Intelligent maintenance and management platform for process data of intelligent manufacturing production line
Through the design of the intelligent manufacturing process data management platform, automated analysis and adjustment of process parameters, the process parameter selection errors caused by manual management are solved, and higher product quality and production efficiency are achieved.
Patent Information
- Application Number
- CN202510562950.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-30
- Publication Date
- 2025-06-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the prior art, the management of production process parameters relies on manual or semi-automatic systems, resulting in errors in parameter selection caused by batch differences in raw materials, which may cause production abnormalities, quality fluctuations and even safety accidents, affecting product consistency and production efficiency.
Design an intelligent maintenance and management platform for process data for intelligent manufacturing. Through the raw material attribute data processing module, process deviation analysis module, process parameter analysis module, joint analysis module and process parameter adjustment module, we automatically obtain and analyze process parameter differences, establish a scatter plot, calculate the stability and degree of impact of quality, and finally update the process parameter combination to improve processing quality.
Through automated analysis and adjustment of process parameters, reduce manual errors, improve the accuracy of process parameters selection, stabilize product quality and production efficiency, and reduce safety risks.
Smart Images

Figure CN120087847A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of process parameter adjustment, and particularly relates to an intelligent maintenance and management platform for process data of an intelligent manufacturing production line. Background Art
[0002] With the rapid development of intelligent manufacturing, manufacturing enterprises have put forward higher requirements for the digital and intelligent management of production data. As the key core data in the manufacturing process, production process data is directly related to product quality, production efficiency, etc. By maintaining and managing production process data, it plays a crucial role in the efficiency and quality during the product production process.
[0003] Currently, the management of process parameters during production mainly relies on manual maintenance or semi-automatic systems. This method usually requires workers or equipment to manually judge and select preset processing parameters according to the characteristics of raw materials (such as hardness, humidity, etc.). However, due to large differences in raw material batches, manual judgment is prone to errors, resulting in improper use of parameters, which in turn leads to production anomalies, quality fluctuations, and even safety accidents, seriously affecting product consistency and production efficiency. Summary of the Invention
[0004] In order to solve the above technical problems, the purpose of the present invention is to provide an intelligent maintenance and management platform for process data of an intelligent manufacturing production line, and the specific technical solutions adopted are as follows: An embodiment of the present invention provides an intelligent maintenance and management platform for process data of an intelligent manufacturing production line, and the platform includes: A raw material attribute data processing module, which is used to obtain the attribute feature vectors of each sample in history; calculate the differences between every two attribute feature vectors to obtain material attribute differences; A process deviation analysis module, which is used to obtain the process deviation values after processing of two samples according to the material attribute differences and quality differences after processing of the two samples; A process parameter analysis module, which is used to obtain the process parameter difference weights of a process parameter when two samples are processed; establish a scatter plot based on the process parameter difference weights of a process parameter of every two samples and the process deviation values after processing; A joint analysis module, which is used to calculate the quality influence stability of a process parameter according to the slope between every two adjacent data points in the scatter plot corresponding to a process parameter; obtain the influence degree of the process parameter based on the quality influence stability and the slope between every two adjacent data points; The process parameter adjustment module is used to normalize the influence degree of a process parameter to obtain an influence weight; obtain the reference values of each process parameter of the current sample; and update the reference values of each process parameter of the current sample according to the influence weight to obtain the optimal process parameter combination of the current sample.
[0005] Preferably, obtain the attribute feature vectors of each sample in history, including: Regard the raw material of a product in history as a sample, obtain the attribute data before processing of each sample in history, normalize the attribute data before processing of each sample, and respectively form the attribute feature vector corresponding to each sample.
[0006] Preferably, obtain the process deviation value after processing of two samples according to the material attribute difference and the quality difference after processing of the two samples, including: After a sample is processed into a product, respectively obtain the absolute value of the difference between each parameter of the processed product and the corresponding standard to obtain the error value corresponding to each parameter; add up the error values corresponding to each parameter and normalize them to obtain a comprehensive error, and use a preset value to perform a negative correlation mapping on the comprehensive error to obtain the quality index of the processed product; Perform a negative correlation mapping on the material attribute difference between the two samples using the exponential function with the natural constant as the base to obtain a mapping result; obtain the absolute value of the difference between the quality indexes of the products processed from the two samples and multiply it by the mapping result to obtain the process deviation value after processing of the two samples.
