Laser surface strengthening intelligent control system based on edge calculation

Abstract

The invention discloses a laser surface strengthening intelligent control system based on edge computing, which comprises a cloud computing layer, an edge computing layer and a terminal layer, and is characterized in that the terminal layer sends collected process parameters and workpiece surface images to the edge computing layer and then transmits the collected process parameters and workpiece surface images to the cloud computing layer; the cloud computing layer sends the constructed workpiece surface defect diagnosis model and the laser surface strengthening remanufacturing model to the edge computing layer for training; and the edge calculation layer monitors and diagnoses the surface defects of the workpiece in real time by using the trained model, calculates laser surface strengthening working parameters, and performs laser surface treatment on the workpiece through a laser strengthening robot. By constructing the workpiece surface defect diagnosis model, real-time monitoring and diagnosis of the product quality in the numerical control machining process are achieved, meanwhile, real-time analysis is conducted based on diagnosis data, product remanufacturing is achieved through laser surface strengthening treatment, and real-time data processing and intelligent decision making in the laser surface strengthening process are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of CNC machine tool processing and laser surface strengthening, and in particular to an intelligent control system for laser surface strengthening based on edge computing. Background Art

[0002] As the key equipment of intelligent manufacturing, CNC machine tools occupy a core position in modern manufacturing. According to statistics, my country currently has more than 10 million CNC machine tools. In order to promote the modernization and intelligence of the manufacturing industry and maximize the energy efficiency of high-end CNC machining, my country has formulated a series of strategic plans.

[0003] my country produces more than 450 million tons of waste every year, including a large number of scrapped parts. A large part of them are scrapped due to quality defects during CNC machining, which also causes huge economic losses. Consciously reducing these wastes of resources can significantly reduce the use of new materials and reduce carbon emissions.

[0004] In the field of CNC machining, the current big data collection methods and deep learning algorithms have achieved a series of research results in fault diagnosis, machining parameter optimization and machining scheduling optimization. However, the main problems are: it is difficult to accurately diagnose workpiece machining quality problems in real time during CNC machining, and it is impossible to carry out timely and effective processing to reduce the amount of scrapped workpieces; the machining data acquisition rate and training efficiency are low, and the computing architecture efficiency is low.

[0005] In view of this, how to conduct timely and effective diagnosis of workpiece quality problems based on the monitoring and collection of processing data during the CNC machining process, and to effectively handle them based on the diagnostic analysis to improve workpiece quality and reduce product defective rate has important economic value and strategic significance. Summary of the invention

[0006] The purpose of the present invention is to provide a laser surface strengthening intelligent control system based on edge computing to solve the above problems.

[0007] To achieve the above object, the present invention adopts the following technical solutions:

[0008] An edge computing-based laser surface strengthening intelligent control system is used to repair surface defects of workpieces in real time during CNC machining. It includes a cloud computing layer, an edge computing layer, and a terminal layer, wherein:

[0009] The cloud computing layer includes a cloud model construction unit, a task sequence calculation unit, a task allocation optimization unit and a data monitoring module; the cloud model construction unit is used to construct a workpiece surface defect diagnosis model and a laser surface strengthening remanufacturing model; the task sequence calculation unit is used to count the data transmission volume, edge computing processing time, and edge layer computing utilization rate, the data transmission volume is the total amount of data transmitted from the edge computing layer to the cloud computing layer, the edge computing processing time is the time required for data transmission and algorithm processing under a specific network environment, and the edge layer computing utilization rate is the utilization efficiency of the edge computing layer; the task allocation optimization unit is used to construct an edge computing task allocation optimization algorithm, use the fruit fly algorithm to solve the best task allocation method and schedule the computing tasks; the data monitoring module is used to determine whether the processing conditions have changed, if not, schedule the edge computing layer to use the deep learning algorithm for model training, if so, schedule the edge computing layer to use the transfer learning algorithm for model training;

[0010] The edge computing layer includes a data analysis unit and an edge-side model training unit, wherein the data analysis unit is used to analyze the workpiece surface image to obtain workpiece surface defect information, wherein the workpiece surface defect information includes at least defect position, defect amount, and defect area geometric information, and the edge-side model training unit is used to train a workpiece surface defect diagnosis model and a laser surface strengthening remanufacturing model, wherein the input of the workpiece surface defect diagnosis model is the processing power, and the output thereof is a workpiece defect degree diagnosis result, and the input of the laser surface strengthening remanufacturing model is a workpiece defect degree diagnosis result and workpiece surface defect information, and the output thereof is a laser surface strengthening working parameter, wherein the laser surface strengthening working parameter includes at least a laser surface strengthening path, a laser power, and a scanning speed;

[0011] The terminal layer includes a CNC machine tool, a process parameter acquisition module, a power data acquisition module, an image acquisition module, and a laser strengthening robot. The process parameter acquisition module is used to acquire process parameters related to CNC machining, and the process parameters at least include machining parameters, workpiece type, and workpiece material category. The power data acquisition module is used to acquire the machining power of the CNC machining process in real time. The image acquisition module is used to acquire the surface image of the workpiece. The laser strengthening robot includes a controller, a mechanical arm, and a laser. The controller is used to control the operation of the mechanical arm, the laser, and the cooling device. The mechanical arm is used to transmit according to the working trajectory. The laser is used to perform laser surface strengthening treatment on defective parts of the workpiece.

