Surface quality management method for stator core mold

Abstract

The invention relates to the technical field of data processing, and provides a surface quality management method for a stator core mold. The method comprises the following steps: performing surface quality detection on a target stator core mold to obtain a quality deviation coefficient; performing multi-dimensional verification based on the deviation coefficient, and outputting a plurality of quality management layers; constructing a teacher model according to hierarchical association; obtaining a superior distillation result through the knowledge distillation channel and the branch optimization model; therefore, a student model is constructed, and stator core mold surface quality management is optimized. The technical problems that a traditional detection method often depends on artificial experience and visual inspection, only the roughness and smoothness of the surface of the mold are concerned, the actual quality of the surface of the mold is difficult to accurately evaluate, and the accuracy and stability of quality management are low are solved. Through comprehensive strategy of multiple links such as the knowledge distillation channel and the quality management teacher model, the technical effects of improving the surface quality of the mold and improving the accuracy and efficiency of quality management are achieved.

Description

Technical Field

[0001] The present application relates to the field of computer science and technology, specifically to the field of data processing technology, and in particular to a surface quality management method for a stator core mold. Background Art

[0002] With the continuous advancement of industrial technology and the rapid development of intelligent manufacturing, the surface quality management of stator core molds has become a key factor affecting product quality and production efficiency. However, the traditional surface quality management methods of stator core molds have many shortcomings. Traditional detection methods mainly rely on manual experience and visual observation, which is highly subjective, inefficient, and difficult to accurately evaluate the actual quality of the mold surface. At the same time, traditional surface quality management methods often only focus on appearance indicators such as the roughness and smoothness of the mold surface, while ignoring more critical factors such as wear resistance and corrosion resistance. These factors are crucial to the service life of the mold and the quality of the stator core. Due to the lack of comprehensive consideration of these key factors, traditional methods cannot effectively prevent problems such as wear and corrosion that may occur during the use of the mold, affecting the overall performance and stability of the mold, and thus making the accuracy and stability of mold quality management low. Summary of the invention

[0003] This application provides a surface quality management method for a stator core mold, aiming to solve the technical problem that traditional detection methods often rely on manual experience and visual observation, only focus on the roughness and smoothness of the mold surface, and find it difficult to accurately evaluate the actual quality of the mold surface, resulting in low accuracy and stability of quality management.

[0004] In view of the above problems, the present application provides a surface quality management method for a stator core mold.

[0005] The present application provides a surface quality management method for a stator core mold, the method comprising: performing surface quality inspection on a target stator core mold to obtain a quality deviation coefficient; performing multi-dimensional verification based on the quality deviation coefficient, and outputting N quality management layers, where N is an integer greater than or equal to 1; connecting the N quality management layers according to a hierarchical association relationship to construct a quality management teacher model; performing distillation optimization on the quality management teacher model through a knowledge distillation channel, the knowledge distillation channel comprising N distillation excellence branches; performing data distillation on the N quality management layers through the N distillation excellence branches to obtain N excellence distillation results; constructing a quality management student model based on the N excellence distillation results, and optimizing the surface quality of the target stator core mold through the quality management student model.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The above-mentioned surface quality management method of a stator core mold detects the surface quality of the target stator core mold and obtains the quality deviation coefficient of the mold surface. Subsequently, the quality deviation coefficient is used to perform multi-dimensional verification analysis, evaluate the quality status of the mold surface from different angles, and output multiple quality management layers. After that, according to the correlation between these quality management layers, they are connected together to construct a quality management teacher model. In order to make the quality management teacher model more efficient and accurate, the knowledge distillation technology is introduced, and the quality management teacher model is distilled and optimized through the knowledge distillation channel. Then, multiple distillation goodness branches in the knowledge distillation channel are used to perform data distillation on multiple quality management layers, and multiple goodness distillation results are obtained by combining the optimization results of the quality management teacher model. Finally, based on these results, a quality management student model is constructed. This quality management student model is then used to optimize the surface quality of the target stator core mold. In this way, it can be ensured that the surface quality of the mold reaches a higher standard, thereby improving the performance and reliability of the stator core.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 It is a schematic flow chart of a surface quality management method of a stator core mold in one embodiment; Figure 2 A schematic diagram of obtaining a quality deviation coefficient of a surface quality management method for a stator core mold in one embodiment. DETAILED DESCRIPTION

[0010] The embodiment of the present application provides a surface quality management method for a stator core mold, aiming to solve the technical problem that traditional detection methods often rely on manual experience and visual observation, only focus on the roughness and smoothness of the mold surface, and find it difficult to accurately evaluate the actual quality of the mold surface, resulting in low accuracy and stability of quality management.

[0011] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0012] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0013] Embodiment 1 like Figure 1 , Figure 2 As shown, the present application provides a surface quality management method for a stator core mold, the method comprising: Conduct surface quality inspection on the target stator core mold to obtain the quality deviation coefficient; With the rapid development of the power industry and the continuous advancement of technology, the surface quality of stator core molds, as the core components of motor manufacturing, plays a vital role in the performance and stability of the motor. However, in the current industrial production environment, the surface quality management of stator core molds faces many challenges.

[0014] In an embodiment of the present application, the system terminal performs surface quality inspection on the target stator core mold, mainly to find out the difference between the mold surface and the ideal state. This process involves testing the roughness, warping, smoothness, and coating uniformity of the mold, and obtaining the corresponding data for each detection direction. Subsequently, the obtained data is calculated for deviation with the corresponding standard, and the calculation results are summarized to obtain the quality deviation coefficient. These coefficients can quantitatively reflect the quality status of the mold surface and help understand whether there are defects, unevenness or other problems on the mold surface. In summary, the surface quality inspection process of the target stator core mold is to find out the potential problems of the target stator core mold and provide a basis for subsequent optimization management.

