Industrial spraying workpiece defective product factor intelligent detection method based on graph neural network model

Abstract

The invention discloses an industrial spraying workpiece defective product factor intelligent detection method based on a graph neural network model, and the method comprises the steps: obtaining various parameters in a workpiece production process, including a carriage environment parameter, a drying chamber environment parameter, a paint quality parameter and a spraying equipment parameter; each workpiece corresponds to one node, and constructing a feature vector of each workpiece node based on various parameters in the production process; aggregating and updating the feature vector of each workpiece node based on a graph neural network model, and performing high-dimensional feature representation on the updated feature vectors; based on the high-dimensional features of the workpiece, optimizing parameters of the graph neural network model by adopting a back propagation algorithm; and predicting defective product factors of the workpiece based on the optimized graph neural network model, and outputting types of the defective product factors. The method is suitable for real-time detection of an industrial production line, data of each workpiece is collected and input into the graph neural network model, and rapid and real-time defective product factor detection is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of workpiece detection, and in particular relates to an intelligent detection method for defective factors of industrial spraying workpieces based on a graph neural network model. Background Art

[0002] In modern industrial manufacturing, spraying technology is widely used in the surface treatment of parts in the automotive, aviation, electronic equipment and other industries. Spraying technology can not only improve the appearance of parts, but also play a functional protective role such as corrosion resistance and wear resistance. However, the quality of spraying technology is easily affected by a variety of external factors, such as environmental humidity, temperature, dust, spraying equipment parameters, etc. These uncertain factors will cause uneven color, particles, scratches and other undesirable phenomena on the sprayed workpiece, thus affecting the final quality of the product.

[0003] At present, the quality inspection methods of industrial spraying workpieces mostly rely on manual experience analysis or simple data statistics methods. These methods have problems such as insufficient detection accuracy, low efficiency, and high labor costs in practical applications. Especially in large-scale production lines, the limitations of manual analysis are more prominent. With the rapid development of intelligent manufacturing and automation technology, automated cause detection solutions based on artificial intelligence (AI) technology have gradually become the mainstream trend to solve the problem of defective product cause analysis.

[0004] Limitations of manual analysis: Traditional detection of defective spray workpieces mainly relies on manual experience analysis, which is not only time-consuming and inefficient, but also often results in inconsistencies due to human subjective judgment. In addition, manual analysis is difficult to meet the real-time needs in mass production environments, especially when multiple environmental factors work together to cause defective products. The analysis is difficult and subtle factors are easily overlooked.

[0005] Limitations of traditional statistical methods: Automated cause analysis methods based on statistics and simple rules can assist manual inspection in some cases, but these methods usually rely on manually designed features and are difficult to cope with complex production environments. Since the quality of sprayed workpieces is affected by a variety of environmental and equipment parameters, simple statistical methods are difficult to capture the complex correlations between these parameters, resulting in low accuracy in cause identification.

[0006] The influence of multiple parameters is difficult to effectively model: The quality of sprayed workpieces is not only limited by the quality of the surface coating, but also affected by multiple factors such as environmental humidity, temperature, air pressure, spraying pressure, etc. Traditional machine learning models have difficulty processing such multi-dimensional parameters, and it is even more difficult to capture the complex relationship between parameters, resulting in the inability to accurately identify the root cause of defective products.

[0007] In recent years, Graph Neural Network (GNN), as an important model in the field of deep learning, has been widely used in many fields. In order to solve the above problems, it is urgent to propose an intelligent detection method for defective factors of industrial spraying workpieces based on the graph neural network model. Summary of the invention

[0008] In order to solve the above technical problems, the present invention proposes an intelligent detection method for defective factors of industrial spraying workpieces based on a graph neural network model to solve the problems existing in the above-mentioned prior art.

[0009] To achieve the above object, the present invention provides an intelligent detection method for defective factors of industrial spraying workpieces based on a graph neural network model, comprising the following steps:

[0010] Obtain various parameters during the workpiece production process, including carriage environment parameters, drying room environment parameters, paint quality parameters and spraying equipment parameters;

[0011] Each workpiece corresponds to a node, and the feature vector of each workpiece node is constructed based on various parameters in the production process;

[0012] Aggregating and updating the feature vector of each workpiece node based on the graph neural network model, and performing high-dimensional feature representation on the updated feature vector;

[0013] Based on the high-dimensional features of the workpiece, a back-propagation algorithm is used to optimize the parameters of the graph neural network model;

[0014] Based on the optimized graph neural network model, the defective factors of the workpiece are predicted and the types of defective factors are output.