[0012] The terminal layer sends the collected process parameters and workpiece surface images to the edge computing layer through the MQTT protocol, and the edge computing layer transmits the process parameters and workpiece surface defect information to the cloud computing layer through the MQTT protocol; the cloud computing layer sends the constructed workpiece surface defect diagnosis model, laser surface strengthening remanufacturing model and original training data set to the edge computing layer for training, and the edge computing layer uses deep learning and transfer learning algorithms to perform model training; the edge computing layer uses the trained workpiece surface defect diagnosis model to monitor and diagnose the workpiece surface defects in real time during the CNC machining process, and the edge computing layer uses the trained laser surface strengthening remanufacturing model to calculate the laser surface strengthening working parameters in real time, and sends them to the controller of the laser strengthening robot to perform laser surface treatment on the workpiece.

[0013] Preferably, the workpiece surface defect diagnosis model is:

[0014] x a =α·ΔP i +β·Q i

[0015]

[0016] Among them, x a is the workpiece defect degree diagnosis result, α and β are weight coefficients, ΔP i is the processing power increment between the i-th processing time point and the i-1-th processing time point, Q i is the defect amount of the workpiece at the i-th processing time point, c, d, g, h, n are model parameters, K cutter is the tool aging coefficient.

[0017] Preferably, the laser surface strengthening remanufacturing model is:

[0018] y m =f(x m )·T

[0019]

[0020] Among them, y m is the output laser surface strengthening path in the working coordinate system of the laser strengthening robot, x m =(v1, v2, ...v j …v M ), which is a set of scanning nodes selected on the defective area of ​​the input workpiece, M is the total number of scanning nodes, T is the three-dimensional translation matrix, y n is the output laser power and scanning speed, x n (ΔP i , Q i) is the input power increment and defect amount, W is the weight matrix, b is the bias parameter, is the activation function, and the activation function adopts one of the ReLU function, PReLU function, Softmax function, and Swish function.

[0021] Preferably, a constraint condition is introduced for the laser surface strengthening remanufacturing model, and the shortest laser surface strengthening path is selected as the optimal result. The length calculation formula of the laser surface strengthening path is:

[0022]

[0023] Where L is the length of the laser surface strengthening path, d(v j , v j+1 ) is the distance from the j+1th scan node to the jth scan node.

[0024] Preferably, the scanning node is selected in the following manner:

[0025] For the workpiece surface image, the geometric information of the workpiece contour and the defect area is obtained by extracting image features;

[0026] Construct a three-dimensional model of the workpiece based on the geometric information of the workpiece contour and the geometric information of the defect area;

[0027] Divide the defect area into multiple equivalent planes based on geometric information of the defect area;

[0028] The area of ​​each equivalent plane is calculated. If the area of ​​the equivalent plane is less than the preset area threshold, the center point of the equivalent plane is selected as the scanning node. If the area of ​​the equivalent plane is greater than the preset area threshold, the number of scanning node selections is calculated based on the preset area threshold. The product of the number of scanning node selections and the preset area threshold is less than and closest to the area of ​​the equivalent plane. Based on the number of scanning node selections, the corresponding scanning nodes are uniformly selected on the equivalent plane.

[0029] Preferably, the uniform selection of corresponding scanning nodes is achieved by the following method: based on the number of selected scanning nodes, the equivalent plane is divided into a corresponding number of grids, and the center point of each grid is used as a scanning node.

[0030] Preferably, the uniform selection of corresponding scanning nodes is achieved by the following method:

[0031] The hash function is used to map the equivalent plane to a uniformly distributed point set, and the scanning nodes are uniformly selected on the point set. The model formula is:

[0032] x=hash(key)mod(x max -x min )

[0033] y=hash(key+1)mod(y max -y min )

[0034] z=hash(key+2))mod(z max -z min )

[0035] Among them, hash is a hash function and key is a variable parameter.

[0036] Preferably, the method of using the fruit fly algorithm to solve the optimal task allocation method is specifically implemented by the following method:

[0037] The data transmission volume, edge computing processing time, and edge layer computing utilization are normalized and used as optimization targets.