[0007] Preferably, obtain the process parameter difference weight of a process parameter of two samples according to the difference of a process parameter when the two samples are processed, including: Obtain the absolute value of the difference of a process parameter when two samples are processed; compare the absolute value of the difference of this process parameter with the sum of the absolute values of the differences of each process parameter when these two samples are processed to obtain the process parameter difference weight of this process parameter of these two samples.
[0008] Preferably, establish a scatter plot according to the process parameter difference weight of a process parameter of every two samples and the process deviation value after processing, including: Use the process parameter difference weight of a process parameter of every two samples as the abscissa and the process deviation value after processing of every two samples as the ordinate to obtain a scatter plot.
[0009] Preferably, calculate the quality influence stability of a process parameter according to the slope between every two adjacent data points in the scatter plot corresponding to a process parameter, including: Obtain the slopes between every two adjacent data points in the scatter plot corresponding to a process parameter to form a slope sequence; Denote the slopes adjacent to a slope in the slope sequence as the left slope and the right slope respectively; Obtain the difference between the left slope and this slope, denoted as the first difference, and obtain the difference between this slope and the right slope, denoted as the second difference; Denote the absolute values of the first difference and the second difference as the local change amount; Calculate the average value of all the local change amounts corresponding to the slope sequence, denoted as the average local change amount; The reciprocal of the sum of the preset value and the average local change amount is the quality influence stability of this process parameter.
[0010] Preferably, obtaining the influence degree of this process parameter based on the quality influence stability and the slope between every two adjacent data points includes: Obtain the mean value of the absolute values of each slope in the slope sequence corresponding to the scatter plot of a process parameter and multiply it by the quality influence stability of this process parameter to obtain the influence degree of this process parameter.
[0011] Preferably, obtaining the reference value of each process parameter of the current sample includes: Cluster all historical samples using material property differences to obtain different clusters; For the current sample to be processed, obtain the material property differences between the current sample and the cluster centers of each cluster. The cluster corresponding to the cluster center with the smallest material property difference is the cluster to which the current sample belongs, denoted as the first cluster; Respectively obtain the material property differences between each sample in the first cluster and the current sample, and compare them with the hyperparameter to obtain the first ratio; Use the exponential function with the natural constant as the base to perform negative correlation mapping on the first ratios corresponding to each sample in the first cluster to obtain the weights corresponding to each sample in the first cluster; Use the first weights corresponding to each sample in the first cluster to calculate the weighted average value of a process parameter when each sample is processed to obtain the reference value of this process parameter of the current sample.
[0012] Preferably, updating the reference values of each process parameter of the current sample according to the influence weights to obtain the optimal process parameter combination of the current sample includes: Take the SVR support vector regression model as the quality prediction model; Use the process parameters of a sample during processing to form the process parameter vector of this sample, multiply each process parameter in the process parameter vector by the corresponding influence weight to obtain the weighted process parameter vector; Concatenate the attribute feature vector and the weighted process parameter vector of a sample and normalize them to form the input vector of this sample; Use the input vectors and corresponding labels of each sample to train the quality prediction model to obtain the trained quality prediction model. The label of each sample is the quality index of the product it is processed into; Multiply the base step size of a process parameter by the difference between the preset value and the process parameter to obtain the update step size of the reference value of the process parameter; use the update step size of the reference value of a process parameter to increase or decrease the reference value of the process parameter within the parameter range of the process parameter to obtain each updated process parameter of the process parameter, and each updated process parameter and the reference value of the process parameter form an updated process parameter set; Randomly select an updated process parameter from each updated process parameter set of the process parameters of the current sample to form an updated process parameter vector; obtain an input vector according to the attribute feature vector of the current sample, an updated process parameter vector, and the influence weights of the process parameters; input the input vector into the trained prediction model to obtain the quality index corresponding to the updated process parameter vector; select the updated process parameters in the updated process parameter vector with the largest quality index to form an optimal process parameter combination.
