Quality control method of rice fine processing based on big data

Abstract

The present invention discloses a method for controlling the quality of rice refined processing based on big data. The method comprises obtaining raw data for rice refined processing, optimizing the raw data, detecting anomalies in rice processing steps, optimizing refined production control parameters, and managing rice processing quality control. The present invention relates to the technical field of rice processing process data processing, and specifically to a method for controlling the quality of rice refined processing based on big data. This method innovatively combines anomaly detection with adjustment of production control parameters, significantly improving the overall quality of rice. It also designs single-step anomaly detection branches and global processing anomaly detection branches to accurately identify and locate the source of anomalies, thereby improving the accuracy and comprehensiveness of anomaly detection. Furthermore, it improves the algorithm for obtaining production control parameters through an optimization strategy that uses search factors and individual position upgrades, thereby obtaining the optimal production control parameters for the rice processing problem step and improving the quality of rice refined processing.

Description

Technical Field

[0001] The present invention relates to the technical field of rice processing flow data processing, and in particular to a rice refined processing quality control method based on big data. Background Art

[0002] As society's requirements for food quality increase, rice, as one of the world's major grains, has a direct impact on consumer health and food safety through its processing quality. Therefore, a rice refined processing quality control method based on big data has emerged. This method is a technical method that uses big data technology and artificial intelligence algorithms to achieve real-time monitoring, anomaly detection and quality optimization of the rice processing process by real-time collection and analysis of multi-dimensional data during the rice processing process, ensuring the refinement, standardization and efficiency of the rice processing process, improving overall production quality and ensuring food safety.

[0003] However, traditional rice processing quality control methods are unable to clearly identify and locate the root causes of rice quality problems, resulting in fluctuations in the quality of the processing process, failure to promptly detect abnormalities in the processing process, and low production efficiency. Existing anomaly detection models for rice processing links have the technical problem of being unable to accurately identify which single link has an abnormality. Existing production control parameter acquisition algorithms may fall into local optimal solutions, resulting in suboptimal production control parameters. Summary of the Invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a rice fine processing quality control method based on big data. In view of the problem that the traditional rice fine processing quality control method cannot clearly identify and locate the root cause of rice quality problems, resulting in fluctuations in the quality of the processing process, failure to timely discover abnormalities in the processing process, and low production efficiency, this solution innovatively proposes to combine anomaly detection and adjustment of production control parameters, which can accurately identify and locate the specific source of rice quality abnormalities, and adjust the production control parameters in a targeted manner through optimization algorithms, solving the problem that quality problems were difficult to trace in the past, ensuring that each processing link is in the optimal configuration, effectively reducing fluctuations in the production process, and improving the overall quality of rice, thereby realizing intelligent and automated control of the production process of rice fine processing; in view of the problem that the existing anomaly detection model applicable to the rice processing link process cannot accurately identify which single link has an abnormality To address the technical problem of abnormalities, this solution innovatively designs a single-link abnormality detection branch and a global processing process abnormality detection branch, which perform abnormality detection on each processing link and the overall production process respectively. This dual detection method can monitor the processing links in real time from both local and global dimensions, accurately identify and locate the source of abnormalities, greatly improving the accuracy and comprehensiveness of abnormality detection, thereby effectively ensuring the stability of the production process; in response to the technical problem that the existing production control parameter acquisition algorithm is trapped in the local optimal solution, resulting in the obtained production control parameters not reaching the optimal solution, this solution improves the algorithm for obtaining production control parameters through the optimization strategy of search factors and individual position upgrades, thereby effectively breaking through the local optimal solution and obtaining the optimal production control parameters for the rice processing problem link, improving the quality of rice fine processing, further enhancing the intelligent control of rice fine processing, and ensuring the maximization of production efficiency and product quality.