[0038] Initialize the number of fruit fly swarms and randomly initialize the positions of the fruit fly swarms;

[0039] Define the fitness function based on the optimization objective:

[0040] F = argmin(e·Q tmn +f·T op +h·D op )

[0041] Among them, F is the fitness function, Q tmn is the normalized value of data transmission volume, T op is the normalized value of edge computing processing time, D op is the normalized value of edge layer computing utilization;

[0042] Based on the smell search method, the crossover and mutation process is introduced to calculate the fitness function value F;

[0043] Construct an iterative mechanism for visual search. If the fitness function value of an individual fruit fly is better than the fitness function value of the center of the fruit fly group, the position of the individual fruit fly is used as the new search center.

[0044] Through iterative optimization, the individual fruit flies with the best fitness function value are determined as the optimal task allocation method.

[0045] Preferably, the model training using the transfer learning algorithm is specifically implemented by the following method: analyzing the difference in feature distribution between source domain data and target domain data, introducing alignment weights for data alignment to achieve cross-domain transfer of data; using the Tensorflow deep learning architecture to build a deep learning algorithm for forward computing training; using parameter transfer learning and retaining most of the training parameters, performing back propagation calculations to optimize feature distribution, using the maximum mean difference to measure the difference in probability distribution between the source domain and the target domain, combining deep learning training cross entropy to characterize the training error results, and using regularized maximum mean difference and cross entropy to construct the optimization target of the algorithm; after training, the final CNC machining fault diagnosis model is obtained, which is used for real-time machining fault diagnosis during the machining process;

[0046] The forward computation training is implemented by the following method:

[0047] Randomly initialize the model’s weight matrix and bias parameters;

[0048] Input source domain data and target domain data into each layer of the neural network and output the diagnosis results;

[0049] The error between the output diagnosis result and the true result is calculated using the loss function;

[0050] The back propagation calculation is implemented by the following method:

[0051] Calculate the gradient of the loss function with respect to each model parameter. Based on the obtained gradient of each model parameter, use the gradient descent algorithm to update the model parameters of the neural network. Specifically:

[0052] Select an initial point θ0, define θ as the model parameter, and θ0 as the initialization parameter;

[0053] Choose a loss function L(θ) that measures the difference between the predictions given the model parameters θ and the actual observations;

[0054] Determine an appropriate learning rate η, which controls the step size of updating the parameters at each iteration;

[0055] Iterative update, for each training sample i, calculate the gradient of the loss function with respect to θ:

[0056]

[0057] Among them, g i is the gradient, x i is the input data, y i is the output result.

[0058] Preferably, the workpiece surface defect diagnosis model and laser surface strengthening remanufacturing model are implemented using a long short-term memory neural network combined with a convolutional neural network, which includes a convolutional layer, a pooling layer, an LSTM layer and a fully connected layer. The convolutional layer and the pooling layer are used for feature extraction, the LSTM layer is used to process time series data, and the fully connected layer is used to output the final diagnosis results or laser surface strengthening process parameters.

[0059] After adopting the above technical solution, the present invention has the following beneficial effects compared with the background technology:

[0060] 1. The present invention constructs a workpiece surface defect diagnosis model to perform effective workpiece surface defect diagnosis based on the machining power increment and the workpiece defect situation, thereby enabling real-time monitoring and diagnosis of product quality during CNC machining. At the same time, by constructing a laser surface strengthening remanufacturing model, the workpiece can be repaired in a timely manner through laser surface strengthening treatment when product quality abnormalities occur, thereby effectively improving workpiece quality and reducing product defective rate.

[0061] 2. Based on the edge computing architecture, the present invention realizes real-time data processing and intelligent decision-making in the process of laser surface strengthening during the complex operation of CNC machining, significantly improves the intelligence level of laser surface strengthening, reduces human intervention, realizes efficient online control, and finally realizes full closed-loop control of the machining process, thereby improving production efficiency and product quality.

[0062] 3. The present invention constructs an edge computing allocation optimization algorithm, uses the fruit fly algorithm to solve the optimal task allocation method and perform task scheduling, thereby improving the computational efficiency of quality diagnosis and remanufacturing planning in the CNC machining process.

[0063] Instruction Manual

[0064] Figure 1 Schematic diagram of the system architecture of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0066] Example

[0067] The technical idea of ​​the present invention is based on digital twin technology. The application of digital twin technology in the field of CNC machine tools can simulate the operating status and processing of machine tools in a virtual environment by creating an accurate digital copy of the physical machine tool, so that the performance, efficiency and potential problems of the machine tool can be comprehensively predicted and analyzed. Emerging technologies such as intelligent sensor measurement systems, big data analysis, deep learning algorithms, data preprocessing, data visualization, edge computing, and processing compensation technology based on digital twins, combined with traditional mechanical processing mechanism models, can effectively realize effective diagnosis, analysis and intervention of CNC machining.