[0015] Furthermore, the present application provides a method for performing surface quality inspection on a target stator core mold to obtain a quality deviation coefficient, and the method further includes: Traversing the qualified mold parameters of the stator core of the big data, formulating the surface quality inspection standard of the target stator core mold, wherein the surface quality inspection standard includes a preset roughness standard, a preset warpage standard, a preset smoothness standard, and a preset coating uniformity standard; Preferably, the system terminal uses big data to collect parameter data of qualified stator core molds from different manufacturers, production batches and use environments, and classifies and organizes the collected data to ensure the accuracy and completeness of the data. Subsequently, the parameters such as roughness, warpage, smoothness and coating uniformity of many qualified stator core molds are traversed from the collected data. Through the analysis of these parameters, the performance of qualified molds in all aspects is fully understood. Based on these analysis results, surface quality inspection standards for target stator core molds are formulated to clarify the numerical range of each standard. The surface quality inspection standard includes multiple dimensions, such as preset roughness standards, preset warpage standards, preset smoothness standards, and preset coating uniformity standards. These standards not only help the system terminal clarify what level the surface quality of qualified molds should reach, but also provide a clear basis and reference for subsequent mold surface quality inspections. In summary, by analyzing a large amount of data and formulating a comprehensive and detailed surface quality inspection standard, the normal subsequent surface quality inspection of stator core molds can be ensured.

[0016] In the first detection phase, the target stator core mold is detected based on the preset roughness standard to generate a first detection result; In the second detection phase, the target stator core mold is detected based on the preset warpage standard to generate a second detection result; Preferably, when the system terminal detects the surface quality of the target stator core mold, it first performs the first detection stage. In this stage, the roughness data of the target stator core mold is compared with the preset roughness standard one by one. It is determined whether each data point falls within the preset qualified range. For the roughness data falling within the qualified range, it is recorded as qualified roughness. For the roughness data exceeding the qualified range, it is recorded as unqualified roughness. Then the recorded qualified and unqualified roughness are sorted and classified to obtain the first detection result. After the first detection stage is completed, the system terminal starts the second detection stage. In this stage, the warpage data is compared with the preset warpage standard using the same method to detect whether there are defects such as unevenness and scratches, and the specific location, size and severity of qualified warpage and unqualified defects are recorded to generate the second detection result. In summary, these two detection stages respectively carry out accurate detection of the roughness and warpage of the mold and generate corresponding detection results. These results will provide an important basis for subsequent quality analysis and optimization.

[0017] In the third detection stage, the target stator core mold is detected based on the preset smoothness standard to generate a third detection result; In the fourth detection stage, the target stator core mold is detected based on the preset coating uniformity standard to generate a fourth detection result; Preferably, the system terminal uses the same method as the above-mentioned method for obtaining the first test result and the second test result. In the third test stage and the fourth test stage, the preset smoothness standard and the preset coating uniformity standard are used to compare the smoothness and coating uniformity of the target stator core mold, and the relevant data of qualified and unqualified are recorded to obtain the third test result and the fourth test result. Among them, the smoothness refers to the measurement of the smoothness and cleanliness of the mold surface, which directly affects the quality and appearance of the mold. The higher the smoothness, the smoother and cleaner the mold surface is, which is conducive to the related operations of mold surface treatment. Coating uniformity refers to the uniformity of the distribution of the coating on the mold surface, that is, the consistency of the coating thickness or coating amount distribution in the coating area. The better the coating uniformity, the smaller the difference in coating thickness in different parts, thereby improving the service life and appearance quality of the mold. In summary, the third test stage and the fourth test stage respectively carry out accurate and comprehensive tests on the smoothness and coating uniformity of the mold, and generate corresponding test results. These results not only provide the system terminal with the specific performance of the mold in key quality indicators, but also provide strong data support for the subsequent acquisition of quality deviation coefficients.

[0018] Integrate the first test result, the second test result, the third test result, and the fourth test result to perform deviation calculation to generate first mass deviation data, second mass deviation data, third mass deviation data, and fourth mass deviation data; The first mass deviation data, the second mass deviation data, the third mass deviation data, and the fourth mass deviation data are added to the mass deviation coefficient.

[0019] Preferably, after completing the detection of the four key quality indicators of the target stator core mold, namely, roughness, warpage, smoothness and coating uniformity, the system terminal obtains four groups of test results, namely the first test result, the second test result, the third test result and the fourth test result. These results respectively reflect the performance of the mold in different quality dimensions. Subsequently, in order to more accurately evaluate the quality status of the mold, the system terminal performs deviation calculation. Specifically, the system terminal calculates the difference between all unqualified indicators in each test result and the corresponding preset standard, and then calculates the ratio with the corresponding preset standard to obtain multiple roughness deviation coefficients, multiple warpage deviation coefficients, multiple smoothness deviation coefficients and multiple uniformity deviation coefficients. Subsequently, the obtained multiple roughness deviation coefficients, multiple warpage deviation coefficients, multiple smoothness deviation coefficients and multiple uniformity deviation coefficients are sorted and classified to obtain the first quality deviation data, the second quality deviation data, the third quality deviation data and the fourth quality deviation data. These deviation data quantitatively reflect the gap between the mold and the standard in each test indicator. Afterwards, the system terminal adds these four groups of quality deviation data to the quality deviation coefficient. The quality deviation coefficient is a comprehensive indicator that integrates the deviation information of the mold in multiple quality dimensions, thus providing a comprehensive quality assessment perspective for the system terminal. By analyzing the quality deviation coefficient, we can better understand the overall quality status of the mold and find out possible quality problems, so that we can more comprehensively and accurately evaluate the quality status of the target stator core mold and provide data support for subsequent quality management.

