Product ingredient data monitoring method, device, electronic device and storage medium

Abstract

The present invention provides a product component data monitoring method, device, electronic device, and storage medium. The method comprises: first, based on preset conditions, screening out-of-control component data sample sets from a control chart sample set; then, training an initial adaptive enhancement-based vector machine model based on the out-of-control component data sample set to obtain corresponding abnormal deviation patterns; and finally, determining the abnormal deviation patterns of the product to be tested based on the fully trained adaptive enhancement-based vector machine model. By converting the product component data into a control chart sample set for judgment, the results are concise and clear, and traceability is facilitated. After obtaining the out-of-control component statistics, the adaptive enhancement-based vector machine model can determine the abnormal deviation patterns, thereby automatically processing abnormal data and determining abnormal product patterns.

Description

Technical Field

[0001] The present invention relates to the field of data monitoring, and in particular to a product ingredient data monitoring method, device, electronic equipment and storage medium. Background Art

[0002] In the quality management of the manufacturing production process, control charts, as a tool of statistical process control (SPC), are often used to monitor abnormal fluctuations in the process. They mainly use mathematical statistical analysis to conduct online real-time monitoring of the product production process. By monitoring the changes in the quality characteristics of the product, it is determined whether the production process is in a controlled or out-of-control state. Once it is out of control, an alarm will be issued, so that production personnel can detect the abnormality in time and make adjustments to achieve the purpose of production quality control.

[0003] In real-world manufacturing scenarios, such as in the chemical and food industries, it's often necessary to control the ratios or concentrations of various product quality characteristics. These quality characteristics are known as component data. The quality of finished products is significantly influenced by the distribution of component data, so multivariate control charts are often used to monitor the distribution of component data and thereby control production quality.

[0004] However, existing statistical process control can only monitor the status of the production process and cannot further identify abnormal patterns based on the monitoring data. This requires manual inspection of the source of abnormalities. Therefore, when monitoring product ingredients, existing technologies are unable to automatically process abnormal data and thus cannot identify abnormal patterns. Summary of the Invention

[0005] In view of this, it is necessary to provide a product ingredient data monitoring method, device, electronic device and storage medium that can effectively monitor the specific abnormal pattern type corresponding to the out-of-control data in the product ingredient data.

[0006] In order to solve the above problems, the present invention provides a product ingredient data monitoring method, comprising:

[0007] Obtain the control chart sample set corresponding to the product component data, and filter out the out-of-control component data sample set based on preset conditions;

[0008] Input the out-of-control component data sample set into the initial vector machine model based on adaptive enhancement, take the abnormal offset pattern corresponding to the out-of-control component data sample set as the output, iteratively train, and determine the fully trained vector machine model based on adaptive enhancement;

[0009] Acquire real-time out-of-control component data and input the real-time out-of-control component data into a well-trained vector machine model based on adaptive enhancement to determine the abnormal deviation pattern of the product under test.

[0010] Furthermore, the initial vector machine model based on adaptive enhancement includes multiple weak classifiers, and the multiple weak classifiers are weighted by the adaptive enhancement algorithm to form a strong classifier; the data of the strong classifier is mapped to a high-dimensional feature space for classification by the support vector machine algorithm.

[0011] Furthermore, the support vector machine algorithm includes kernel parameters and penalty factors, and the kernel parameters and penalty factors are optimized by a differential evolution algorithm.

[0012] Furthermore, the performance indicators of the control chart sample set include an average running length in a controlled state and an average running length in an out-of-control state.

[0013] Furthermore, the average run length in the controlled state is solved by the Markov chain algorithm; the average run length in the out-of-control state is solved by the Monte Carlo simulation method.

[0014] Furthermore, the step of obtaining a control chart sample set includes:

[0015] Obtain product ingredient data;

[0016] Perform isometric logarithmic ratio transformation on product composition data samples to determine mean coordinate vector samples;

[0017] Determine the statistic sample based on the mean coordinate vector sample and construct the control chart sample set.