[0068] Edge computing aims to move computing power and storage resources closer to users or data sources. "Edge" is relative to "center". Edge computing aims to make data processing as close as possible to the device that generates data, reducing the response delay between the device and the server. Edge computing greatly reduces the dependence on centralized cloud servers by processing data on edge devices, sensors and edge nodes. This distributed computing model can provide faster, more efficient and secure computing services. The combination of edge computing and CNC machining can improve the intelligence level and response speed of CNC machine tools by processing data near the data source. By deploying edge computing nodes on CNC machine tools, machine tool operation data can be collected and processed in real time, thereby realizing real-time monitoring and optimization of the machining process. At the same time, combined with big data analysis and artificial intelligence technology, it can provide intelligent decision-making for CNC machining operations and realize a more efficient and intelligent machining and manufacturing process.

[0069] The present invention discloses an edge computing-based laser surface strengthening intelligent control system, which is used for real-time repair and remanufacturing of workpiece surface defects during CNC machining. The control system includes a cloud computing layer 100, an edge computing layer 200, and a terminal layer 300, wherein:

[0070] The cloud computing layer 100 includes a cloud model building unit 110, a task sequence calculation unit 120, a task allocation optimization unit 130 and a data monitoring module 140, wherein:

[0071] The cloud model building unit 110 is used to build a workpiece surface defect diagnosis model and a laser surface strengthening remanufacturing model. The workpiece surface defect diagnosis model and the laser surface strengthening remanufacturing model are implemented using a long short-term memory neural network combined with a convolutional neural network, which includes a convolution layer, a pooling layer, an LSTM layer and a fully connected layer. The convolution layer and the pooling layer are used for feature extraction, the LSTM layer is used to process time series data, and the fully connected layer is used to output the final diagnosis result or laser surface strengthening process parameters.

[0072] The workpiece surface defect diagnosis model is:

[0073] x a =α·ΔPi +β·Q i

[0074]

[0075] Among them, x a is the workpiece defect degree diagnosis result, α and β are weight coefficients, ΔP i is the processing power increment between the i-th processing time point and the i-1-th processing time point, Q i is the defect amount of the workpiece at the i-th processing time point, c, d, g, h, n are model parameters, K cutter is the tool aging coefficient, A i is the tool wear at the i-th processing time point. i and tool wear A i The calculation principle is the same.

[0076] The present invention has been found in previous studies that tool wear is the main cause of workpiece surface defects. When workpiece quality defects occur under tool wear, the tool cutting force is also abnormal, which in turn causes fluctuations in machining power. Therefore, the relationship between the machining power increment and the workpiece quality defects can be established through the incremental change of machining power. In order to better understand the present invention, the correlation model between the machining power increment and the tool wear degree involved in the present invention is described in detail:

[0077] Through the wear curve of the cutting tool, the cutting tool processing life cycle is divided into three intervals: running-in period, steady-state period, and failure period. The wear curve of the cutting tool is used to characterize the change of wear amount with processing time during the processing life cycle. For the running-in period and steady-state period, a correlation model between tool wear and tool cutting force is established. The correlation model between tool wear and tool cutting force is:

[0078]

[0079] Where i is the i-th tool wear state in the cutting tool processing life cycle, A i is the wear amount of the i-th tool under wear state, which is obtained through image analysis, F i is the cutting force corresponding to the i-th tool wear data, F0 is the initial cutting force of the cutting tool, and a and b are constants determined based on the wear curve of the cutting tool.

[0080] Matlab software is used to estimate the cutting force during the machining life cycle, analyze the nonlinear relationship between cutting tool wear and cutting force increment, and define the tool aging coefficient K cutter , K cutter The solution model is:

[0081]

[0082] Among them, K cutter is the tool aging coefficient, c and d are model parameters.

[0083] Based on K cutter , determine ΔP i With K cutter The laser surface strengthening remanufacturing model is:

[0084] y m =f(x m )·T

[0085] Among them, y m is the output laser surface strengthening path in the working coordinate system of the laser strengthening robot 350, x m =(v1, v2, ...v j …v M ), which is a set of scanning nodes selected on the defective area of ​​the input workpiece, M is the total number of scanning nodes, T is the three-dimensional translation matrix, y n is the output laser power and scanning speed, x n (ΔP i , Q i ) is the input power increment and loss,

[0086] W is the weight matrix, b is the bias parameter, is the activation function, and the activation function adopts one of the ReLU function, PReLU function, Softmax function, and Swish function. Constraints are introduced for the above laser surface strengthening remanufacturing model, and the shortest laser surface strengthening path is selected as the optimal result. The calculation formula for the length of the laser surface strengthening path is:

[0087]

[0088] Where L is the length of the laser surface strengthening path, D(v j , v j+1 ) is the distance from the j+1th scan node to the jth scan node.