[0020] Perform multi-dimensional verification based on the quality deviation coefficient and output N quality management layers, where N is an integer greater than or equal to 1; In one embodiment, the system terminal performs multi-dimensional verification analysis based on the calculated quality deviation coefficient, thereby outputting multiple quality management layers, in order to comprehensively evaluate the quality status of the stator core mold from multiple angles and levels, and ensure that each link is effectively monitored and managed. The process of multi-dimensional verification is actually an in-depth analysis and interpretation of the quality deviation coefficient. Starting from the first quality deviation data, the second quality deviation data, the third quality deviation data, and the fourth quality deviation data, the system terminal draws relevant curves, performs dimensional verification, and assigns different weights to the four drawn curves according to factors such as the severity of the quality problem and the verification results, and then outputs N quality management layers. These quality management layers represent different levels of quality monitoring and management, from basic quality control to advanced quality optimization, and each layer has its specific goals and requirements. By implementing these different levels of quality management, the quality of the stator core mold can be fully controlled to ensure that the requirements of the preset standards are met. In general, multi-dimensional verification based on the quality deviation coefficient and output of multiple quality management layers can help the system terminal better understand the quality status of the mold, find potential problems, and improve the quality of the mold.

[0021] Furthermore, the present application provides multi-dimensional verification based on the quality deviation coefficient, outputting N quality management layers, and the method also includes: Respectively analyzing the change trends of the first mass deviation data, the second mass deviation data, the third mass deviation data, and the fourth mass deviation data to obtain a first mass change curve, a second mass change curve, a third mass change curve, and a fourth mass change curve; Optionally, in the surface quality inspection, the system terminal obtains the first quality deviation data, the second quality deviation data, the third quality deviation data and the fourth quality deviation data, which respectively represent the quality deviation of the stator core mold at different inspection stages and different inspection indicators. In order to more intuitively understand the changing trend of these deviation data, the system terminal further analyzes these data and draws the corresponding quality change curve. First, for the first quality deviation data, the system terminal marks the roughness deviation coefficient in the first quality deviation data in the coordinate system according to the corresponding batch information. Each data point represents the quality deviation value of a batch. Subsequently, the parameters of the curve in each interval are calculated using the cubic spline interpolation method. This involves solving the corresponding linear equations to determine the polynomial coefficients of the curve in each interval. After that, these parameters are used to construct a polynomial function in each interval, and these functions are connected to form an overall spline interpolation curve, that is, the first quality change curve. This curve shows the fluctuation of the quality deviation of the mold in the roughness inspection. By observing the ups and downs and slope of the curve, it can be judged whether the mold quality is stable in this link and whether there are obvious quality problems. Then, the system terminal uses the same method to analyze the second, third and fourth quality deviation data, and draws the second, third and fourth quality change curves respectively. These curves reflect the quality changes of the mold in other detection links and indicators. By comparing the shapes and characteristics of different curves, the quality performance differences of the mold in different aspects can be judged, so as to find out possible quality problems or weak links. In summary, by analyzing the first, second, third and fourth quality deviation data and drawing the corresponding quality change curves, we can fully and intuitively understand the quality changes of the stator core mold in multiple aspects.

[0022] Determine a verification dimension based on the first quality change curve, the second quality change curve, the third quality change curve, and the fourth quality change curve; Testing and recording the quality deviation coefficient according to the verification dimension to generate a multi-dimensional verification result; Optionally, the system terminal performs a detailed characteristic analysis on each quality change curve. This includes analyzing the overall trend of the curve, the location of the peaks and troughs, the slope change of the curve, etc. These characteristics can reflect the changing rules of quality deviation under different detection links and indicators. Based on the curve characteristic analysis, the system terminal identifies the key points on each curve. These key points include the extreme points and inflection points of the curve. Subsequently, by comparing and analyzing the key points of different curves, the verification dimensions are preliminarily determined. These dimensions involve specific production links, raw material types, process parameters, etc., and their performance on different curves is significantly different. After preliminarily determining the verification dimensions, the system terminal adjusts and improves the dimensions based on actual production conditions and historical data. For example, a certain production link can be further split into multiple sub-links, or multiple related process parameters can be combined into one dimension for verification. After refinement and optimization, the system terminal determines the verification dimensions. These dimensions can fully cover the key influencing factors of product quality and provide clear guidance for subsequent quality deviation coefficient test records. Further, the system terminal sets the parameter range, constraint conditions and initial state of each dimension according to the determined verification dimensions to build a simulation environment. This simulation environment can accurately reflect the key factors in the actual production process. Subsequently, the system terminal inputs the quality deviation coefficient into the simulation environment, and the simulation environment performs multiple simulation runs based on the deviation coefficients of multiple indicators in the quality deviation coefficient. In each run, observe and record the changes in the quality deviation coefficient in each dimension. Afterwards, extract the recorded data from the simulation environment, organize them according to the dimensions, and generate multi-dimensional verification results. These results show the specific values ​​and change trends of the quality deviation coefficient in different dimensions, providing the system terminal with rich information to comprehensively judge the product quality status. In summary, this process aims to verify and analyze the quality deviation coefficient from multiple dimensions to obtain more comprehensive and accurate product quality assessment results. This not only helps the system terminal to gain an in-depth understanding of the root causes of product quality problems, but also provides strong data support for subsequent quality management.