[0015] Optionally, the cabin environment parameters include cabin humidity, cabin temperature, cabin dust particle size, cabin air pressure, cabin air flow and cabin electrostatic voltage;

[0016] The drying chamber environmental parameters include drying chamber humidity, drying chamber temperature, drying chamber airflow, drying chamber dust particle size, drying chamber air pressure, drying chamber electrostatic voltage and drying chamber drying time;

[0017] The paint quality parameters include paint brand, paint generation time and paint to thinner ratio;

[0018] The spraying equipment parameters include spray gun spraying pressure, nozzle size, spraying angle, spraying speed and spraying distance.

[0019] Optionally, the formula for aggregating and updating the feature vector of each workpiece node based on the graph neural network is as follows:

[0020]

[0021] in, represents the feature representation of node i at the kth layer, N(i) is the set of neighbor nodes of node i, and W (k) and is a learnable weight matrix, σ is a nonlinear activation function, and d i and d j are the degrees of node i and node j respectively.

[0022] Optionally, based on the high-dimensional features of the workpiece, the process of optimizing the parameters of the graph neural network model using a back propagation algorithm includes:

[0023] The high-dimensional features of the artifacts are fed back to the reverse learning layer to generate a new representation of the original data, based on which the aggregation and update process of the feature vector is guided.

[0024] Optionally, the formula for guiding the aggregation and updating process of the feature vector based on the new representation is as follows:

[0025]

[0026] Among them, P is the target matrix; s ij is the incidence matrix element; ||·|| F represents the Frobenius norm; S is the auxiliary matrix; is the original data matrix; is a low-dimensional embedding matrix; is the cluster center matrix.

[0027] Optionally, it also includes performing feature weight analysis on high-dimensional features to obtain the influence of various parameters on defective product factors.

[0028] Optionally, the loss function of the graph neural network model is a cross entropy loss function, and the formula is as follows:

[0029]

[0030] Among them, y i is the true label, is the predicted value of the model.

[0031] The present invention also provides an electronic device, comprising: a memory and a processor; the memory is used to store programs; the processor is used to execute the programs to implement the various steps of the intelligent detection method of defective factors of industrial spray workpieces based on the graph neural network model.

[0032] The present invention also provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the various steps of the intelligent detection method of defective factors of industrial spraying workpieces based on a graph neural network model are implemented.

[0033] Compared with the prior art, the present invention has the following advantages and technical effects:

[0034] Multi-factor correlation modeling: Traditional methods find it difficult to handle multi-dimensional production parameters. The present invention uses a graph neural network model to integrate multiple factors such as ambient humidity, temperature, and spray equipment parameters into a graph structure, capturing the complex relationship between the parameters in the graph, thereby more comprehensively reflecting the interaction of multiple factors in the production process.

[0035] Automated and high-precision cause identification: Through the aggregation and learning mechanism of the graph neural network model, the present invention can automatically analyze the information of each workpiece and its neighboring nodes, and deeply explore the potential causes of defective products. Compared with manual detection and traditional data statistical methods, the graph neural network model has significant advantages in extracting and associating complex features, which helps to improve the accuracy of cause detection.

[0036] Real-time and high efficiency: The present invention can be deployed in real time on the production line to perform real-time analysis on the quality factors of each workpiece, which not only greatly improves the detection efficiency, but also can promptly identify and feedback the key factors that may lead to defective products, providing support for timely adjustment of production parameters, thereby effectively improving production quality.

[0037] Adapt to complex production environments: The node and edge structure of the graph neural network model is suitable for the complex environment in industrial manufacturing. It can not only process the characteristics of individual workpieces, but also reflect the production association relationship between workpieces. Especially in large-scale production scenarios, this method can maintain stability and robustness and adapt to complex and changing production conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0039] Figure 1 This is a schematic flow chart of an intelligent detection method for defective factors of industrial spraying workpieces based on a graph neural network model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0041] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0042] Embodiment 1

[0043] like Figure 1 As shown, this embodiment provides an intelligent detection method for defective factors of industrial spraying workpieces based on a graph neural network model, including the following steps:

[0044] Obtain various parameters during the workpiece production process, including carriage environment parameters, drying room environment parameters, paint quality parameters and spraying equipment parameters;

[0045] Each workpiece corresponds to a node, and the feature vector of each workpiece node is constructed based on various parameters in the production process;

[0046] Aggregating and updating the feature vector of each workpiece node based on the graph neural network model, and performing high-dimensional feature representation on the updated feature vector;

[0047] Based on the high-dimensional features of the workpiece, a back-propagation algorithm is used to optimize the parameters of the graph neural network model;

[0048] Based on the optimized graph neural network model, the defective products of the workpiece are predicted and the type of defective products is output.