[0089] The basic work of the laser surface strengthening path is to select the scanning nodes of the defective area of ​​the workpiece. Since the defective area is usually irregular in shape and not an ideal plane, and there is complexity and diversity, how to reasonably select the scanning nodes becomes particularly important for the subsequent laser surface strengthening treatment. In this embodiment, the scanning node is selected as follows: for the workpiece surface image, the geometric information of the workpiece contour and the geometric information of the defective area are obtained by image feature extraction; based on the geometric information of the workpiece contour and the geometric information of the defective area, a three-dimensional model of the workpiece is constructed; based on the geometric information of the defective area, it is divided into multiple equivalent planes; the area of ​​each equivalent plane is calculated, if the area of ​​the equivalent plane is less than the preset area threshold, the center point of the equivalent plane is selected as the scanning node, if the area of ​​the equivalent plane is greater than the preset area threshold, the number of scanning node selections is calculated based on the preset area threshold, the product of the number of scanning node selections and the preset area threshold is less than and closest to the area of ​​the equivalent plane, and the corresponding scanning nodes are uniformly selected on the equivalent plane based on the number of scanning node selections. The uniform selection of scanning nodes involved here can be based on the number of scanning nodes selected, the equivalent plane can be divided into a corresponding number of grids, and the center point of each grid is used as a scanning node; or the equivalent plane can be mapped to a uniformly distributed point set by using a hash function, and the scanning nodes are uniformly selected on the point set. The model formula is:

[0090] x=hash(key)mod(x max -x min )

[0091] y=hash(key+1))mod(y max -y min )

[0092] z=hash(key+2)mod(z max -z min )

[0093] Among them, hash is a hash function and key is a variable parameter.

[0094] The task sequence calculation unit 120 is used to count the data transmission volume, edge computing processing time, and edge layer computing utilization rate. The data transmission volume is the total amount of data transmitted from the edge computing layer 200 to the cloud computing layer 100, the edge computing processing time is the time required for data transmission and algorithm processing under a specific network environment, and the edge layer computing utilization rate is the utilization efficiency of the edge computing layer 200. The task allocation optimization unit 130 is used to construct an edge computing task allocation optimization algorithm, and use the fruit fly algorithm to solve the best task allocation method and schedule the computing tasks. In this embodiment, the fruit fly algorithm is used to solve the best task allocation method, which is specifically implemented by the following method:

[0095] The data transmission volume, edge computing processing time, and edge layer computing utilization are normalized and used as optimization targets.

[0096] Initialize the number of fruit fly swarms and randomly initialize the positions of the fruit fly swarms;

[0097] Define the fitness function based on the optimization objective:

[0098] F = argmin(e·Q tmn +f·T op +h·D op )

[0099] Among them, F is the fitness function, Q tmn is the normalized value of data transmission volume, T op is the normalized value of edge computing processing time, D op is the normalized value of edge layer computing utilization;

[0100] Based on the smell search method, the crossover and mutation process is introduced to calculate the fitness function value F;

[0101] Construct an iterative mechanism for visual search. If the fitness function value of an individual fruit fly is better than the fitness function value of the center of the fruit fly group, the position of the individual fruit fly is used as the new search center.

[0102] Through iterative optimization, the individual fruit flies with the best fitness function value are determined as the optimal task allocation method.

[0103] The data monitoring module 140 is used to determine whether the processing conditions have changed. If not, the edge computing layer 200 is scheduled to use the deep learning algorithm for model training. If so, the edge computing layer 200 is scheduled to use the transfer learning algorithm for model training. The use of the transfer learning algorithm for model training involved here is specifically implemented by the following methods: analyzing the feature distribution differences between the source domain data and the target domain data, introducing alignment weights for data alignment to achieve cross-domain transfer of data; using the Tensorflow deep learning architecture to build a deep learning algorithm for forward computing training; using parameter transfer learning and retaining most of the training parameters, performing back propagation calculations to optimize feature distribution, using the maximum average difference to measure the difference in probability distribution between the source domain and the target domain, combining deep learning training cross entropy to characterize the training error results, and using regularized maximum average difference and cross entropy to construct the optimization target of the algorithm; after training, the final CNC machining fault diagnosis model is obtained for real-time machining fault diagnosis during the machining process.

[0104] Forward computation training is achieved through the following methods:

[0105] Randomly initialize the model's weight matrix and bias parameters;

[0106] Input source domain data and target domain data into each layer of the neural network and output the diagnosis results;

[0107] The error between the output diagnosis result and the true result is calculated using the loss function;

[0108] The back propagation calculation is implemented by the following method:

[0109] Calculate the gradient of the loss function with respect to each model parameter. Based on the obtained gradient of each model parameter, use the gradient descent algorithm to update the model parameters of the neural network. Specifically:

[0110] Select an initial point θ0, define θ as the model parameter, and θ0 as the initialization parameter;

[0111] Choose a loss function L(θ) that measures the difference between the predictions given the model parameters θ and the actual observations;

[0112] Determine an appropriate learning rate η, which controls the step size of updating the parameters at each iteration;

[0113] Iterative update, for each training sample i, calculate the gradient of the loss function with respect to θ:

[0114]

[0115] Among them, g i is the gradient, x i is the input data, y i is the output result.