[0013] The embodiments of the present invention have at least the following beneficial effects: The present application obtains the attribute feature vector composed of the attribute data of the samples before processing in history, and then obtains the material attribute difference according to the difference between every two attribute feature vectors. Then, the process deviation value is obtained by correcting the quality difference after processing of the two samples using the material attribute difference between the two samples, excluding the influence of the material attribute difference on the quality difference, and can better illustrate the influence of the process parameters on the quality difference; further, obtain the process parameter difference weight of a process parameter of two samples, combine the process deviation values after processing of the two samples to establish a scatter plot, and then analyze the slope between every two adjacent data points in the scatter plot, and analyze the process parameter difference weight and the process deviation value jointly to obtain the stability of the influence of each process parameter on the quality, that is, the quality influence stability; then analyze based on the quality influence stability to obtain the influence degree of each process parameter on the processing quality of the sample, and can determine the importance of each process parameter for improving the quality during sample processing; then obtain the influence weights of the process parameters and the reference values of the process parameters of the current sample, and update the reference values of the process parameters of the current sample according to the influence weights to obtain the optimal process parameter combination of the current sample, improving the stability and quality of the processing quality of the current sample. Description of the Drawings
[0014] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.
[0015] Figure 1It is a block diagram of a module of an intelligent maintenance management platform for process data of an intelligent manufacturing production line provided by an embodiment of the present invention. Detailed implementation manners
[0016] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following, in combination with the accompanying drawings and preferred embodiments, details the specific implementation manners, structures, features, and effects of an intelligent maintenance management platform for process data of an intelligent manufacturing production line proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.
[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.
[0018] The following specifically describes the specific solution of an intelligent maintenance management platform for process data of an intelligent manufacturing production line provided by the present invention in conjunction with the accompanying drawings.
[0019] Embodiment: The main application scenario of the present invention is: During the product production process, the selection of process parameters directly affects the production efficiency and quality of the product. Most of the existing process data maintenance methods rely on manual maintenance or semi - automated system management, resulting in problems such as low data maintenance efficiency and untimely changes, which are extremely likely to cause incorrect use of process parameters, production anomalies, and even quality accidents, thus affecting the production efficiency and quality of the product.
[0020] Therefore, there is an urgent need to build a system with the capabilities of intelligent analysis, automatic adjustment, and continuous optimization of process parameters to meet the requirements of high - quality, high - efficiency, and flexible production in the context of intelligent manufacturing and achieve precise processing under different raw material conditions.
[0021] Please refer to Figure 1 , which shows a block diagram of a module of an intelligent maintenance management platform for process data of an intelligent manufacturing production line provided by an embodiment of the present invention. The platform includes the following modules: A raw material attribute data processing module, which is used to obtain the attribute feature vectors of each sample in history; calculate the differences between every two attribute feature vectors to obtain the material attribute differences.
[0022] During the manufacturing process, due to different property parameters of raw materials, such as the hardness of the material and the change of dimensions, it often leads to unstable product quality when different process parameters are selected for processing and producing products. Therefore, it is necessary to analyze the relevant attribute data of the raw materials to be processed and the process parameter data, etc., and realize the intelligent recommendation of the optimal parameter combination, so as to achieve the purpose of "tailoring the art according to the material", in order to avoid the improper selection due to the fluctuation of raw material data when traditional production methods process different raw materials based on the production process parameters preset by manual or equipment, resulting in unstable product quality.
[0023] First, the raw materials required for a product are called a sample. Furthermore, the relevant attribute data before processing the sample (raw materials) are obtained, such as data on hardness, geometric parameters (length, width, and height), etc. Here, the raw materials can be regarded as a sample to be processed. The attribute data of the raw materials and the process parameters of processing can be automatically obtained by integrating a variety of intelligent sensing and data acquisition devices. For example, the attribute data such as the size, surface thickness, and hardness of the raw materials can be collected in real time by non-contact sensors such as laser rangefinders, three-dimensional profilers, and ultrasonic thickness gauges or manually measured and input; the process parameters during processing, such as rotational speed, pressure, and temperature, are automatically recorded and uploaded by the process sensing modules (such as encoders, force sensors, and temperature sensors) inside the numerical control equipment.