[0023] Based on the multi-dimensional verification result, weight allocation is performed according to the first mass change curve, the second mass change curve, the third mass change curve, and the fourth mass change curve to generate a weight configuration result; The hierarchical division is determined according to the weight configuration result to generate the N quality management layers.

[0024] Optionally, the system terminal analyzes the specific performance of the quality deviation coefficient under each dimension of the multi-dimensional verification result, and identifies the key features of each quality change curve, such as trend, fluctuation amplitude, extreme point, etc. According to the quality management requirements, determine the key factors for weight allocation, such as the severity of the quality problem, the scope of influence, the difficulty of solving, etc., and analyze the correlation between these key factors and the four quality change curves. In this way, the system terminal can more accurately determine the importance and priority of the quality problem represented by each curve in the overall quality management, and complete the weight allocation to obtain the weight configuration result. After completing the weight allocation, the system terminal further divides the levels according to these weights, thereby generating multiple quality management layers. For example, a curve with a higher weight value is divided into a higher level to highlight its importance. Subsequently, sorting and classification are performed according to the weight configuration results so that similar curves are classified into the same level, thereby determining the final N quality management layers. Each quality management layer contains quality change curves with similar weights and characteristics to facilitate targeted quality management and control.

[0025] Connect the N quality management layers according to the hierarchical association relationship to build a quality management teacher model; In one embodiment, the system terminal identifies and determines the association relationship between each quality management layer based on the judgment of information flow. Subsequently, the N quality management layers are connected by using the obtained association relationship, and the quality management teacher model is integrated. The various levels are connected to each other through specific rules to ensure the consistency of the format of data exchange. This model not only shows the hierarchical relationship between the quality management layers, but also reveals the interactions and influences between them. Ultimately, this teacher model can be used to guide the practical work of quality management. It can help understand quality problems and assist the system terminal in obtaining goodness distillation results.

[0026] Further, the present application provides a method for connecting the N quality management layers according to the hierarchical association relationship to construct a quality management teacher model, and the method further includes: Determining the hierarchical association relationship between the N quality management layers based on mapping the information flow between the N quality management layers; Preferably, the system terminal determines the hierarchical association relationship between N quality management layers based on mapping the information flow between them, which is a process of clarifying the logical relationship between each layer by analyzing the transmission and interaction of information during parameter adjustment. In this process, the system terminal pays attention to the information flow caused by parameter adjustment. When adjusting a certain parameter, if this adjustment affects other parameters, there is a corresponding relationship between the affected parameter and the adjusted parameter. This corresponding relationship actually reveals the hierarchical association relationship between different quality management layers. By analyzing these corresponding relationships, the mutual influence and dependency between the quality management layers can be clarified. For example, a change in a surface roughness-related parameter of the surface roughness management layer will affect a process in the surface finish management layer that relies on this parameter. Through such information flow analysis, the system terminal can build a more accurate quality management hierarchy. This structure not only reflects the logical relationship between the layers, but also provides an important reference for formulating and adjusting quality management strategies. Therefore, determining the hierarchical association relationship between quality management layers based on the mapping of information flow during parameter adjustment is a key step to deeply understand the mutual influence and interaction between the layers, which helps to implement quality management measures more accurately and improve the overall quality management level.

[0027] Constructing a hierarchical connection rule according to the hierarchical association relationship, and connecting the N quality management layers according to the hierarchical connection rule, wherein the N quality management layers include a surface roughness management layer, a surface warpage management layer, a surface finish management layer, and a surface coating uniformity management layer; The surface roughness management layer, the surface warpage management layer, the surface finish management layer, and the surface coating uniformity management layer are integrated to obtain the quality management teacher model.

[0028] Preferably, constructing hierarchical connection rules according to hierarchical association relationships is to achieve effective connection between N quality management layers by formulating unified data exchange standards. Doing so can avoid duplication of work and the occurrence of information islands, and ensure that each quality management layer can maintain consistency and accuracy during data transmission and sharing. Subsequently, the system terminal closely connects the N quality management layers together by following these hierarchical connection rules, and constructs the framework of the quality management teacher model. This framework can accommodate all management layers and support interaction and collaboration between layers. Afterwards, the functions of the connected quality management layer are integrated into the quality management teacher model to ensure that each function can work together in the model to jointly achieve the quality management goal. Among them, the quality management teacher model includes a surface roughness management layer, a surface warpage management layer, a surface finish management layer, and a surface coating uniformity management layer. In the quality management teacher model, each level can share data in real time, collaborate with each other, and jointly realize comprehensive monitoring and management of product quality. Under the unified data exchange standard, these levels can transmit relevant data about the surface quality of the product in real time, jointly analyze the surface quality of the mold, achieve efficient quality management, and help the system terminal to understand the product quality status more comprehensively. Therefore, formulating unified data exchange standards as hierarchical connection rules is a key step to ensure effective connection and collaboration between various quality management levels, which helps to improve the overall quality management level.

[0029] Furthermore, the present application provides a method for constructing a hierarchical connection rule according to the hierarchical association relationship, and the method further includes: Based on the hierarchical association relationship, the shared data sets of the surface roughness management layer, the surface warpage management layer, the surface finish management layer, and the surface coating uniformity management layer are sequentially extracted; Perform cross-layer collaborative analysis based on the shared data set to construct collaborative framework information; The connection process is sorted out according to the collaborative framework information, and the hierarchical connection rules are formulated.