[0116] The edge computing layer 200 includes a data analysis unit 210 and an edge-side model training unit 220, wherein:

[0117] The data analysis unit 210 is used to analyze the workpiece surface image to obtain workpiece surface defect information, which includes at least defect position, defect amount, and defect area geometric information. The workpiece surface defect information is obtained in the following manner:

[0118] 1) Obtaining a workpiece surface quality image, and extracting workpiece feature information using an image feature extraction method, the workpiece feature information including workpiece contour features, workpiece surface texture features, workpiece defect area texture features, and workpiece defect area edge features;

[0119] 2) Based on the workpiece feature information, obtain the workpiece geometric information and the position and geometric information of the defect area;

[0120] 3) Calculate the area and depth of the defect area, and then calculate the volume of the defect area as the wear amount.

[0121] The edge-side model training unit 220 is used to train the workpiece surface defect diagnosis model and the laser surface strengthening remanufacturing model. The input of the workpiece surface defect diagnosis model is the processing power, and its output is the workpiece defect degree diagnosis result. The input of the laser surface strengthening remanufacturing model is the workpiece defect degree diagnosis result and the workpiece surface defect information, and the output is the laser surface strengthening working parameters. The laser surface strengthening working parameters at least include the laser surface strengthening path, laser power and scanning speed.

[0122] The terminal layer 300 includes a CNC machine tool 310, a process parameter acquisition module 320, a power data acquisition module 330, an image acquisition module 340, and a laser strengthening robot 350, wherein:

[0123] The process parameter acquisition module 320 is used to collect process parameters related to CNC machining, which include at least machining parameters, workpiece type and workpiece material category. The power data acquisition module 330 is used to collect the machining power of the CNC machining process in real time. The image acquisition module 340 is used to collect the surface image of the workpiece. The process parameter acquisition module 320 and the power data acquisition module 330 send the collected data to the edge layer in real time.

[0124] The laser strengthening robot 350 includes a controller, a robotic arm, and a laser. The controller is used to control the operation of the robotic arm, the laser, and the cooling device. The robotic arm is used to transmit according to the working trajectory. The laser is used to perform laser surface strengthening treatment on defective parts of the workpiece.

[0125] The overall architecture coordination relationship of the present invention is that the terminal layer 300 sends the collected process parameters and workpiece surface images to the edge computing layer 200 through the MQTT protocol, and the edge computing layer 200 transmits the process parameters and workpiece surface defect information to the cloud computing layer 100 through the MQTT protocol; the cloud computing layer 100 sends the constructed workpiece surface defect diagnosis model, laser surface strengthening remanufacturing model and original training data set to the edge computing layer 200 for training, and the edge computing layer 200 uses deep learning and transfer learning algorithms to perform model training; the edge computing layer 200 uses the trained workpiece surface defect diagnosis model to monitor and diagnose the workpiece surface defects in real time during the CNC machining process, and the edge computing layer 200 uses the trained laser surface strengthening remanufacturing model to calculate the laser surface strengthening working parameters in real time, and sends it to the controller of the laser strengthening robot 350 to perform laser surface treatment on the workpiece. In this way, during the workpiece processing, real-time analysis can be performed based on the diagnostic data to control the laser surface strengthening operation to repair the workpiece quality defects. Depending on the CNC processing scenario, the laser surface strengthening operation can be performed simultaneously during the CNC processing process or at the workpiece discharge station. The laser surface strengthening operation does not require the CNC equipment to be shut down.

[0126] It can be seen from the above description that the present invention, by constructing a workpiece surface defect diagnosis model, performs effective workpiece surface defect diagnosis based on the machining power increment and workpiece defect situation, and can realize real-time monitoring and diagnosis of product quality during CNC machining; at the same time, by constructing a laser surface strengthening remanufacturing model, the workpiece is repaired in time through laser surface strengthening treatment when product quality abnormalities occur, effectively improving the workpiece quality and reducing the product defect rate. In addition, based on the edge computing architecture, the present invention realizes real-time data processing and intelligent decision-making in the laser surface strengthening process during the complex operation process of CNC machining, significantly improving the intelligence level of laser surface strengthening, reducing human intervention, and realizing efficient online control, and finally realizing full closed-loop control of the machining process, improving production efficiency and product quality.