[0005] The technical solution adopted by the present invention is as follows: The method for controlling the quality of rice refined processing based on big data provided by the present invention comprises the following steps:

[0006] Step S1: obtaining raw data of rice fine processing;

[0007] Step S2: optimizing the original data;

[0008] Step S3: Detecting abnormalities in the rice processing process;

[0009] Step S4: Optimizing and refining production control parameters;

[0010] Step S5: Rice processing quality control management.

[0011] Furthermore, in step S1, the rice fine processing raw data is obtained by collecting the rice fine processing raw data from the rice fine processing factory management system; the rice fine processing raw data includes fine production link data, rice quality data after fine processing, fine processing environment data and fine production control parameter data.

[0012] Furthermore, in step S2, the optimized raw data is used to optimize the raw data of rice fine processing, specifically performing data cleaning optimization, data standardization optimization and feature optimization on the raw data of rice fine processing to obtain rice fine processing optimized data, and dividing training data and data to be tested; the data cleaning optimization is used to eliminate invalid and inaccurate data, specifically performing missing value processing, data outlier processing and data duplicate value deletion processing on the data; the data standardization optimization is specifically to standardize the data using the maximum and minimum normalization method; the feature optimization is specifically to perform statistical feature selection on the data to screen out features related to quality impact.

[0013] Furthermore, in step S3, the rice processing process anomaly detection is used to identify production links with problems in the rice processing process, specifically by establishing a processing link anomaly detection model, anomaly detection model training and processing link anomaly detection, to obtain problematic links in the rice processing process;

[0014] Step S31: Establishing a processing link process anomaly detection model, which is used to construct a processing link process anomaly detection model, by analyzing the rice quality data after fine processing, the fine processing environment data and the fine production link data, determining which specific link in the rice fine processing process has an abnormal problem, and obtaining the processing link process anomaly detection model; specifically comprising the following steps:

[0015] Step S311: Single-link anomaly detection branch design is used to monitor a single link in the rice processing process, thereby identifying the specific link that causes poor quality of rice after fine processing, specifically including the following steps:

[0016] Step S3111: Extract local features of a single processing step. Specifically, use depthwise separable convolution and pointwise convolution to extract local features of each processing step. The formula used is as follows:

[0017] ;

[0018] ;

[0019] Where, It represents the feature map obtained after the depth convolution operation of the cth channel in the i-th processing link. Indicates the input data of the cth channel in the i-th processing link, represents the depth convolution kernel of the cth channel in the i-th processing link, represents the convolution operation, It represents the output single-link local features after the point-by-point convolution operation in the i-th processing link, Represents the point-by-point convolution kernel for the i-th processing link, which is used to fuse features from different channels;

[0020] Step S3112: Obtain weighted single-link features to highlight the important production control parameters of each link. Specifically, global average pooling and channel weighting mechanisms are introduced to calculate the weighted features of each processing link. The formula used is as follows:

[0021] ;

[0022] ;

[0023] Where, represents the global characteristics of the i-th processing link, represents the eigenvalue in the feature map of the i-th processing link, H and W represent the height and width in the feature map respectively, and w represent the height and width coordinates of the feature map respectively, represents the weighted feature of the i-th link, represents the Sigmoid activation function, and represents the weight matrix;

[0024] Step S3113: Obtain the single-link abnormality probability. Specifically, the abnormality probability of each processing link is calculated by combining the fully connected layer. The formula used is as follows:

[0025] ;

[0026] Where, represents the abnormal probability of the i-th link, Represents the weight matrix of the abnormal probability output of the i-th link, Represents the bias parameter of the abnormal probability output of the i-th link;

[0027] Step S312: Designing a global processing anomaly detection branch, which is used to analyze each link in the rice processing process from a global perspective and identify the link that causes quality problems in the rice after processing; specifically, the following steps are included:

[0028] Step S3121: Capture the temporal dependencies between processing links. Specifically, the forward and backward dependencies between processing links can be captured through the Bi-LSTM bidirectional structure. The formula used is as follows:

[0029] ;