[0018] Furthermore, product ingredient data is obtained, including:

[0019] The collected product production data is subjected to wavelet transform noise reduction processing to obtain product composition data.

[0020] In order to solve the above problems, the present invention further provides a product ingredient data monitoring device, comprising:

[0021] The out-of-control component data sample set acquisition module is used to obtain the control chart sample set corresponding to the product component data and filter out the out-of-control component data sample set based on preset conditions;

[0022] A model training module is used to input the out-of-control component data sample set into the initial adaptive enhancement-based vector machine model, and iteratively train the abnormal offset pattern corresponding to the out-of-control component data sample set to determine a fully trained adaptive enhancement-based vector machine model;

[0023] The abnormal deviation pattern determination module is used to obtain real-time out-of-control component data, input the real-time out-of-control component data into a well-trained vector machine model based on adaptive enhancement, and determine the abnormal deviation pattern of the product to be tested.

[0024] In order to solve the above problems, the present invention also provides an electronic device, including a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the product ingredient data monitoring method described above is implemented.

[0025] In order to solve the above problems, the present invention also provides a computer-readable storage medium, which stores computer program instructions. When the computer program instructions are executed by a computer, the computer executes the product ingredient data monitoring method as described above.

[0026] The beneficial effects of adopting the above technical solution are as follows: The present invention provides a product component data monitoring method, device, electronic device, and storage medium. The method includes: first, based on preset conditions, screening out-of-control component data sample sets from a control chart sample set; then, training an initial adaptive enhancement-based vector machine model based on the out-of-control component data sample set to obtain corresponding abnormal offset patterns; finally, determining the abnormal offset pattern of the product to be tested based on the fully trained adaptive enhancement-based vector machine model. By converting the product component data into a control chart sample set for judgment, not only are the results concise and clear, but they are also easy to trace; after obtaining the out-of-control component statistics, the adaptive enhancement-based vector machine model can determine the abnormal offset pattern, thereby automatically processing abnormal data and determining the abnormal pattern of the product. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A schematic flow chart of an embodiment of a method for monitoring product composition data provided by the present invention;

[0028] Figure 2 The structural component data control provided by the present invention Figure 1 Schematic diagram of the process of the embodiment;

[0029] Figure 3 A schematic structural diagram of an embodiment of an initial adaptive enhancement-based vector machine model provided by the present invention;

[0030] Figure 4 A schematic structural diagram of an embodiment of a product composition data monitoring device provided by the present invention;

[0031] Figure 5 This is a structural block diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0032] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings, wherein the accompanying drawings constitute a part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, and are not used to limit the scope of the present invention.

[0033] Before describing the examples, the abnormal patterns of product ingredients are described first:

[0034] Taking the ternary process as an example, since the product ingredient data has fixed and restricted values, if one variable is offset, it will cause another or variables to be offset. Therefore, the situation where only a single variable is offset does not exist.

[0035] For a ternary process (i.e., a product involving only three components), the following abnormal patterns exist: shifts in both the first and second variables, shifts in both the first and third variables, shifts in both the second and third variables, and shifts in all three variables. These 12 abnormal patterns are: (1,-1,0), (1,0,-1), (0,1,-1), (-1,1,0), (-1,0, 1), (0,-1,1); (1,-1,-1), (-1,1,-1), (1,-1,1), (1,1,-1), (1,-1,1), (-1,1, 1). Normal and abnormal states are represented by 0 and ±1, respectively. + represents an upward shift, and - represents a downward shift.