[0127] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A laser surface strengthening intelligent control system based on edge computing, used for real-time repair of workpiece surface defects during CNC machining, characterized in that: It includes cloud computing layer, edge computing layer and terminal layer, among which: The cloud computing layer includes a cloud model construction unit, a task sequence calculation unit, a task allocation optimization unit and a data monitoring module; the cloud model construction unit is used to construct a workpiece surface defect diagnosis model and a laser surface strengthening remanufacturing model; the task sequence calculation unit is used to count the data transmission volume, edge computing processing time, and edge layer computing utilization rate, the data transmission volume is the total amount of data transmitted from the edge computing layer to the cloud computing layer, the edge computing processing time is the time required for data transmission and algorithm processing under a specific network environment, and the edge layer computing utilization rate is the utilization efficiency of the edge computing layer; the task allocation optimization unit is used to construct an edge computing task allocation optimization algorithm, use the fruit fly algorithm to solve the best task allocation method and schedule the computing tasks; the data monitoring module is used to determine whether the processing conditions have changed, if not, schedule the edge computing layer to use the deep learning algorithm for model training, if so, schedule the edge computing layer to use the transfer learning algorithm for model training; The edge computing layer includes a data analysis unit and an edge-side model training unit, wherein the data analysis unit is used to analyze the workpiece surface image to obtain workpiece surface defect information, wherein the workpiece surface defect information includes at least defect position, defect amount, and defect area geometric information, and the edge-side model training unit is used to train a workpiece surface defect diagnosis model and a laser surface strengthening remanufacturing model, wherein the input of the workpiece surface defect diagnosis model is the processing power, and the output thereof is a workpiece defect degree diagnosis result, and the input of the laser surface strengthening remanufacturing model is a workpiece defect degree diagnosis result and workpiece surface defect information, and the output thereof is a laser surface strengthening working parameter, wherein the laser surface strengthening working parameter includes at least a laser surface strengthening path, a laser power, and a scanning speed; The terminal layer includes a CNC machine tool, a process parameter acquisition module, a power data acquisition module, an image acquisition module, and a laser strengthening robot. The process parameter acquisition module is used to acquire process parameters related to CNC machining, and the process parameters at least include machining parameters, workpiece type, and workpiece material category. The power data acquisition module is used to acquire the machining power of the CNC machining process in real time. The image acquisition module is used to acquire the surface image of the workpiece. The laser strengthening robot includes a controller, a mechanical arm, and a laser. The controller is used to control the operation of the mechanical arm and the laser. The mechanical arm is used to transmit according to the working trajectory. The laser is used to perform laser surface strengthening treatment on defective parts of the workpiece. The terminal layer sends the collected process parameters and workpiece surface images to the edge computing layer through the MQTT protocol, and the edge computing layer transmits the process parameters and workpiece surface defect information to the cloud computing layer through the MQTT protocol; the cloud computing layer sends the constructed workpiece surface defect diagnosis model, laser surface strengthening remanufacturing model and original training data set to the edge computing layer for training, and the edge computing layer uses deep learning and transfer learning algorithms to perform model training; the edge computing layer uses the trained workpiece surface defect diagnosis model to monitor and diagnose the workpiece surface defects in real time during the CNC machining process, and the edge computing layer uses the trained laser surface strengthening remanufacturing model to calculate the laser surface strengthening working parameters in real time, and sends them to the controller of the laser strengthening robot to perform laser surface treatment on the workpiece.

2. The laser surface strengthening intelligent control system based on edge computing according to claim 1, characterized in that: The workpiece surface defect diagnosis model is: x a =α·ΔP i +β·Q i Among them, x a is the workpiece defect degree diagnosis result, α and β are weight coefficients, ΔP i is the processing power increment between the i-th processing time point and the i-1-th processing time point, Q i is the defect amount of the workpiece at the i-th processing time point, c, d, g, h, n are model parameters, K cutter is the tool aging coefficient.

3. The laser surface strengthening intelligent control system based on edge computing according to claim 1, characterized in that: The laser surface strengthening remanufacturing model is: y m =f(x m )·T Among them, y m is the output laser surface strengthening path in the working coordinate system of the laser strengthening robot, x m =(v1, v2, ...v j …v M ), which is a set of scanning nodes selected on the defective area of ​​the input workpiece, M is the total number of scanning nodes, T is the three-dimensional translation matrix, y n is the output laser power and scanning speed, x n (ΔP i , Q i ) is the input power increment and defect amount, W is the weight matrix, b is the bias parameter, is the activation function, and the activation function adopts one of the ReLU function, PReLU function, Softmax function, and Swish function.

4. The laser surface strengthening intelligent control system based on edge computing according to claim 3, characterized in that: Constraints are introduced into the laser surface strengthening remanufacturing model, and the shortest laser surface strengthening path is selected as the optimal result. The length calculation formula of the laser surface strengthening path is: Where L is the length of the laser surface strengthening path, D(v j , v j+1 ) is the distance from the j+1th scan node to the jth scan node.