[0030] Optionally, the system terminal analyzes the hierarchical associations between the various quality management layers. This includes understanding the input-output relationships of each layer, the degree of mutual dependence, etc. By analyzing these associations, it is determined which data is common to each layer, that is, the shared data set. Subsequently, based on the analysis results of the hierarchical associations, the data items of common concern to each layer are identified. These data items may involve basic measurement data on the product surface, historical quality problem records, etc. After that, the data related to the shared data set is extracted from the databases of the surface roughness management layer, the surface warpage management layer, the surface finish management layer, and the surface coating uniformity management layer in turn, and stored in the corresponding location in the shared data set. This shared data set is key information shared between the various management layers. They reflect different aspects of product quality and are also the basis for cross-layer collaborative analysis.

[0031] After obtaining the shared data set, the system terminal conducts an in-depth analysis of the shared data set to reveal the relationship and interaction between each management level in the quality control process. Subsequently, based on the results of the cross-layer collaborative analysis, the collaborative mechanism between each management level is designed to determine the specific implementation methods of the collaborative mechanism, including data sharing process, collaborative decision-making mechanism, problem feedback channel, etc., to eliminate duplication of work, break information silos, and achieve resource sharing and efficient collaboration. After that, combined with the collaborative mechanism design, detailed collaborative framework information is constructed. This includes clarifying the roles and responsibilities of each management level, collaborative workflow, information exchange standards, etc. This collaborative framework not only describes the logical relationship between each management level, but also provides guiding principles and methods for collaborative work.

[0032] After obtaining the collaborative framework information, the system terminal combs the connection process between each management layer based on the collaborative framework information. This includes determining the data flow direction, information transmission path and key nodes of collaborative work to ensure that each management layer can receive and send necessary information at the right time according to the process requirements and achieve seamless connection with other management layers. Subsequently, specific hierarchical connection rules are formulated according to the combed connection process. The hierarchical connection rules clarify the data exchange standards, information sharing methods, decision support processes, etc. between each management layer. After that, after the hierarchical connection rules are formulated, their feasibility is verified through simulation experiments. This helps to discover possible problems or deficiencies in the rules and make timely adjustments. The verification process can also help each management layer better understand and accept these rules, laying the foundation for subsequent collaborative work. Then, based on the verification results, the hierarchical connection rules are improved and optimized, including adjusting the data exchange format, optimizing the information sharing process or adding new collaborative mechanisms, etc., to ensure that the rules can adapt to the changing quality control needs and maintain their long-term effectiveness.

[0033] In summary, by extracting shared data sets, conducting cross-layer collaborative analysis, and formulating hierarchical connection rules, the system terminal successfully built a hierarchical connection rule that can integrate resources at each management level and optimize the quality management process. This hierarchical connection rule provides strong support for improving product quality and enhancing the collaborative capabilities between various management levels.

[0034] Distill and optimize the quality management teacher model through a knowledge distillation channel, wherein the knowledge distillation channel includes N distillation goodness branches; In one embodiment, knowledge distillation is a model optimization technique whose core idea is to transfer the knowledge of a large and complex teacher model to a small and simple student model, so that the student model can reduce the amount of computation and storage requirements while maintaining performance. In the process of distilling and optimizing the quality management teacher model, the system terminal uses a knowledge distillation channel containing N distillation excellence branches. Specifically, the system terminal determines the knowledge distillation target that each of the N distillation excellence branches is responsible for. For example, the surface roughness branch will focus on extracting knowledge and features related to surface roughness from the teacher model; the surface warpage branch focuses on the relevant information of warpage; the surface finish branch and the surface coating uniformity branch focus on the knowledge of finish and coating uniformity, respectively. Subsequently, the network structure of each branch is designed, and specific convolutional layers, pooling layers, fully connected layers, etc. are designed according to the docking management layer to extract and encode the quality management features of each concern. After that, the interface between the teacher model and each distillation excellence branch is defined to ensure that the output of the teacher model can be correctly input into each branch, and the data flow is designed so that the output of the teacher model can flow to the corresponding distillation excellence branches respectively. Through the above process, the system terminal successfully constructs N distillation excellence branches. Then, the system terminal optimizes each management layer in the quality management teacher model through these distillation goodness branches to obtain the corresponding optimization results. In this way, the subsequent student model can quickly absorb the essence of the teacher model in quality management, and show similar or even better performance than the teacher model in practical applications. At the same time, due to the small size of the student model, it is easier to deploy in an environment with limited resources, which provides greater convenience for actual production. In summary, the distillation optimization of the quality management teacher model through the knowledge distillation channel successfully achieved the lightweight and performance improvement of the model, injecting new vitality into the development of the quality management field.

[0035] Performing data distillation on the N quality management layers through the N distillation goodness branches to obtain N goodness distillation results; In one embodiment, the system terminal performs data distillation on N different quality management layers in the quality management teacher model by constructing N distillation excellence branches. Each branch targets a specific quality management layer and is responsible for extracting and encoding the key knowledge and features of that layer. Through distillation, the system terminal obtains N excellence distillation results, which represent the essence of knowledge and experience transferred from the teacher model to the student model. These results not only retain the performance advantages of the teacher model, but also optimize computing efficiency and storage space through the simplified structure of the student model. In summary, data distillation through N distillation excellence branches can effectively extract key knowledge of multiple quality management layers from the teacher model, and convert it into a form that can be understood by the student model, thereby improving the performance of the student model.