[0024] Thus, the attribute data before processing each sample in history and various process parameters during processing can be obtained. The attribute data before processing each sample are normalized and respectively composed into attribute feature vectors, expressed as: , where represents the attribute feature vector of the a-th sample (raw material) among the same type of products produced, respectively represent the attribute values of the i-th dimension in the a-th sample, that is to say, the values after normalization of the i-th type of attribute data. For example, if the i-th dimension represents the hardness of the sample, then the attribute value of the i-th dimension represents the value after normalization of the hardness value.
[0025] For any two samples, the difference in their attribute data can be analyzed and calculated by using the cosine similarity method for the attribute feature vectors constructed by the corresponding attribute data of these two samples, so as to obtain the difference in the corresponding attributes of these two samples, that is, the difference before processing. Specifically, calculate the cosine similarity between the attribute feature vectors of the two samples, and use a preset value to perform a negative correlation mapping on the cosine similarity to obtain the material attribute difference between the two samples. The specific calculation model is: , where, represents the material attribute difference before processing between the a-th sample and the b-th sample; , respectively represent the attribute value of the i-th dimension of the attribute feature vectors corresponding to the a-th and b-th samples, m represents the number of attribute dimensions corresponding to the current sample, represents the cosine similarity value between the attribute feature vectors of the a-th sample and the b-th sample. The smaller this value is, the less similar these two samples are, that is, the greater the attribute difference between them. 1 represents a preset value used for negative correlation mapping of the cosine similarity. Thus, the material attribute differences between every two samples can be obtained.
[0026] The process deviation analysis module is used to obtain the process deviation value after processing of two samples based on the material attribute differences and the quality differences after processing of the two samples.
[0027] Through the above, the material attribute differences before processing of any two samples can be obtained. That is, when the difference in attribute data between these two samples is smaller, it indicates that the similarity of the attribute features before processing between these two samples is higher. Therefore, if the difference in attribute data before processing between these two samples is smaller, and there is a difference in their quality after processing, then it may indicate that it is caused by improper selection of process parameters during processing.
[0028] First, obtain the quality index of each sample after processing. Specifically, after a sample is processed into a product, respectively obtain the absolute value of the difference between each parameter of the processed product and the corresponding standard to obtain the error value corresponding to each parameter; add up the error values corresponding to each parameter and normalize them to obtain a comprehensive error, and use a preset value to perform negative correlation mapping on the comprehensive error to obtain the quality index of the processed product. The preset value is 1.
[0029] Furthermore, obtain the process deviation value after processing of two samples based on the material attribute differences and the quality differences after processing of the two samples. Specifically, use the exponential function with the natural constant as the base to perform negative correlation mapping on the material attribute differences between the two samples to obtain a mapping result; obtain the absolute value of the difference between the quality indexes of the products processed from the two samples and multiply it by the mapping result to obtain the process deviation value after processing of the two samples.
[0030] The specific calculation model of its process deviation value is as follows: , where, respectively represent the process deviation values after processing of the a-th sample and the b-th sample, and also represent the difference value of their processing quality caused by improper selection of process parameters, and respectively represent the quality indexes of the a-th sample and the b-th sample after processing, represents the absolute value of the difference between the quality indexes of the a-th and b-th samples after processing. If this value is larger, and when the material attribute differences between these two samples are smaller, that is When the value is larger, the corresponding difference in processing quality between the two samples caused by different process parameters may be greater. Therefore, The larger the value, the greater the possible quality difference in processing due to process parameters.
[0031] A process parameter analysis module, which is used to obtain the process parameter difference weight of a certain process parameter of two samples according to the difference in a certain process parameter during the processing of the two samples; and establish a scatter plot based on the process parameter difference weight of a certain process parameter of each two samples and the process deviation value after processing.
[0032] Through the above, the difference value of the possible quality difference caused by processing parameters for any two samples to be processed (raw materials) can be obtained. However, during the processing of samples, the influence degree of the selection and change of processing process parameters on processing quality may be different. For example, the influence of processing temperature and processing speed data on the processing quality of samples may be different. Therefore, in order to further obtain the influence weight of process parameters on processing quality, the process parameter difference weight of a certain process parameter of two samples is obtained according to the difference in a certain process parameter during the processing of the two samples.
[0033] Specifically, obtain the absolute value of the difference in a certain process parameter when processing two samples; compare the absolute value of the difference in this process parameter with the sum of the absolute values of the differences in each process parameter when processing these two samples to obtain the process parameter difference weight of this process parameter of these two samples.