[0049] It is feasible to first number and classify the key variables in the production process in order to accurately model and detect defective products. It is mainly divided into four categories of parameters: cabin environment parameters, drying room environment parameters, paint quality parameters, and spraying equipment parameters. The cabin environment parameters include cabin humidity, cabin temperature, cabin dust particle size, cabin air pressure, cabin air flow, and cabin electrostatic voltage; the drying room environment parameters include drying room humidity, drying room temperature, drying room air flow, drying room dust particle size, drying room air pressure, drying room electrostatic voltage, and drying room drying time; paint quality parameters include paint brand, paint generation time, and the ratio of paint to thinner; spraying equipment parameters include spray gun spray pressure, nozzle size, spray angle, spray speed, and spray distance.

[0050] Specifically, the environmental conditions in the carriage (i.e., the assembly line area where the workpiece is located) have a direct impact on the spraying quality. The environmental parameters of carriage 1, carriage 2, ... carriage n are as follows, and the design variable number is C ij , where i represents the carriage number, j represents the environmental parameter, expressed as the ambient humidity C i1 , Temperature C i2 , dust particle size C i3, air pressure C i4 ,, airflow C i5 , electrostatic voltage C i6 The environmental conditions of the drying room directly affect the curing effect of the coating. The environmental parameters of the drying room are numbered as D j , where j represents the parameter type, which is represented by ambient humidity D1, temperature D2, airflow D3, dust particle size D4, air pressure D5, electrostatic voltage D6, and drying time D7. The quality of paint affects the final effect of the sprayed workpiece. The paint quality parameter number is P j , expressed as paint brand P1, production time P2, paint and thinner ratio P3. The operating parameters of the spraying equipment are important factors that directly affect the spraying results. The spraying equipment parameter number is E j , expressed as spray gun spray pressure E1, nozzle size E2, spray angle E3, spray speed E4, and spray distance E5.

[0051] As a specific embodiment, this embodiment integrates a real-time data acquisition system, uses sensor networks and wireless communication technology to dynamically collect multi-dimensional data of assembly line workpieces, and inputs it into a graph neural network model for online detection, thereby realizing real-time quality monitoring on the production line.

[0052] This embodiment uses the following method to represent the various types of parameter data collected:

[0053] Node: Each workpiece corresponds to a node, and the attributes of the node include parameters such as environment, equipment, paint, etc. related to the workpiece.

[0054] Edge: The edge connecting different workpieces represents the relationship between the workpieces in the same production batch. The influence between workpieces can be modeled by different carriages, spraying equipment, etc.

[0055] Node features: feature vector X of each workpiece node i It consists of various parameters of the workpiece:

[0056] X i =[C i1 ,C i2 ,C i3 ,C i4 ,C i5 ,C i6 ,D1,D2,D3,D4,D5,D6,D7,P1,P2,P3,E1,E2,E3,E4,E5];

[0057] Edge features: Edge features represent the mutual influence between adjacent workpieces and can be modeled based on the process connection relationship between carriages or spraying batches.

[0058] Graph construction:

[0059] Each workpiece is modeled as a node in a graph G = (V, E), where V is a set of nodes and E is a set of edges. The edge (i, j) ∈ E (i, j) between workpieces indicates some kind of association between workpieces, such as workpieces from the same batch or in the same carriage. The existence and weight of edges can be based on different physical connections or production processes.

[0060] This embodiment jointly models multiple parameters involved in the spraying process, such as ambient humidity, temperature, airflow, equipment parameters, etc., to construct a dynamic graph structure. Each workpiece node not only has its own characteristics, but can also be combined with the characteristic information of neighboring nodes for comprehensive analysis.

[0061] Adjacency matrix A: The adjacency matrix A represents the connection relationship between nodes. ij Indicates whether there is an edge connection between artifacts i and j.

[0062]

[0063] Based on the adjacency relationship between artifacts, the graph neural network (GNN) aggregates and updates the features of nodes through graph convolution operations. The representation of nodes in each layer is updated by the features of neighboring nodes. In the kth layer, the representation of node i is Updated by the following formula:

[0064]

[0065] in, represents the feature representation of node i at the kth layer, N(i) is the set of neighbor nodes of node i, and W (k) and is a learnable weight matrix, σ is a nonlinear activation function (such as ReLU), and d i and d j are the degrees of node i and node j respectively.

[0066] Node Embedding and Back Propagation:

[0067] After multiple layers of graph convolution, the feature vector of the workpiece node It contains multi-level information and is embedded in high-dimensional space.