[0036] Further, the present application provides performing data distillation on the N quality management layers through the N distillation goodness branches to obtain N goodness distillation results, and the method further includes: The surface roughness management layer in the quality management teacher model is distilled and optimized through the surface roughness distillation goodness branch in the knowledge distillation channel to obtain a surface roughness goodness distillation result; The surface warpage management layer is distilled and optimized through a surface warpage distillation goodness branch to obtain a surface warpage goodness distillation result; Performing distillation optimization on the surface finish management layer through the surface finish distillation goodness branch to obtain a surface finish goodness distillation result; The surface coating uniformity management layer is distilled and optimized through a surface coating uniformity distillation goodness branch to obtain a surface coating uniformity goodness distillation result; Preferably, the system terminal uses four specific distillation goodness branches to perform distillation optimization on different management layers in the quality management teacher model through the constructed knowledge distillation channel. These branches are distillation goodness branches for surface roughness, surface warpage, surface finish and surface coating uniformity. First, the surface roughness distillation goodness branch extracts key knowledge and features from the surface roughness management layer of the teacher model, and uses the extracted data as soft labels, which contain the teacher model's rich knowledge and experience in surface roughness recognition. Subsequently, the weight distribution and connection mode of each neuron in the surface roughness management layer are analyzed. Determine which weights are critical to the recognition of surface roughness, and mark the key weights that need to be optimized. After that, the surface roughness distillation goodness branch adjusts the weights of relevant neurons in the surface roughness management layer according to the extracted key features. For example, neurons that respond strongly to key features are given higher weights to enhance the model's ability to recognize surface roughness, or these weights are fine-tuned using a gradient descent algorithm to improve the recognition performance of surface roughness. Then, the optimized surface roughness management layer is output as a surface roughness goodness distillation result. This result reflects the performance advantage of the teacher model in terms of surface roughness, which is successfully passed on to the student model. Secondly, the surface warpage distillation goodness branch performs a similar distillation optimization process on the surface warpage management layer, and obtains the goodness distillation result of surface warpage. This result also retains the excellent performance of the teacher model in this aspect. Similarly, the surface finish distillation goodness branch performs the same distillation optimization on the surface finish management layer, extracts the key knowledge and experience about surface finish, and forms the goodness distillation result of surface finish. Finally, the surface coating uniformity distillation goodness branch performs distillation optimization on the surface coating uniformity management layer, and obtains the goodness distillation result of surface coating uniformity. This result reflects the professional knowledge and experience of the teacher model in terms of coating uniformity. Through the collaborative work of these four distillation goodness branches, the key knowledge and experience about surface roughness, surface warpage, surface finish and surface coating uniformity are successfully distilled from the teacher model, and the corresponding goodness distillation results are formed. These results not only improve the performance of the student model at the corresponding quality management level, but also provide strong support for further model optimization and application.

[0037] The N goodness distillation results are determined based on the surface roughness goodness distillation result, the surface warpage goodness distillation result, the surface smoothness goodness distillation result, and the surface coating uniformity goodness distillation result.

[0038] Preferably, based on the obtained surface roughness goodness distillation results, surface warpage goodness distillation results, surface finish goodness distillation results, and surface coating uniformity goodness distillation results, the system terminal regards these goodness distillation results as a quantitative reflection of the optimization effect of the quality management teacher model in surface quality assessment. These goodness distillation results are obtained by optimizing the relevant management layers in the teacher model, and they respectively reflect the performance improvement of the model in identifying key quality features such as surface roughness, warpage, finish, and coating uniformity. Combining these N goodness distillation results, the overall optimization effect of the model can be comprehensively evaluated. These results not only show the degree of optimization of the model on a single quality feature, but also provide the performance of the model in comprehensive quality assessment. By comparing the results before and after optimization, the degree of improvement of the model in surface quality assessment can be quantitatively understood, so as to better guide subsequent quality management work. In summary, these N goodness distillation results are quantitative indicators of the optimization effect of the quality management teacher model in surface quality assessment, and they together constitute a comprehensive evaluation of the comprehensive performance of the model.

[0039] A quality management student model is constructed based on the N goodness distillation results, and the surface quality of the target stator core mold is optimized through the quality management student model.

[0040] In one embodiment, the system terminal uses a BP (back propagation) neural network based on the N goodness distillation results obtained to construct a quality management student model, and optimizes it through a preset distillation loss calculation formula. This quality management student model is inspired by the knowledge after the teacher model is optimized. It integrates the goodness distillation results of multiple key quality features such as surface roughness, warpage, smoothness, and coating uniformity. These goodness distillation results not only contain the experience of the quality management teacher model in identifying these quality features, but also reflect the learning results of the quality management teacher model in the optimization process. By using the quality management student model, the system terminal can optimize the surface quality of the target stator core mold. The quality management student model can use the key knowledge and features distilled from the quality management teacher model to accurately evaluate the surface quality of the stator core mold. Based on these evaluation results, possible problems on the mold surface can be found, and corresponding optimization measures can be taken, such as adjusting processing parameters, improving coating processes, etc., to improve the surface quality of the mold. In summary, by constructing a quality management student model and applying N goodness distillation results, the surface quality of the target stator core mold can be optimized and the product quality can be improved.