[0036] The details are shown in the figure below:

[0037] Abnormal Mode Abnormal deviation pattern (1,-1,0) Component 1 is shifted upward, component 2 is shifted downward (1,0,-1) Component 1 is shifted upward, component 3 is shifted downward (0,1,-1) Component 2 is shifted upward, and component 3 is shifted downward. (-1,1,0) Component 1 is offset downward, component 2 is offset upward (-1,0,1) Component 1 is offset downward, component 3 is offset upward (0,-1,1) Component 2 is shifted downward, and component 3 is shifted upward. (1,-1,-1) Component 1 is shifted upward, component 2 is shifted downward, and component 3 is shifted downward. (-1,1,-1) Component 1 is offset downward, component 2 is offset upward, and component 3 is offset downward. (1,-1,1) Component 1 is shifted upward, component 2 is shifted downward, and component 3 is shifted upward. (1,1,-1) Component 1 is shifted upward, component 2 is shifted upward, and component 3 is shifted downward. (1,-1,1) Component 1 is shifted upward, component 2 is shifted downward, and component 3 is shifted upward. (-1,1,1) Component 1 is offset downward, component 2 is offset upward, and component 3 is offset upward.

[0038] Currently, product composition data is typically monitored to control the ratios and concentrations of various quality characteristics. However, existing multivariate control charts lack high recognition accuracy, making it difficult to automatically process abnormal data from production data, let alone identify abnormal product patterns based on abnormal production data.

[0039] Therefore, in the process of monitoring product ingredients, the existing technology has the problem of being unable to determine the corresponding specific abnormal pattern based on the out-of-control data in the product ingredient data.

[0040] To solve the above problems, the present invention provides a product ingredient data monitoring method, device, electronic device and storage medium, which are described in detail below.

[0041] like Figure 1 As shown, Figure 1 A flow chart of an embodiment of a method for monitoring product composition data provided by the present invention includes:

[0042] Step S101: Obtain a control chart sample set corresponding to product component data, and filter out an out-of-control component data sample set based on preset conditions.

[0043] Step S102: inputting the out-of-control component data sample set into the initial vector machine model based on adaptive enhancement, taking the abnormal offset pattern corresponding to the out-of-control component data sample set as output, iteratively training, and determining a fully trained vector machine model based on adaptive enhancement.

[0044] Step S103: acquiring real-time out-of-control component data, inputting the real-time out-of-control component data into a well-trained vector machine model based on adaptive enhancement, and determining the abnormal deviation pattern of the product to be tested.

[0045] In this embodiment, based on preset conditions, a sample set of out-of-control component data for a product is obtained from a control chart sample set. Then, an adaptively enhanced vector machine model is used to determine the abnormal excursion pattern corresponding to the out-of-control component data sample set. By using the control chart to determine the product component data sample set, the out-of-control component data sample set provides concise and clear results and facilitates traceability. After obtaining the out-of-control component data sample set, the adaptively enhanced vector machine model can determine the abnormal excursion pattern, enabling automated processing of abnormal data and automatically determining the product's abnormal excursion pattern based on the abnormal data.

[0046] As a preferred embodiment, in step S101, in order to obtain multiple sets of product composition data and ensure the accuracy of the product composition data, it is necessary to pre-process the collected product production data and eliminate abnormal values.

[0047] In a specific embodiment, the collected product production data is subjected to wavelet transform noise reduction processing to obtain product composition data.

[0048] Through wavelet transform noise reduction, when the signal and noise in product production data are decomposed by wavelet at different scales, the signal and noise exhibit diametrically opposite transmission characteristics. That is, the maximum modulus of noise decreases as the wavelet scale increases, while the maximum modulus of signal increases with the scale. This characteristic can be used to remove the noise portion of the signal, and then reconstruct the original signal from the denoised maximum modulus, achieving the purpose of noise removal and making product composition data more accurate.

[0049] Furthermore, after obtaining the product composition data, it is also necessary to obtain a control chart sample set.

[0050] In a specific embodiment, if Figure 2 As shown, Figure 2 The flowchart of an embodiment of obtaining a control diagram sample set provided by this application includes:

[0051] Step S111: Obtain product ingredient data.

[0052] In a specific embodiment, random sampling is used to extract p-dimensional component data with a sample size of n, which is recorded as X i,j ={X i,1 ,X i,2 ,L,X i,n}; where each X i,j All obey normal distribution.