5. The laser surface strengthening intelligent control system based on edge computing according to claim 3, characterized in that: The scanning node is selected as follows: For the workpiece surface image, the geometric information of the workpiece contour and the defect area is obtained by extracting image features; Construct a three-dimensional model of the workpiece based on the geometric information of the workpiece contour and the geometric information of the defect area; Divide the defect area into multiple equivalent planes based on geometric information of the defect area; The area of ​​each equivalent plane is calculated. If the area of ​​the equivalent plane is less than the preset area threshold, the center point of the equivalent plane is selected as the scanning node. If the area of ​​the equivalent plane is greater than the preset area threshold, the number of scanning node selections is calculated based on the preset area threshold. The product of the number of scanning node selections and the preset area threshold is less than and closest to the area of ​​the equivalent plane. Based on the number of scanning node selections, the corresponding scanning nodes are uniformly selected on the equivalent plane.

6. The laser surface strengthening intelligent control system based on edge computing according to claim 5, characterized in that: The uniform selection of corresponding scanning nodes is achieved by the following method: based on the number of selected scanning nodes, the equivalent plane is divided into a corresponding number of grids, and the center point of each grid is used as a scanning node.

7. The laser surface strengthening intelligent control system based on edge computing according to claim 5, characterized in that: The uniform selection of corresponding scanning nodes is achieved by the following method: The hash function is used to map the equivalent plane to a uniformly distributed point set, and the scanning nodes are uniformly selected on the point set. The model formula is: x=hash(key)mod(x max -x min ) y=hash(key+1)mod(y max -y min ) z=hash(key+2))mod(z max -z min ) Among them, hash is a hash function and key is a variable parameter.

8. The laser surface strengthening intelligent control system based on edge computing according to claim 1, characterized in that: The method of using the fruit fly algorithm to solve the optimal task allocation method is specifically implemented by the following method: The data transmission volume, edge computing processing time, and edge layer computing utilization are normalized and used as optimization targets. Initialize the number of fruit fly swarms and randomly initialize the positions of the fruit fly swarms; Define the fitness function based on the optimization objective: F=argmin(e·Q tmn +f·T op +h·D op) Among them, F is the fitness function, Q tmn is the normalized value of data transmission volume, T op is the normalized value of edge computing processing time, D op is the normalized value of edge layer computing utilization; Based on the smell search method, the crossover and mutation process is introduced to calculate the fitness function value F; Construct an iterative mechanism for visual search. If the fitness function value of an individual fruit fly is better than the fitness function value of the center of the fruit fly group, the position of the individual fruit fly is used as the new search center. Through iterative optimization, the individual fruit flies with the best fitness function value are determined as the optimal task allocation method.

9. The laser surface strengthening intelligent control system based on edge computing according to claim 1, characterized in that: The model training using the transfer learning algorithm is specifically implemented by the following method: analyzing the feature distribution difference between the source domain data and the target domain data, introducing alignment weights to perform data alignment to achieve cross-domain data transfer; The Tensorflow deep learning architecture is used to build a deep learning algorithm for forward computing training; parameter transfer learning is used and most of the training parameters are retained, and back propagation calculations are performed to optimize feature distribution. The maximum mean difference is used to measure the difference in probability distribution between the source domain and the target domain. The training error results are characterized by combining deep learning training cross entropy, and the regularized maximum mean difference and cross entropy are used to construct the optimization target of the algorithm; after training, the final CNC machining fault diagnosis model is obtained, which is used for real-time machining fault diagnosis during the machining process; The forward computation training is implemented by the following method: Randomly initialize the model's weight matrix and bias parameters; Input source domain data and target domain data into each layer of the neural network and output the diagnosis results; The error between the output diagnosis result and the true result is calculated using the loss function; The back propagation calculation is implemented by the following method: Calculate the gradient of the loss function with respect to each model parameter. Based on the obtained gradient of each model parameter, use the gradient descent algorithm to update the model parameters of the neural network. Specifically: Select an initial point θ0, define θ as the model parameter, and θ0 as the initialization parameter; Choose a loss function L(θ) that measures the difference between the predictions given the model parameters θ and the actual observations; Determine an appropriate learning rate η, which controls the step size of updating the parameters at each iteration; Iterative update, for each training sample i, calculate the gradient of the loss function with respect to θ: Among them, g i is the gradient, x i is the input data, y i is the output result.

10. The laser surface strengthening intelligent control system based on edge computing according to claim 1, characterized in that: The workpiece surface defect diagnosis model and laser surface strengthening remanufacturing model are implemented using a long short-term memory neural network combined with a convolutional neural network, which includes a convolutional layer, a pooling layer, an LSTM layer and a fully connected layer. The convolutional layer and the pooling layer are used for feature extraction, the LSTM layer is used to process time series data, and the fully connected layer is used to output the final diagnosis result or laser surface strengthening process parameters.

Images