[0030] Where, Indicates the unit running function, Indicates the hidden state at the current moment, Represents the input data at the current moment, Indicates the hidden state in the forward direction at the previous moment, Indicates the reverse hidden state at the next moment;

[0031] Step S3122: Calculate the weighted features of the processing links. Specifically, the multi-head self-attention mechanism is used to adaptively assign weights to different processing links, and finally obtain the weighted features of the processing links. The formula used is as follows:

[0032] ;

[0033] Where H represents the temporal dependency between processing links, represents the r-th query matrix, represents the rth bond matrix, represents the r-th value matrix, represents the weight matrix mapped to the query matrix, represents the weight matrix mapped to the key matrix, represents the weight matrix mapped to the value matrix, represents the output of the rth attention head, represents the dimension of the key matrix, T represents the transpose operation, represents the weighted characteristics of the processing link, represents the concatenation function, Represents the weight matrix of the weighted feature output of the processing link, represents the output of the first attention head, represents the output of the second attention head, represents the output of the h-th attention head, and h represents the total number of attention heads;

[0034] Step S3123: Obtain the abnormal probability of each link globally, using the following formula:

[0035] ;

[0036] Where, represents the abnormal probability of the global i-th link, Represents the weight matrix of the global abnormal probability output of the i-th link, Represents the bias parameter of the global abnormal probability output of the i-th link;

[0037] Step S313: Obtain the probability fusion weight, using the following formula:

[0038] ;

[0039] Where, represents the abnormal probability fusion weight of the i-th link, Represents the weight matrix for obtaining the abnormal probability fusion weight, Indicates the bias parameter for obtaining the abnormal probability fusion weight;

[0040] Step S314: Obtain the final abnormality probability of each processing link. The formula used is as follows:

[0041] ;

[0042] Where, represents the final abnormal probability of the i-th link;

[0043] Step S315: Determine if the processing link is abnormal. , then the link is judged to be abnormal, indicating that there are potential problems in this link, which will affect the quality of rice after fine processing; if , then the link is judged to be normal, indicating that the link will not affect the quality of rice after fine processing;

[0044] Step S32: anomaly detection model training, used to train the processing anomaly detection model, specifically using a binary cross entropy loss function as the loss function of the model, and using the training data to train the processing anomaly detection model, updating the processing anomaly detection model parameters through the back propagation algorithm and the gradient descent method to obtain a trained processing anomaly detection model;

[0045] Step S33: Processing link anomaly detection is used to obtain the results of process problems in the rice processing link. Specifically, the data to be detected is used as input data of the trained processing link process anomaly detection model to obtain the problem link in the rice processing process. The problem link in the rice processing process is specifically the link that affects the quality of rice after fine processing.

[0046] Furthermore, in step S4, the optimization and refinement of production control parameters is used to optimize the production control parameters in the problematic links in the rice processing process; specifically, the following steps are included:

[0047] Step S41: Initialize the individual positions of the population, specifically randomly initialize the position of each search individual in the population, and the individual position represents the production control parameter in the production process; the formula used is as follows:

[0048] ;

[0049] Where, represents the initialization position of the dth dimension of the i-th individual, n represents the number of individuals in the population, represents the initialization position of the first dimension in the i-th individual, represents the initialization position of the second dimension in the i-th individual, represents the initialization position of the Dth dimension in the i-th individual, D represents the number of production control parameters in the production link, and each dimension represents each production control parameter in the production link;

[0050] Step S42: Calculate the individual fitness value, specifically calculate the individual fitness value f in the population i ; The quality of rice after fine processing is used as the fitness value of the individual, and the individuals are sorted from best to worst according to the fitness value, and the position of the individual with the highest current global fitness value is obtained. ;

[0051] Step S43: Individual location update, specifically by searching for factors to control the individual location update strategy; the formula used is as follows:

[0052] ;

[0053] ;