[0053] Step S112: Perform an equidistant logarithmic ratio transformation on the product component data samples to determine the mean coordinate vector samples.

[0054] In a specific embodiment, each component data is subjected to an equidistant logarithmic ratio transformation to obtain a mean coordinate vector, wherein the formula for calculating the mean coordinate vector is:

[0055]

[0056] Among them, Z i,j Obey additive logistic normal distribution.

[0057] Step S113: Determine the statistic sample according to the mean coordinate vector sample and construct a control chart sample set.

[0058] In a specific embodiment, the mean coordinate vector is processed by a multivariate cumulative sum algorithm to obtain a statistical sample; and a control chart sample set is constructed based on the statistical sample.

[0059] The formula for calculating the statistical sample is as follows:

[0060] C i ={(Y i-1 +Z i -μ0)'Σ0 -1 (Y i-1 +Z i -μ0)} 1 / 2

[0061]

[0062] Q i =[Y i 'Σ0 -1 Y i ] 1 / 2

[0063] Among them, Q i is the statistical sample, k is the process offset coefficient, Z i is the mean coordinate vector of the sample, Z i ∈R p-1 ,i=1,2,L.

[0064] In this embodiment, by performing an equidistant logarithmic ratio transformation on the product component data, the product component data is made to obey additive logical normalization, so that the mean coordinate vector can better reflect the characteristics of the component data; the statistical sample of the product is obtained through the multivariate cumulative sum algorithm, and a control chart sample set is constructed to facilitate intuitive acquisition of monitoring results.

[0065] Furthermore, the performance indicators of the control chart sample set include: average run length in a controlled state and average run length in an out-of-control state. A larger average run length in an out-of-control state effectively reduces the false alarm rate of the control chart sample set, while a smaller average run length in a controlled state effectively reduces the false alarm rate of the control chart sample set.

[0066] In order to obtain the optimal value of the average run length under controlled conditions, the Markov chain is used for solution.

[0067] In one embodiment, the formula for the average run length under controlled conditions is:

[0068] ARL0=h'(IP) -1 E

[0069] where ARL0 is the average run length under control, h = (1, 0, 0, ...) is the initial vector of dimension m + 1, I is the identity matrix of dimension (m + 1, m + 1), P is the probability transition matrix of dimension (m + 1, m + 1), and E is the column vector of dimension m + 1.

[0070] In order to obtain the optimal value of the average running length under the out-of-control state, the Monte Carlo simulation method is used to solve it.

[0071] In a specific embodiment, the value of the average run chain length ARL0 under the controlled state is set, and a set of parameters k and H based on the variable cumulative sum control chart of the component data is selected; a random number that obeys the normal distribution and has a deviation coefficient of δ is generated; based on the given parameter k, the statistic sample Q is calculated i ;Record the length of the run chain in the out-of-control state;Repeat the above steps 10 4 times, get 10 4 The expected value of RL is the average running length in the out-of-control state, which is recorded as ARL1.

[0072] In this embodiment, the optimal value of the average running length in the controlled state and the optimal value of the average running length in the uncontrolled state are obtained by calculation, thereby ensuring the performance of the control chart and improving the reliability of the result.

[0073] Furthermore, in step S101, after obtaining the control chart sample set, it is still necessary to screen out the out-of-control component data sample set based on preset conditions.

[0074] In a specific embodiment, an initial controlled average chain length is set based on experience or historical records; then, a Markov chain is used to solve the actual value of the controlled average chain length, an offset reference coefficient k=1 / 2*δ is taken, and the fsolve function in Matlab software is used to solve the nonlinear equation of the actual value of the controlled average chain length in the Markov chain to obtain the control limit at this time, which is the preset condition in this embodiment.

[0075] Furthermore, after determining the preset conditions, the control chart sample set is traversed and compared with the preset conditions. If the control chart sample set is less than the preset conditions, the control chart sample set is judged to be under control; if it is greater than the preset conditions, the control chart sample set is judged to be out of control, recorded as the out-of-control component data sample set, and recorded.