[0068] The embedding layer Z is used to represent the node features after multi-layer convolution into high-dimensional features. These high-dimensional features can characterize the overall quality status of the workpiece. After embedding, the representation of node i is:

[0069]

[0070] During the back propagation process, the graph learning layer generates a new graph S, which is combined with the original graph A to form an updated graph A′. The updated graph and input features are then fed into the GNN layer for feature learning. The node embedding Z generated by the GNN layer is further fed into the reverse learning layer to generate a new representation ZP of the original data for L RGL and L RL Guided graph learning layer. The dotted line indicates the path of the loss.

[0071]

[0072] Among them, P is the target matrix, which is the representation of the input data; P i is the predicted value of the i-th sample; s ij is an element of the correlation matrix, indicating the similarity or correlation between sample i and sample j; ||·|| F represents the Frobenius norm, which measures the size of the matrix; S is the auxiliary matrix, which is used to adjust the calculation of the correlation degree. is the original data matrix, each row represents the features of a data point; Is a low-dimensional embedding matrix, which means that the data Features mapped to low-dimensional space; is the cluster center matrix.

[0073] Defective product detection and classification:

[0074] After the forward propagation of the graph neural network, by inputting the type of defective products, the output layer will classify the defective products based on the node embedding features and analyze the output state point [C i1 ,C i2 ,C i3 ,C i4 ,C i5 ,C i6 , D1, D2, D3, D4, D5, D6, D7, P1, P2, P3, E1, E2, E3, E4, E5] are represented by black and white. The ones close to black indicate that the correlation between the defective product type and this node is strong, and those close to white indicate that the correlation is not strong. The goal of the model is to predict the relevant factors of each workpiece being a defective product and give the specific correlation factors of the defective type (such as uneven color, particles, scratches, etc.).

[0075] Classification output: The embedding representation of the node is used through the fully connected layer and the softmax function to classify the defective product.

[0076]

[0077] Among them, y i is the node result of node i, W fis the weight of the classification layer, b f is the bias term, Represents the feature representation of the i-th sample at the K-th layer.

[0078] Loss function: The model uses the cross entropy loss function to measure the gap between the predicted results and the true labels:

[0079]

[0080] Among them, y i is the true label, is the predicted value of the model and N is the number of samples.

[0081] Result output and optimization:

[0082] After the forward propagation and back propagation of the model, the model finally outputs the associated factors of whether each workpiece is a defective product, that is, the factor results that cause this category of defective products, and the production process can be further optimized based on the results. The specific steps include:

[0083] Defective product category factor prediction: The model outputs the prediction results, showing the associated factors of each workpiece with problems such as uneven color, particles, scratches, etc.

[0084] Feature weight analysis: By analyzing the high-dimensional features embedded in the nodes, it is possible to identify which environmental or equipment parameters have a greater impact on defective products.

[0085] The key point of the present invention is to detect the spraying quality of workpieces in real time by combining the complex environment and equipment parameters in the industrial spraying process through the graph neural network model. The key point is that it can effectively process multidimensional data in industrial production lines, especially the correlation between workpieces and the dynamic changes of environmental parameters, thereby realizing the automation of spraying quality and intelligent detection of defective product factors.

[0086] The specific key points are as follows:

[0087] 1. Data modeling and detection method based on graph neural network model: By constructing the multi-dimensional production process data of the workpiece (including environmental parameters, equipment parameters, paint quality, etc.) into a graph structure, the graph neural network is used to perform convolution aggregation on the node (workpiece) features, and the complex relationship between the workpiece and the neighboring nodes is learned layer by layer, finally realizing the intelligent detection of defective factors of the workpiece spraying quality.

[0088] 2. Multi-parameter joint modeling and dynamic graph structure: Multiple parameters involved in the spraying process (such as ambient humidity, temperature, airflow, equipment parameters, etc.) are jointly modeled to construct a dynamic graph structure. Each workpiece node not only has its own characteristics, but can also be combined with the characteristic information of neighboring nodes for comprehensive analysis.

[0089] 3. Classification and prediction of defective products in industrial production: Through the embedding layer and back-propagation mechanism of the graph neural network, the model can identify defective phenomena in the workpiece in real time, including surface defects such as uneven color, particles, scratches, etc., and classify and predict different types of defective products by factors.

[0090] 4. Intelligent feedback and optimization based on environmental and equipment data: By correlating and analyzing the collected environmental data, equipment parameters and workpiece spraying quality, the system can feedback which parameters affect the quality of the workpiece based on the learning results of the GNN model, and make optimization suggestions, automatically adjust production parameters and reduce the defective rate.

