Integrating the operation parameters and the job performance data into a data set D, d= { (Prun,Pperf) }, wherein D is a complete data packet after the single job is completed;
Uploading the integrated data to the cloud end through an industrial Internet of things IIoT platform, controlling the uploading frequency by adopting a time window method, and uploading the data to the cloud end through a IIoT platform;
Specifically, real-time operation parameters and operation performance data obtained through the data acquisition module are integrated and encoded, and then uploaded to the cloud according to a time window by utilizing an industrial Internet of things IIoT platform, so that data integrity and timeliness are guaranteed.
Step S4, the cloud receives the adjusted operation parameters and the adjusted working performance through an industrial Internet of things IIoT platform, and uses a deep learning transducer model to analyze load balancing and operation sequences among tightening tools;
the steps of analyzing load balancing and work order among tightening tools using a deep learning transducer model are,
The integrated operation parameters and operation performance data are received through an industrial internet of things IIoT platform, and uploading data Dcloud of all tightening tools are received from a IIoT platform:
Wherein Dcloud is all tool data integrated in the cloud, K is the number of tightening tools, and decoding (-) is a decoding function;
extracting the operation parameters and performance data of each tool, wherein the extraction formula is as follows:
Xk＝{(Tact,t,θact,t,Qtight,t,ΔTt,Δθt)∣t＝1,2,…,Nk},
Wherein Xk is the data set of the kth tool, and Nk is the total number of sampling points of the kth tool;
Sorting the tool data into a time series form Sk＝Sequence(Xk, wherein Sk is time series data of a kth tool, sequence (-) represents formatting the multidimensional data in time steps;
analyzing the multi-tool data by using a transducer model, mining rules of load unbalance and operation sequence optimization, converting the time series data Sk into an embedded representation, and converting the embedded representation into a conversion formula:
wherein,For initial embedding matrix, the embedded (·) is a feature embedding function, the multidimensional feature is mapped into a fixed dimension vector, posEnc (·) is a position code;
the transducer captures the correlation between tools through a multi-head attention mechanism, the capture process being expressed as:
wherein q=zkWQ,K＝ZkWk,V＝ZkWV is a query, key and value matrix, obtained by projection of a weight matrix WQ,WK,WV, dk is the dimension of the key, and Attention (·) is the Attention score;
After L-layer transform coding, the final output is obtained
The data for all tools are combined, expressed as:
The output is used for representing the load balance state Bk and the recommended operation order Ok among Z prediction tools, and the prediction formula is as follows:
wherein, Bk is the load score of the kth tool, Ok is the recommended order of the kth tool, and MLPB,MLPO is a multi-layer perceptron;
updating an optimization algorithm based on the analysis result, and issuing the optimized model to an edge calculation module,
According to the load balancing and operation sequence analysis results, updating parameters and strategies of an edge calculation model, and expressing the optimized tightening task allocation strategy as follows:
wherein Wk is the task weight of the kth tool, task allocation is adjusted according to Wk, and Bk is the load score of the kth tool;
The tool operation order is reordered, and the ordering formula is:
O=Sort(O1,O2,…,OK),
Wherein O is the ordered tool operation sequence,
And updating model parameters of the edge computing module by combining load optimization and sequence adjustment results, wherein an updating formula is as follows:
Wherein, thetanew is the updated model parameter, thetaold is the old model parameter, eta is the learning rate,Issuing the updated model to an edge calculation module as a loss function;
Specifically, the load and the operation sequence among the tools are analyzed by using a transducer model, task allocation and model parameters are dynamically adjusted by combining an optimization result, and the task allocation and the model parameters are issued to an edge calculation module through a IIoT platform, so that the efficiency and the synergy of the whole production line are ensured.
It should be noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention may be modified or substituted without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered in the scope of the claims of the present invention.