[0076] Through the above method, not only the component data is distinguished and abnormal component data is extracted, but also the abnormal component data is sorted out and a component data control chart is generated for subsequent use.

[0077] Furthermore, in step S102 , after determining the out-of-control component data sample set, it is necessary to analyze the out-of-control component data sample set to determine the abnormal deviation pattern corresponding to the out-of-control component data sample set.

[0078] As a preferred embodiment, in step S102, as Figure 3 As shown, Figure 3 This is a structural diagram of an embodiment of an initial vector machine model based on adaptive enhancement provided by the present invention. The initial vector machine model based on adaptive enhancement includes multiple weak classifiers (A1, A2, A3, A4), and a strong classifier (A) composed of them.

[0079] In this embodiment, a weighted operation is performed on multiple weak classifiers (A1, A2, A3, A4) using an adaptive boosting algorithm to form a strong classifier (A). Then, the data of the strong classifier is mapped into a high-dimensional feature space for classification using a support vector machine algorithm.

[0080] In one specific embodiment, the adaptive boosting algorithm weights weak classifiers to form a strong classifier. Therefore, the performance of the strong classifier depends on the classification results of the weak classifiers. The support vector machine algorithm maps low-dimensional data into a high-dimensional feature space for classification, which to some extent compensates for the problem that the adaptive boosting algorithm's error increases with the number of iterations. Therefore, choosing the support vector machine algorithm as the weak classifier of the adaptive boosting algorithm can improve the classification accuracy of the classification algorithm.

[0081] In other words, by fusing the adaptive enhancement algorithm with the support vector machine algorithm to form an integrated algorithm, the strengths of the two algorithms are complemented to improve the performance of the algorithm.

[0082] In one specific embodiment, the support vector machine algorithm includes its kernel parameters and penalty factors. If the kernel parameters are not appropriately selected, "overlearning" or "underlearning" may occur; if the penalty factors are not appropriately selected, "overfitting" or "underfitting" may occur. To improve the performance of the support vector machine algorithm and ensure the reliability of the results, it is necessary to select the optimal kernel parameters and penalty factors.

[0083] Furthermore, in order to optimize the kernel parameters and penalty factors, the differential evolution algorithm is used.

[0084] The differential evolution algorithm is a heuristic algorithm that mimics the natural process of biological evolution. By randomly searching for differences within a population, it iteratively approaches the optimal solution, ultimately reaching global convergence. This method is currently primarily used to solve problems requiring a global optimal solution or to combine with other algorithms to form hybrid algorithms. The key idea is to take the vector difference between two randomly selected individuals in a population and add a third individual as a mutant. This mutant is then crossed with the target individual to form a new individual, which is then compared with the target individual. The superior individual is then selected for the next generation of the population. This iterative process continuously improves the quality of the population, moving it closer to the optimal solution.

[0085] The differential evolution algorithm primarily seeks the global optimal solution through three steps: mutation, crossover, and selection. Mutation involves obtaining a new individual vector from three individual vectors. The smaller the difference between the first and second individuals, the smaller the impact on the new individual. In the early stages of the algorithm's iterations, due to the large differences between pairs, the random search range becomes larger. In the middle and later stages of the algorithm's iterations, as the differences between pairs decrease, the search range also decreases, and the algorithm approaches the optimal value. Crossover aims to increase the diversity of individuals in the population. During the crossover process, mutant individuals are crossed with target individuals to form new test individuals, ensuring that the test individuals can obtain at least one value from the mutant individuals, thus avoiding invalid crossovers. The goal of selection is to retain the best individuals. Based on the greedy selection principle, fitness values ​​are compared and the better individual between the test and target individuals is retained, ensuring that the fitness value of the offspring individuals is always better than that of the parent individuals.