[0041] Furthermore, the present application provides that before the frosting area distribution and frosting thickness are input into the defrost control analysis module for analysis and processing, the method further includes: Based on the BP neural network, a quality management student model is constructed through the N goodness distillation results; Calculate and output the distillation loss data set based on the distillation loss calculation formula; Preferably, the system terminal uses N goodness distillation results based on BP neural network to construct a quality management student model, so that it can learn the key knowledge and features distilled from the teacher model. BP neural network is a powerful learning algorithm that adjusts the network weights by back-propagating errors to achieve classification or regression of input data. Specifically, the system terminal first determines the network structure of the quality management student model, including the number of nodes in the input layer, hidden layer and output layer. Then, according to the number of features of the N goodness distillation results, the number of nodes in the input layer is determined. Subsequently, according to the complexity of the problem and the characteristics of the data, the appropriate number of hidden layers and nodes are selected, and the number of nodes in the output layer corresponds to the number of quality management indicators. After that, the connection weights in the neural network are randomly assigned initial values. Then, the N goodness distillation results are used as input data, and the output of the network is calculated by the forward propagation algorithm. Then, the error between the network output and the output of the teacher model is calculated. Using the back-propagation algorithm, the error is back-propagated from the output layer to the input layer, and the weights between the nodes in each layer are adjusted according to the error size. Finally, the forward propagation and back-propagation processes are repeated, and the weights are iteratively updated until the preset maximum number of iterations is reached, and the constructed quality management student model is output.

[0042] After obtaining the quality management student model, the system terminal uses the pre-set distillation loss calculation formula to calculate the distillation loss. Distillation loss is a measure of the difference between the output of the student model and the teacher model. By comparing the outputs of the two, a loss value is calculated, which reflects the performance of the student model when imitating the teacher model. The distillation loss calculation formula is designed based on the difference between the outputs of the two to ensure that the student model can be as close to the performance of the teacher model as possible. Through calculation, the system terminal successfully obtains the distillation loss dataset. This distillation loss dataset will be used to further adjust and optimize the parameters of the student model. Through iterative training, the distillation loss can be continuously reduced, thereby improving the performance of the student model.

[0043] In summary, based on the BP neural network, the quality management student model can be constructed and the calculation results obtained according to the distillation loss calculation formula can be used to evaluate it, so that it can make full use of N optimality distillation results to improve the management effect of the target stator core mold surface quality.

[0044] Evaluating the effectiveness of the quality management student model based on the distillation loss data set to generate effective evaluation results; Optimizing the quality management student model according to the effective evaluation results.

[0045] Preferably, in the process of optimizing the quality management student model, the distillation loss data set is of great significance. This data set is used to measure the performance difference of the student model when imitating the teacher model, by storing the error between the outputs of the two, that is, the distillation loss. This loss value reflects the effect of the student model in learning and inheriting the knowledge of the teacher model. In order to evaluate the effectiveness of the quality management student model, the system terminal uses the distillation loss data set for in-depth analysis. By observing the size and change trend of the loss value, it is judged whether the performance of the student model meets expectations and whether there is room for further optimization. A smaller distillation loss value means that the student model can better approximate the performance of the teacher model, indicating that the model is more effective. Based on the effective evaluation results, the system terminal optimizes the quality management student model in a targeted manner. If the distillation loss is large, it means that the student model is still insufficient in some aspects and needs further optimization. At this time, the system terminal tries to adjust the network structure of the model, increase the number of hidden layer nodes, change the learning rate and other strategies to improve the performance of the model. By continuously optimizing the quality management student model, it can be better adapted to the actual quality management needs and improve the optimization management level of the surface quality of the stator core mold. In summary, using the distillation loss dataset to evaluate the effectiveness of the quality management student model and optimizing it based on the evaluation results are key steps to improve model performance and achieve high-quality management.

[0046] Further, the present application provides a formula for calculating distillation loss, and the method also includes: ; in, is the distillation loss, is the weight factor, is a smooth probability distribution, For any category, For the quality management teacher model, For the quality management student model, is the output probability distribution of the quality management teacher model, Output probability distribution of the quality management student model.

[0047] Optionally, the distillation loss calculation formula is as follows: ; in, Represents the distillation loss function, which measures the student model Imitating the teacher model The performance gap here is Teacher Model and student models Specifically, Is a weight factor used to adjust the influence of distillation loss in the overall loss function. It can be adjusted according to the needs of specific tasks to balance the importance of different loss terms. Represents a smoothed probability distribution, which is used to soften the output of the teacher model, making it easier for the student model to learn. Smoothing helps prevent the student model from overfitting to the hard-labeled output of the teacher model. Teacher Model For The system terminal first calculates the logarithm of the output probability distribution of the teacher model for each sample or category, then multiplies these logarithms by the weight factor and the square of the smoothed probability distribution, and finally sums the weighted logarithms of all samples to obtain the overall distillation loss. The purpose of this formula is to quantify the performance of the student model in the imitation process by comparing the output probability distributions of the teacher model and the student model, and to improve the accuracy and generalization ability of the student model by optimizing the distillation loss, so as to achieve more efficient model training and performance improvement.

[0048] In summary, the embodiments of the present application have at least the following technical effects: The embodiment of the present application obtains the quality deviation coefficient through surface quality detection, and performs multi-dimensional verification based on the coefficient to obtain N quality management layers. Subsequently, these quality management layers are connected according to the hierarchical association relationship to construct a quality management teacher model. After that, the teacher model is optimized through knowledge distillation, N goodness distillation results are generated, and a quality management student model is constructed. Then, the BP neural network is used to evaluate the effectiveness of the student model, and the model is optimized according to the evaluation results. This method improves the surface quality of the stator core mold through a multi-level and multi-dimensional management and optimization process, and realizes efficient quality management. These technical effects jointly solve the technical problems that traditional detection methods often rely on manual experience and naked eye observation, only focus on the roughness and smoothness of the mold surface, and it is difficult to accurately evaluate the actual quality of the mold surface, resulting in low accuracy and stability of quality management. Through the comprehensive measures of multiple links such as knowledge distillation channels and quality management teacher models, the technical effects of improving the surface quality of the mold and improving the accuracy and efficiency of quality management are achieved.