[0034] The specific calculation model is: , Among them, represents the process parameter difference weight of the s-th process parameter when processing the a-th sample and the b-th sample, and respectively represent the values of the s-th process parameter when processing the a-th sample and the b-th sample; n represents the number of types of process parameters, represents the ratio of the absolute value of the difference between the values of the s-th process parameter when processing the a-th and b-th samples to the sum of the absolute values of the differences between the values of each process parameter when processing the a-th and b-th samples. The larger this value is, the higher the relative difference proportion between the values of the s-th process parameter when processing the a-th and b-th samples during processing.
[0035] Through the above, the weight coefficient of the difference in any process parameter when processing any two samples in the overall difference of the process parameters of these two samples is obtained, that is, the process parameter difference weight of a certain process parameter of two samples.
[0036] Further, a scatter plot is established based on the process parameter difference weight of a process parameter for every two samples and the process deviation value after processing. Specifically, taking the process parameter difference weight of a process parameter for every two samples as the abscissa and the process deviation value after processing of every two samples as the ordinate, a scatter plot is obtained.
[0037] Plot the process parameter difference weight corresponding to a process parameter and the process deviation value after processing of every two samples in the constructed two-dimensional coordinate system to obtain a scatter plot. If the data points in the figure change more smoothly and have a consistent regularity, it indicates that the above-mentioned process parameter has a more stable influence on the process deviation during sample processing. Then, the weight for adjusting this process parameter in the future may be greater to ensure the processing quality of the sample and also avoid the instability and unreliability of the overall analysis due to individual data when analyzing the influence degree of this process parameter on the processing quality later. This scatter plot reflects the relationship between the change of this process parameter and the change of the process deviation during sample processing. By analyzing this scatter plot, the influence of the adjustment of this process parameter on the quality during sample processing can be obtained more intuitively and accurately.
[0038] The joint analysis module is used to calculate the quality influence stability of a process parameter according to the slope between every two adjacent data points in the scatter plot corresponding to a process parameter; and obtain the influence degree of this process parameter based on the quality influence stability and the slope between every two adjacent data points.
[0039] In the above, obtain the scatter plot corresponding to each process parameter, and conduct a joint analysis of the quality difference caused by the process parameter between a process parameter and the sample after processing in the scatter plot.
[0040] Calculate the quality influence stability of a process parameter according to the slope between every two adjacent data points in the scatter plot corresponding to a process parameter. Specifically, obtain the slope between every two adjacent data points in the scatter plot corresponding to a process parameter to form a slope sequence; denote the slope adjacent to a slope on the left and right in the slope sequence as the left slope and the right slope respectively; obtain the difference between the left slope and this slope, denoted as the first difference, and obtain the difference between this slope and the right slope, denoted as the second difference; denote the absolute value of the first difference and the second difference as the local change amount; calculate the average value of all local change amounts corresponding to the slope sequence, denoted as the average local change amount; the reciprocal of the sum of the preset value and the average local change amount is the quality influence stability of this process parameter.
[0041] The specific calculation model is: , Among them, It represents the stability of the influence of the s-th process parameter on the processing quality, that is, the quality influence stability of the s-th process parameter. and and represent the (j - 1)-th, j-th, and (j + 1)-th slope values in the slope sequence formed by the slopes between adjacent data points in the scatter plot corresponding to the s-th process parameter. and represent the slope on the left and the slope on the right of the j-th slope. z represents the number of slopes in the slope sequence, and 1 represents a preset value. represents the first difference. represents the second difference. The absolute value of the difference between the two is the local change amount. represents the average value of all local change amounts corresponding to the slope sequence. The smaller this value is, that is when the value is larger, it indicates that the influence of this process parameter on the processing quality may be more stable.
[0042] Thus, the quality influence stability corresponding to each process parameter can be obtained. Further, based on the quality influence stability and the slope between every two adjacent data points, the influence degree of this process parameter is obtained. Specifically, the mean value of the absolute values of each slope in the slope sequence corresponding to the scatter plot of a process parameter is multiplied by the quality influence stability of this process parameter to obtain the influence degree of this process parameter.