[0054] Where, represents the search factor for the t-th iteration, Represents the initialization search factor, t represents the current number of iterations, represents the maximum number of iterations, A constant that adjusts the decay rate of the balance factor. A constant that controls the degree of influence of group fitness on the balance factor. represents the group fitness of the tth iteration, which is the average fitness value of all individuals. represents the number of individuals i in the tth generation population. dimensional positions, represents a random dimension in dimension, represents a random search individual, represents the d-th dimension position of the i-th individual in the t+1 generation population, represents the d-th dimension position of the i-th individual in the t-th generation population, represents the position of the individual with the best fitness value in the population in the tth iteration on dimension d, 、 、 、 and Indicates between A random number uniformly distributed in the range, Represents an even number, It is represented as an odd number;

[0055] Step S44: Individual position upgrade, specifically, individual position upgrade is performed based on the current fitness and search status. The formula used is as follows:

[0056] ;

[0057] Where, and Indicates between A random number uniformly distributed in the range, shows the position of the i-th individual in the t+1 generation population after the d-th dimension upgrade;

[0058] Step S45: updating the current optimal solution, specifically recalculating the fitness values ​​of the individuals after position upgrade and the current individual, and comparing them with the fitness of the current optimal solution. If the fitness of the current solution is better than the global optimal solution, then updating the optimal solution;

[0059] Step S46: Obtain the optimal individual position, specifically by i When the fitness threshold is higher than the set fitness threshold and the maximum number of iterations is reached, the search is terminated and the individual global optimal position is obtained. The individual global optimal position specifically refers to the optimal production control parameters of the rice processing problem link.

[0060] Furthermore, in step S5, the rice processing quality control management readjusts the production control parameters of various links in the rice fine processing process according to the optimal production control parameters of the rice processing problem link, so as to ensure the refinement and standardization of the rice fine processing process and improve the processing quality of rice.

[0061] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0062] (1) In view of the fact that the traditional rice fine processing quality control method cannot clearly identify and locate the root cause of rice quality problems, resulting in fluctuations in the quality of the processing process, failure to timely detect abnormalities in the processing process, and low production efficiency, this solution innovatively proposes to combine abnormality detection and adjustment of production control parameters, which can accurately identify and locate the specific source of rice quality abnormalities, and adjust production control parameters in a targeted manner through optimization algorithms, solving the problem of difficult tracing of quality problems in the past, ensuring that each processing link is in the optimal configuration, effectively reducing fluctuations in the production process, and improving the overall quality of rice, thereby realizing intelligent and automated control of the production process of rice fine processing.

[0063] (2) The existing anomaly detection model for rice processing has the technical problem of being unable to accurately identify which single link has an anomaly. This solution innovatively designs a single-link anomaly detection branch and a global processing anomaly detection branch to perform anomaly detection on each processing link and the overall production process respectively. This dual detection method can monitor the processing links in real time from both local and global dimensions, accurately identify and locate the source of anomalies, greatly improving the accuracy and comprehensiveness of anomaly detection, thereby effectively ensuring the stability of the production process.

[0064] (3) Aiming at the technical problem that the existing production control parameter acquisition algorithm falls into the local optimal solution, resulting in the obtained production control parameters not reaching the optimal level, this scheme improves the algorithm for obtaining production control parameters through the optimization strategy of search factor and individual position upgrade, so as to effectively break through the local optimal solution and obtain the optimal production control parameters for the rice processing problem link, thereby improving the quality of rice fine processing, further enhancing the intelligent control of rice fine processing, and ensuring the maximization of production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A schematic diagram of the process of the rice fine processing quality control method based on big data provided by the present invention;

[0066] Figure 2 Schematic diagram of the process of step S3;

[0067] Figure 3 Schematic diagram of the process of step S4;

[0068] Figure 4 is a flow chart of step S31;

[0069] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0070] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0071] In the description of the present invention, it should be understood that terms such as "up", "down", "front", "back", "left", "right", "top", "bottom", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, they should not be understood as limiting the present invention.