[0091] 5. Real-time data acquisition and online detection system: The real-time data acquisition system is integrated, and the multi-dimensional data of the assembly line workpieces is dynamically collected by using sensor networks and wireless communication technologies, and the data is input into the GNN model for online detection to realize real-time quality monitoring on the production line.

[0092] 6. Embedding layer and back propagation optimization of the GNN model: The high-dimensional features of the workpiece are extracted through the GNN embedding layer, and the back propagation algorithm optimizes the model parameters, gradually improving the accuracy of workpiece quality prediction and the generalization ability of the model, ensuring that the system can accurately judge the quality of the sprayed workpiece.

[0093] Embodiment 2

[0094] The present invention also provides an electronic device, comprising: a memory and a processor; the memory is used to store programs; the processor is used to execute the programs to implement the various steps of the intelligent detection method of defective factors of industrial spray workpieces based on the graph neural network model.

[0095] Embodiment 3

[0096] The present invention also provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the various steps of the intelligent detection method of defective factors of industrial spraying workpieces based on a graph neural network model are implemented.

[0097] The above are only preferred specific implementations of the present application, but the protection scope of the present application 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 in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An intelligent detection method for defective factors of industrial spraying workpieces based on a graph neural network model, characterized in that: The following steps are involved: Obtain various parameters during the workpiece production process, including carriage environment parameters, drying room environment parameters, paint quality parameters and spraying equipment parameters; Each workpiece corresponds to a node, and the feature vector of each workpiece node is constructed based on various parameters in the production process; Aggregating and updating the feature vector of each workpiece node based on the graph neural network model, and performing high-dimensional feature representation on the updated feature vector; Based on the high-dimensional features of the workpiece, a back-propagation algorithm is used to optimize the parameters of the graph neural network model; Based on the optimized graph neural network model, the defective factors of the workpiece are predicted and the types of defective factors are output.

2. The intelligent detection method for defective factors of industrial spraying workpieces based on the graph neural network model according to claim 1 is characterized in that: The cabin environment parameters include cabin humidity, cabin temperature, cabin dust particle size, cabin air pressure, cabin air flow and cabin static voltage; The drying chamber environmental parameters include drying chamber humidity, drying chamber temperature, drying chamber airflow, drying chamber dust particle size, drying chamber air pressure, drying chamber electrostatic voltage and drying chamber drying time; The paint quality parameters include paint brand, paint generation time and paint to thinner ratio; The spraying equipment parameters include spray gun spraying pressure, nozzle size, spraying angle, spraying speed and spraying distance.

3. The intelligent detection method for defective factors of industrial spraying workpieces based on the graph neural network model according to claim 1 is characterized in that: The formula for aggregating and updating the feature vector of each workpiece node based on the graph neural network is as follows: in, represents the feature representation of node i at the kth layer, N(i) is the set of neighbor nodes of node i, and W (k) and is a learnable weight matrix, σ is a nonlinear activation function, and d i and d j are the degrees of node i and node j respectively.

4. The intelligent detection method for defective factors of industrial spraying workpieces based on the graph neural network model according to claim 1 is characterized in that: Based on the high-dimensional features of the workpiece, the process of optimizing the parameters of the graph neural network model using the back propagation algorithm includes: The high-dimensional features of the artifacts are fed back to the reverse learning layer to generate a new representation of the original data, based on which the aggregation and update process of the feature vector is guided.

5. The intelligent detection method for defective factors of industrial spraying workpieces based on the graph neural network model according to claim 4 is characterized in that: The formula for guiding the aggregation and updating process of feature vectors based on the new representation is as follows: Where P is the target matrix; s ij is the incidence matrix element; ||·|| F represents the Frobenius norm; S is the auxiliary matrix; is the original data matrix; is a low-dimensional embedding matrix; is the cluster center matrix.

6. The intelligent detection method for defective factors of industrial spraying workpieces based on the graph neural network model according to claim 4 is characterized in that: It also includes feature weight analysis of high-dimensional features to obtain the influence of various parameters on defective product factors.

7. The intelligent detection method for defective factors of industrial spraying workpieces based on graph neural network model according to claim 1 is characterized in that: The loss function of the graph neural network model is the cross entropy loss function, and the formula is as follows: Among them, y i is the true label, is the predicted value of the model.

8. An electronic device, characterized in that: include: Memory and processor; The memory is used to store programs; the processor is used to execute the programs to implement the various steps of the intelligent detection method for defective factors of industrial spray workpieces based on a graph neural network model as described in any one of claims 1 to 7.

9. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the intelligent detection method for defective factors of industrial spraying workpieces based on the graph neural network model as described in any one of claims 1 to 7 are implemented.

Images