[0043] The specific calculation model is: , where represents the influence degree of the s-th process parameter. represents the quality influence stability of the s-th process parameter. represents the average value of the absolute values of the slopes in the slope sequence corresponding to the scatter plot of the s-th process parameter. The larger this value is, it indicates that the influence of this process parameter on the sample processing quality may be greater. At the same time, when the influence of this process parameter is more stable, that is, it is more credible, and the influence degree is greater, it indicates that the final influence of this process parameter on the sample processing quality is greater.
[0044] Thus, the influence degree of each process parameter on the quality of the product obtained after processing the sample (raw material) to be processed can be obtained.
[0045] The process parameter adjustment module is used to normalize the influence degree of a process parameter to obtain an influence weight; obtain the reference values of the process parameters of the current sample; and update the reference values of the process parameters of the current sample according to the influence weight to obtain the optimal process parameter combination of the current sample.
[0046] After obtaining the influence degree of each process parameter, normalize the influence degree of each process parameter. Specifically, divide the influence degree of a process parameter by the sum of the influence degrees of all process parameters to obtain the influence weight of this process parameter.
[0047] The calculation model of the influence weight of process parameters is: , Where, represents the influence weight of the s-th process parameter, represents the influence degree of the s-th process parameter, and n represents the number of types of process parameters.
[0048] Furthermore, based on the known material property differences between every two samples above, cluster all historical samples using the material property differences to obtain different clusters; for the current sample to be processed, obtain the material property differences between the current sample and the cluster centers of each cluster. The cluster corresponding to the cluster center with the smallest material property difference is the cluster to which the current sample belongs, denoted as the first cluster; respectively obtain the material property differences between each sample in the first cluster and the current sample, and compare them with the hyperparameter to obtain the first ratio; use the exponential function with the natural constant as the base to perform a negative correlation mapping on the first ratio corresponding to each sample in the first cluster to obtain the weight corresponding to each sample in the first cluster; use the first weight corresponding to each sample in the first cluster to calculate the weighted average value of a process parameter when each sample is processed to obtain the reference value of this process parameter for the current sample.
[0049] Thus, obtain the reference values of each process parameter when the current sample is processed.
[0050] Furthermore, construct a quality prediction model and obtain the quality indicators of each historical sample. Preferably, the quality prediction model in the embodiments of the present invention is an SVR support vector regression model, and the loss function of the quality prediction model is the ε-Insensitive loss function.
[0051] Furthermore, use the process parameters of a sample when it is processed to form the process parameter vector of this sample. For example, , represents the process parameter vector corresponding to the a-th sample. Multiply each process parameter in the process parameter vector by the corresponding influence weight to obtain the weighted process parameter vector. For example, , D represents the weighted process parameter vector, represents the influence weight of the n-th process parameter of the a-th sample, represents the n-th process parameter of the a-th sample. Concatenate the attribute feature vector and the weighted process parameter vector of a sample end to end and normalize them to form the input vector of this sample.
[0052] For example, the input vector of the a-th sample is as follows: Multiplying the process parameters by their corresponding influence weights can reflect the different influences of different process parameters on quality, thereby improving the accuracy of quality prediction.
[0053] Furthermore, the label of each sample is the quality index obtained above. Using the input vectors and corresponding labels of each sample to train the quality prediction model, a trained quality prediction model is obtained.
[0054] Furthermore, it is necessary to update the reference values of the process parameters of the current sample according to the influence weights. Specifically, multiplying the basic step of a process parameter by the difference between the preset value and this process parameter is used as the update step of the reference value of this process parameter; using the update step of the reference value of a process parameter to increase or decrease the reference value of this process parameter within the parameter range of this process parameter, and obtaining each updated process parameter of this process parameter. Each updated process parameter and the reference value of this process parameter form an updated process parameter set.
[0055] Among them, the specific calculation model of the update step is: , represents the update step of the reference value of the s-th process parameter, represents the influence weight of the s-th process parameter, u represents the basic step of the process parameter, and the implementer needs to set it according to the actual situation. For process parameters with large influence weights, smaller update steps are required to improve the accuracy of determining the process parameters. For process parameters with small influence weights, larger update steps are required to accelerate the traversal speed of optimizing this process parameter.