[0072] Example 1, see Figure 1 The technical solution adopted by the present invention is as follows: The method for controlling the quality of rice processing based on big data provided by the present invention comprises the following steps:

[0073] Step S1: obtaining raw data of rice fine processing, specifically collecting raw data of rice fine processing from a management system of a rice fine processing factory;

[0074] Step S2: Optimizing the raw data, which is used to optimize the rice fine processing raw data, specifically performing data cleaning optimization, data standardization optimization and feature optimization on the rice fine processing raw data to obtain rice fine processing optimized data;

[0075] Step S3: rice processing process anomaly detection, which is used to identify production links with problems in the rice processing process, specifically by establishing a processing process anomaly detection model and training the anomaly detection model, and finally using the data to be detected as input data for the trained processing process anomaly detection model to obtain problematic links in the rice processing process;

[0076] Step S4: optimizing and refining production control parameters, which are used to optimize the production control parameters in the problematic link in the rice processing process, and obtaining the optimal production control parameters for the problematic link in rice processing by improving the algorithm for obtaining the production control parameters;

[0077] Step S5: rice processing quality control management, readjusting the production control parameters of each link according to the optimal production control parameters of the rice processing problem link.

[0078] By performing the above operations, the traditional rice fine processing quality control method cannot clearly identify and locate the root cause of rice quality problems, resulting in fluctuations in the quality of the processing process, the inability to timely discover abnormalities in the processing process, and low production efficiency. This solution innovatively proposes to combine abnormality detection and adjustment of production control parameters. It can accurately identify and locate the specific source of rice quality abnormalities, and adjust production control parameters in a targeted manner through optimization algorithms, solving the problem of difficult tracing of quality problems in the past, ensuring that each processing link is in the optimal configuration, effectively reducing fluctuations in the production process, and improving the overall quality of rice, thereby realizing intelligent and automated control of the production process of rice fine processing.

[0079] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, the raw data of rice fine processing is obtained, specifically, the raw data of rice fine processing is obtained by collecting from the management system of the rice fine processing factory; the raw data of rice fine processing includes data of fine production links, quality data of rice after fine processing, fine processing environment data and fine production control parameter data; the fine production link data includes rice raw material data, cleaning link control parameter data, shelling link control parameter data, peeling link control parameter data, polishing link control parameter data, grading link control parameter data, drying link control parameter data, cooling link control parameter data and rice quality data of each link; the quality data of rice after fine processing includes rice appearance quality, rice particle size, rice whiteness, rice moisture content, breakage rate, nutritional components and rice impurity content; the fine processing environment data includes temperature, humidity, air flow rate, etc. degree, air flow temperature, light intensity and air quality; the cleaning link control parameter data includes screen aperture parameters, vibration frequency parameters, temperature parameters, time parameters and flow parameters; the shelling link control parameter data includes pressure parameters, screen aperture parameters, time parameters, speed parameters and temperature parameters; the peeling link control parameter data includes pressure parameters, time parameters and temperature parameters; the polishing link control parameter data includes time parameters, speed parameters, pressure parameters and temperature parameters; the grading link control parameter data includes screen aperture parameters and vibration frequency parameters; the drying link control parameter data includes temperature parameters, time parameters and flow parameters; the cooling link control parameter data includes temperature parameters and time parameters; the refined production control parameter data specifically refers to the control parameters of the processing link, including temperature parameters, humidity parameters, pressure parameters, screen aperture parameters, speed parameters, flow parameters, time parameters and vibration frequency parameters.

[0080] Example 3, see Figure 1This embodiment is based on the above embodiment. In step S2, the optimized original data is used to optimize the original data of rice fine processing, specifically performing data cleaning optimization, data standardization optimization and feature optimization on the original data of rice fine processing to obtain rice fine processing optimized data, and dividing training data and data to be tested; the data cleaning optimization is used to eliminate invalid and inaccurate data, specifically performing missing value processing, data outlier processing and data duplicate value deletion processing on the data; the data standardization optimization is specifically to standardize the data using the maximum and minimum normalization method; the feature optimization is specifically to perform statistical feature selection on the data to screen out features related to quality impact.