[0049] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. Other embodiments are within the scope of the attached claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0050] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0051] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A surface quality management method for a stator core mold, characterized in that: The method comprises: Conduct surface quality inspection on the target stator core mold to obtain the quality deviation coefficient; Perform multi-dimensional verification based on the quality deviation coefficient and output N quality management layers, where N is an integer greater than or equal to 1; Connect the N quality management layers according to the hierarchical association relationship to build a quality management teacher model; Distill and optimize the quality management teacher model through a knowledge distillation channel, wherein the knowledge distillation channel includes N distillation goodness branches; Performing data distillation on the N quality management layers through the N distillation goodness branches to obtain N goodness distillation results; A quality management student model is constructed based on the N goodness distillation results, and the surface quality of the target stator core mold is optimized through the quality management student model.

2. A surface quality management method for a stator core mold according to claim 1, characterized in that: Perform surface quality inspection on the target stator core mold to obtain the quality deviation coefficient. The method includes: Traversing the qualified mold parameters of the stator core of the big data, formulating the surface quality inspection standard of the target stator core mold, wherein the surface quality inspection standard includes a preset roughness standard, a preset warpage standard, a preset smoothness standard, and a preset coating uniformity standard; In the first detection phase, the target stator core mold is detected based on the preset roughness standard to generate a first detection result; In the second detection phase, the target stator core mold is detected based on the preset warpage standard to generate a second detection result; In the third detection stage, the target stator core mold is detected based on the preset smoothness standard to generate a third detection result; In the fourth detection stage, the target stator core mold is detected based on the preset coating uniformity standard to generate a fourth detection result; Integrate the first test result, the second test result, the third test result, and the fourth test result to perform deviation calculation to generate first mass deviation data, second mass deviation data, third mass deviation data, and fourth mass deviation data; The first mass deviation data, the second mass deviation data, the third mass deviation data, and the fourth mass deviation data are added to the mass deviation coefficient.

3. A surface quality management method for a stator core mold according to claim 2, characterized in that: Based on the quality deviation coefficient, multi-dimensional verification is performed to output N quality management layers, and the method includes: Respectively analyzing the change trends of the first mass deviation data, the second mass deviation data, the third mass deviation data, and the fourth mass deviation data to obtain a first mass change curve, a second mass change curve, a third mass change curve, and a fourth mass change curve; Determine a verification dimension based on the first quality change curve, the second quality change curve, the third quality change curve, and the fourth quality change curve; Testing and recording the quality deviation coefficient according to the verification dimension to generate a multi-dimensional verification result; Based on the multi-dimensional verification result, weight allocation is performed according to the first mass change curve, the second mass change curve, the third mass change curve, and the fourth mass change curve to generate a weight configuration result; The hierarchical division is determined according to the weight configuration result to generate the N quality management layers.

4. A surface quality management method for a stator core mold according to claim 1, characterized in that: The N quality management layers are connected according to the hierarchical association relationship to construct a quality management teacher model, and the method includes: Determining the hierarchical association relationship between the N quality management layers based on mapping the information flow between the N quality management layers; Constructing a hierarchical connection rule according to the hierarchical association relationship, and connecting the N quality management layers according to the hierarchical connection rule, wherein the N quality management layers include a surface roughness management layer, a surface warpage management layer, a surface finish management layer, and a surface coating uniformity management layer; The surface roughness management layer, the surface warpage management layer, the surface finish management layer, and the surface coating uniformity management layer are integrated to obtain the quality management teacher model.

5. A surface quality management method for a stator core mold according to claim 4, characterized in that: The hierarchical connection rules are constructed according to the hierarchical association relationship, and the method includes: Based on the hierarchical association relationship, the shared data sets of the surface roughness management layer, the surface warpage management layer, the surface finish management layer, and the surface coating uniformity management layer are sequentially extracted; Perform cross-layer collaborative analysis based on the shared data set to construct collaborative framework information; The connection process is sorted out according to the collaborative framework information, and the hierarchical connection rules are formulated.

6. A surface quality management method for a stator core mold according to claim 4, characterized in that: The N quality management layers are subjected to data distillation through the N distillation goodness branches to obtain N goodness distillation results, the method comprising: The surface roughness management layer in the quality management teacher model is distilled and optimized through the surface roughness distillation goodness branch in the knowledge distillation channel to obtain a surface roughness goodness distillation result; The surface warpage management layer is distilled and optimized through a surface warpage distillation goodness branch to obtain a surface warpage goodness distillation result; Performing distillation optimization on the surface finish management layer through the surface finish distillation goodness branch to obtain a surface finish goodness distillation result; The surface coating uniformity management layer is distilled and optimized through a surface coating uniformity distillation goodness branch to obtain a surface coating uniformity goodness distillation result; The N goodness distillation results are determined based on the surface roughness goodness distillation result, the surface warpage goodness distillation result, the surface smoothness goodness distillation result, and the surface coating uniformity goodness distillation result.

7. A surface quality management method for a stator core mold according to claim 1, characterized in that: Methods include: Based on the BP neural network, a quality management student model is constructed through the N goodness distillation results; Calculate and output the distillation loss data set based on the distillation loss calculation formula; Evaluating the effectiveness of the quality management student model based on the distillation loss data set to generate effective evaluation results; Optimizing the quality management student model according to the effective evaluation results.

8. A surface quality management method for a stator core mold according to claim 7, characterized in that: The distillation loss calculation formula is: ； in, is the distillation loss, is the weight factor, is a smooth probability distribution, For any category, For the quality management teacher model, For the quality management student model, is the output probability distribution of the quality management teacher model, Output probability distribution of the quality management student model.

Images