[0056] In the actual production process, each process parameter has a parameter range, which is determined by the implementer according to the actual processing situation. Increasing or decreasing the reference value of each process parameter by using the update step, and then the obtained value and the reference value form a set, which is the updated process parameter set of this process parameter. For example, the reference value of the process parameter is 5, the parameter range is from 0 to 7, and the update step is 2, then the updated process parameter set is (3, 1, 5, 7).
[0057] Arbitrarily select an updated process parameter from each updated process parameter set of the current sample's process parameters to form an updated process parameter vector; where the number of obtained updated parameter vectors is the product of the number of elements in the updated process parameter sets of each process parameter, that is, permutation and combination is performed. For example, if the updated process parameter set of one process parameter is (a, b), and the updated process parameter set of another process parameter is (c, d, e), then the number of updated process parameter vectors is 6, namely ac, ad, ae, bc, bd, and be.
[0058] Obtain an input vector based on the attribute feature vector of the current sample, an updated process parameter vector, and the influence weights of each process parameter; input this input vector into the trained prediction model to obtain the quality index corresponding to the updated process parameter vector; select each updated process parameter in the updated process parameter vector with the largest quality index to form an optimal process parameter combination.
[0059] Furthermore, use each process parameter in the optimal process parameter combination to process the current sample to improve the processing quality of the current sample.
[0060] It should be noted that: the above sequence of embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. And the above has described specific embodiments of this specification. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0061] Each embodiment in this specification is described in a progressive manner. For the same or similar parts between each embodiment, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.
[0062] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.
Claims
1. An intelligent maintenance and management platform for process data of intelligent manufacturing production lines, characterized in that: The platform includes: The raw material attribute data processing module is used to obtain the attribute characteristic vector of each sample in history; the difference between every two attribute characteristic vectors is calculated to obtain the material attribute difference; A process deviation analysis module is used to obtain the process deviation values of the two samples after processing according to the material property difference and the quality difference after processing of the two samples; A process parameter analysis module is used to obtain a process parameter difference weight of a process parameter of two samples according to the difference of the process parameter of the two samples during processing; a scatter plot is established based on the process parameter difference weight of a process parameter of each two samples and the process deviation value after processing; A joint analysis module is used to calculate the quality impact stability of a process parameter according to the slope between every two adjacent data points in a scatter plot corresponding to the process parameter; and obtain the impact degree of the process parameter based on the quality impact stability and the slope between every two adjacent data points; The process parameter adjustment module is used to normalize the influence of a process parameter to obtain the influence weight; obtain the baseline value of each process parameter of the current sample; and update the baseline value of each process parameter of the current sample according to the influence weight to obtain the optimal process parameter combination of the current sample.
2. According to claim 1, the intelligent maintenance management platform for process data of intelligent manufacturing production lines is characterized in that: The step of obtaining the attribute feature vector of each sample in history includes: The raw material of a product in history is called a sample, and the attribute data of each sample before processing in history is obtained. The attribute data of each sample before processing is normalized to form the attribute feature vector corresponding to each sample.
3. According to claim 1, the intelligent maintenance management platform for process data of intelligent manufacturing production lines is characterized in that: The method of obtaining the process deviation values of the two samples after processing according to the material property difference of the two samples and the quality difference after processing includes: After a sample is processed into a product, the absolute values of the differences between the parameters of the processed product and the corresponding standards are calculated to obtain the error value corresponding to each parameter; the error values corresponding to each parameter are added and normalized to obtain the comprehensive error, and the preset value is used to negatively correlate the comprehensive error to obtain the quality index of the processed product; The material property difference of the two samples is negatively correlated and mapped using an exponential function with a natural constant as the base to obtain a mapping result; the absolute value of the difference in quality indicators of the products processed from the two samples is calculated and multiplied by the mapping result to obtain the process deviation value of the two samples after processing.
4. According to claim 1, the intelligent maintenance management platform for process data of intelligent manufacturing production lines is characterized in that: The step of obtaining a process parameter difference weight of a process parameter of two samples according to a difference in a process parameter when the two samples are processed includes: Obtain the absolute value of the difference in a process parameter when two samples are processed; compare the absolute value of the difference in the process parameter with the sum of the absolute values of the differences in various process parameters when the two samples are processed to obtain the process parameter difference weight of the process parameter of the two samples.