[0086] The optimal kernel parameters and the optimal penalty factor are determined by the differential evolution algorithm, which improves the reliability of the support vector machine algorithm and ensures the judgment accuracy of the vector machine model based on adaptive enhancement determined by the support vector machine algorithm, thereby effectively ensuring the reliability of the obtained abnormal offset pattern.

[0087] Through the above method, first, by converting the product composition data into a control chart sample set, the composition data is graphically displayed, thereby improving the intuitiveness of the results; then, based on preset conditions, whether the product composition is controllable is judged according to the control chart sample set, thereby realizing preliminary monitoring of the product composition; further, for the out-of-control component data sample set, according to the vector machine model based on adaptive enhancement, its corresponding abnormal deviation pattern is obtained, thereby realizing automatic processing of abnormal data according to the product composition data and determining the abnormal deviation pattern.

[0088] In order to solve the above problems, the present invention also provides a product ingredient data monitoring device, such as Figure 4 As shown, Figure 4 This is a schematic structural diagram of an embodiment of a product composition data monitoring device provided by the present invention. The product composition data monitoring device 400 includes:

[0089] The out-of-control component data sample set acquisition module 401 is used to acquire the control chart sample set corresponding to the product component data and filter out the out-of-control component data sample set based on preset conditions;

[0090] The model training module 402 is used to input the out-of-control component data sample set into the initial adaptive enhancement-based vector machine model, and use the abnormal offset pattern corresponding to the out-of-control component data sample set as the output, iteratively train to determine a fully trained adaptive enhancement-based vector machine model;

[0091] The abnormal deviation pattern determination module 403 is used to obtain real-time out-of-control component data, input the real-time out-of-control component data into a well-trained adaptive enhancement-based vector machine model, and determine the abnormal deviation pattern of the product to be tested.

[0092] The present invention also provides an electronic device, such as Figure 5 As shown, Figure 5 This is a block diagram of an electronic device according to an embodiment of the present invention. Electronic device 500 can be a computing device such as a mobile terminal, desktop computer, notebook, PDA, or server. Electronic device 500 includes a processor 501 and memory 502, wherein memory 502 stores a product ingredient data monitoring program 503.

[0093] In some embodiments, the memory 502 can be an internal storage unit of a computer device, such as a hard disk or memory of the computer device. In other embodiments, the memory 502 can also be an external storage device of the computer device, such as a plug-in hard disk equipped on the computer device, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Furthermore, the memory 502 can also include both an internal storage unit of the computer device and an external storage device. The memory 502 is used to store application software and various types of data installed on the computer device, such as program codes installed on the computer device. The memory 502 can also be used to temporarily store data that has been output or is to be output. In one embodiment, the product ingredient data monitoring program 503 can be executed by the processor 501, thereby realizing the product ingredient data monitoring method of each embodiment of the present invention.

[0094] In some embodiments, the processor 501 may be a central processing unit (CPU), a microprocessor, or other data processing chip, configured to execute program codes or process data stored in the memory 502 , such as executing a product ingredient data monitoring program.

[0095] This embodiment also provides a computer-readable storage medium on which a product ingredient data monitoring program is stored. When the program is executed by a computer processor, the product ingredient data monitoring method described in any of the above technical solutions is implemented.

[0096] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0097] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with this technical field within the technical scope disclosed by the present invention should be covered by the scope of protection of the present invention.