5. According to claim 1, the intelligent maintenance management platform for process data of intelligent manufacturing production lines is characterized in that: The step of establishing a scatter plot based on the process parameter difference weight of a process parameter of every two samples and the process deviation value after processing includes: The process parameter difference weight of a process parameter between every two samples is used as the horizontal axis, and the process deviation value after processing every two samples is used as the vertical axis to obtain a scatter plot.
6. According to claim 1, the intelligent maintenance management platform for process data of intelligent manufacturing production lines is characterized in that: The step of calculating the quality impact stability of a process parameter according to the slope between every two adjacent data points in a scatter plot corresponding to the process parameter comprises: Obtain the slope between every two adjacent data points in the scatter plot corresponding to a process parameter to form a slope sequence; record the slopes adjacent to a slope in the slope sequence on the left and right as the left slope and the right slope respectively; obtain the difference between the left slope and the slope, record it as the first difference, obtain the difference between the slope and the right slope, record it as the second difference; record the absolute values of the first difference and the second difference as the local variation; calculate the average value of all local variations corresponding to the slope sequence, record it as the average local variation; the reciprocal of the sum of the preset value and the average local variation is the quality influence stability of the process parameter.
7. The intelligent maintenance management platform for process data of intelligent manufacturing production lines according to claim 1 is characterized in that: The method of obtaining the influence of the process parameter based on the quality influence stability and the slope between every two adjacent data points includes: The influence degree of a process parameter is obtained by obtaining the mean of the absolute value of each slope in the slope sequence corresponding to the scatter plot of a process parameter and multiplying it by the quality influence stability of the process parameter.
8. The intelligent maintenance management platform for process data of intelligent manufacturing production lines according to claim 1 is characterized in that: The step of obtaining the reference values of various process parameters of the current sample includes: All historical samples are clustered using material property differences to obtain different clusters; for the current sample to be processed, the material property difference between the current sample and the cluster centers of each cluster is obtained, and the cluster center corresponding to the smallest material property difference is the cluster to which the current sample belongs, recorded as the first cluster; the material property difference between each sample in the first cluster and the current sample is obtained, and compared with the hyperparameter to obtain a first ratio; an exponential function with a natural constant as the base is used to negatively correlate the first ratio corresponding to each sample in the first cluster to obtain the weight corresponding to each sample in the first cluster; the first weight corresponding to each sample in the first cluster is used to calculate the weighted average of a process parameter when each sample is processed to obtain the baseline value of the process parameter of the current sample.
9. The intelligent maintenance management platform for process data of intelligent manufacturing production lines according to claim 1 is characterized in that: The updating of the reference values of various process parameters of the current sample according to the influence weights to obtain the optimal process parameter combination of the current sample includes: The SVR support vector regression model is used as a quality prediction model; various process parameters of a sample during processing are used to form a process parameter vector of the sample, and each process parameter in the process parameter vector is multiplied by the corresponding influence weight to obtain a weighted process parameter vector; the attribute feature vector of a sample and the weighted process parameter vector are connected end to end and normalized to form an input vector of the sample; The quality prediction model is trained using the input vector and the corresponding label of each sample to obtain a trained quality prediction model, wherein the label of each sample is a quality indicator of the product processed therefrom; The basic step length of a process parameter is multiplied by the difference between a preset value and the process parameter as the update step length of the reference value of the process parameter; the reference value of the process parameter is increased or decreased within the parameter range of the process parameter by using the update step length of the reference value of the process parameter to obtain each updated process parameter of the process parameter, and each updated process parameter and the reference value of the process parameter constitute an updated process parameter set; An updated process parameter is arbitrarily selected from the updated process parameter set of each process parameter of the current sample to form an updated process parameter vector; an input vector is obtained according to the attribute feature vector of the current sample, an updated process parameter vector and the influence weights of each process parameter; the input vector is input into the trained prediction model to obtain the quality index corresponding to the updated process parameter vector; and each updated process parameter in the updated process parameter vector with the largest quality index is selected to form an optimal process parameter combination.
Citation Information
Cited By
Industrial production line fault self-diagnosis system based on digital twinning
CN120722874A
Fault self-diagnosis system for industrial production line based on digital twinning
CN120722874B
Feed production and processing detection method based on big data
CN121544134A