Claims

1. A product ingredient data monitoring method, characterized in that: include: Obtain a control chart sample set corresponding to the product component data, and filter out a sample set of out-of-control component data based on preset conditions; wherein filtering out the sample set of out-of-control component data specifically includes: Obtain product ingredient data, and collect samples of a certain size through random sampling. of dimensional component data, denoted as ; Each All obey the normal distribution; Perform an equidistant logarithmic ratio transformation on the product ingredient data samples to determine the mean coordinate vector samples. The formula for calculating the mean coordinate vector is: in, Obey additive logistic normal distribution; The mean coordinate vector is processed by multivariate cumulative sum algorithm to obtain statistical samples; the control chart sample set is constructed based on the statistical samples; The formula for calculating the statistical sample is as follows: in, is the sample statistic, is the process offset coefficient, is the mean coordinate vector of the sample, ; Inputting the out-of-control component data sample set into an initial adaptive enhancement-based vector machine model, taking the abnormal offset pattern corresponding to the out-of-control component data sample set as output, iteratively training to determine a fully trained adaptive enhancement-based vector machine model, wherein the initial adaptive enhancement-based vector machine model includes multiple weak classifiers, and weighted operations are performed on the multiple weak classifiers through an adaptive enhancement algorithm to form a strong classifier; mapping the data of the strong classifier into a high-dimensional feature space for classification through a support vector machine algorithm; Real-time out-of-control component data is acquired, and the real-time out-of-control component data is input into the well-trained vector machine model based on adaptive enhancement to determine the abnormal deviation pattern of the product to be tested.

2. The product ingredient data monitoring method according to claim 1, characterized in that: The support vector machine algorithm includes kernel parameters and penalty factors, and the kernel parameters and the penalty factors are optimized by a differential evolution algorithm.

3. The product ingredient data monitoring method according to claim 1, characterized in that: The performance indicators of the control chart sample set include an average running length in a controlled state and an average running length in an out-of-control state.

4. The product composition data monitoring method according to claim 3, characterized in that: The average running length in the controlled state is solved by a Markov chain algorithm; and the average running length in the out-of-control state is solved by a Monte Carlo simulation method.

5. The product ingredient data monitoring method according to claim 1, characterized in that: The step of obtaining the control chart sample set includes: Obtaining the product ingredient data; Performing an equidistant logarithmic ratio transformation on a sample of the product composition data to determine a mean coordinate vector sample; A statistical quantity sample is determined according to the mean coordinate vector sample, and a control chart sample set is constructed.

6. The product composition data monitoring method according to claim 5, characterized in that: The obtaining of the product ingredient data includes: The collected product production data is subjected to wavelet transform noise reduction processing to obtain the product composition data.

7. A product ingredient data monitoring device, characterized in that: include: The out-of-control component data sample set acquisition module is used to obtain the control chart sample set corresponding to the product component data and filter out the out-of-control component data sample set based on preset conditions; wherein, filtering out the out-of-control component data sample set specifically includes: Obtain product ingredient data, and collect samples of a certain size through random sampling. of dimensional component data, denoted as ; Each All obey the normal distribution; Perform an equidistant logarithmic ratio transformation on the product ingredient data samples to determine the mean coordinate vector samples. The formula for calculating the mean coordinate vector is: in, Obey additive logistic normal distribution; The mean coordinate vector is processed by multivariate cumulative sum algorithm to obtain statistical samples; the control chart sample set is constructed based on the statistical samples; The formula for calculating the statistical sample is as follows: in, is the sample statistic, is the process offset coefficient, is the mean coordinate vector of the sample, ; A model training module is used to input the out-of-control component data sample set into an initial adaptive enhancement-based vector machine model, and use the abnormal offset pattern corresponding to the out-of-control component data sample set as output, iteratively train to determine a fully trained adaptive enhancement-based vector machine model, wherein the initial adaptive enhancement-based vector machine model includes multiple weak classifiers, and the multiple weak classifiers are weighted by an adaptive enhancement algorithm to form a strong classifier; the data of the strong classifier is mapped into a high-dimensional feature space for classification by a support vector machine algorithm; The abnormal deviation pattern determination module is used to obtain real-time out-of-control component data, input the real-time out-of-control component data into the well-trained adaptive enhancement-based vector machine model, and determine the abnormal deviation pattern of the product to be tested.

8. An electronic device, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the method for monitoring product ingredient data according to any one of claims 1 to 6 is implemented.

9. A storage medium, characterized in that: The storage medium stores computer program instructions, and when the computer program instructions are executed by a computer, the computer is caused to execute the product ingredient data monitoring method according to any one of claims 1 to 6.

Images