[0081] Example 4, see Figure 1 、 Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S3, the rice processing process anomaly detection is used to identify the production links with problems in the rice processing process. Specifically, a processing link anomaly detection model is established, the anomaly detection model is trained, and the processing link anomaly detection is performed to obtain the problem link in the rice processing process.

[0082] Step S31: Establishing a processing link process anomaly detection model, which is used to construct a processing link process anomaly detection model, by analyzing the rice quality data after fine processing, the fine processing environment data and the fine production link data, determining which specific link in the rice fine processing process has an abnormal problem, and obtaining the processing link process anomaly detection model; specifically comprising the following steps:

[0083] Step S311: Single-link anomaly detection branch design is used to monitor a single link in the rice processing process, thereby identifying the specific link that causes poor quality of rice after fine processing, specifically including the following steps:

[0084] Step S3111: Extract local features of a single processing step. Specifically, use depthwise separable convolution and pointwise convolution to extract local features of each processing step. The formula used is as follows:

[0085] ;

[0086] ;

[0087] Where, It represents the feature map obtained after the depth convolution operation of the cth channel in the i-th processing link. Indicates the input data of the cth channel in the i-th processing link, represents the depth convolution kernel of the cth channel in the i-th processing link, represents the convolution operation, It represents the output single-link local features after the point-by-point convolution operation in the i-th processing link, Represents the point-by-point convolution kernel for the i-th processing link, which is used to fuse features from different channels;

[0088] Step S3112: Obtain weighted single-link features to highlight the important production control parameters of each link. Specifically, global average pooling and channel weighting mechanisms are introduced to calculate the weighted features of each processing link. The formula used is as follows:

[0089] ;

[0090] ;

[0091] Where, represents the global characteristics of the i-th processing link, represents the eigenvalue in the feature map of the i-th processing link, H and W represent the height and width in the feature map respectively, and w represent the height and width coordinates of the feature map respectively, represents the weighted feature of the i-th link, represents the Sigmoid activation function, and represents the weight matrix;

[0092] Step S3113: Obtain the single-link abnormality probability. Specifically, the abnormality probability of each processing link is calculated by combining the fully connected layer. The formula used is as follows:

[0093] ;

[0094] Where, represents the abnormal probability of the i-th link, Represents the weight matrix of the abnormal probability output of the i-th link, Represents the bias parameter of the abnormal probability output of the i-th link;

[0095] Step S312: Designing a global processing anomaly detection branch, which is used to analyze each link in the rice processing process from a global perspective and identify the link that causes quality problems in the rice after processing; specifically, the following steps are included:

[0096] Step S3121: Capture the temporal dependencies between processing links. Specifically, the forward and backward dependencies between processing links can be captured through the Bi-LSTM bidirectional structure. The formula used is as follows:

[0097] ;

[0098] Where, Indicates the unit running function, Indicates the hidden state at the current moment, Represents the input data at the current moment, Indicates the hidden state in the forward direction at the previous moment, Indicates the reverse hidden state at the next moment;

[0099] Step S3122: Calculate the weighted features of the processing links. Specifically, the multi-head self-attention mechanism is used to adaptively assign weights to different processing links, and finally obtain the weighted features of the processing links. The formula used is as follows:

[0100] ;

[0101] Where H represents the temporal dependency between processing links, represents the r-th query matrix, represents the rth bond matrix, represents the r-th value matrix, represents the weight matrix mapped to the query matrix, represents the weight matrix mapped to the key matrix, represents the weight matrix mapped to the value matrix, represents the output of the rth attention head, represents the dimension of the key matrix, T represents the transpose operation, represents the weighted characteristics of the processing link, represents the concatenation function, Represents the weight matrix of the weighted feature output of the processing link, represents the output of the first attention head, represents the output of the second attention head, represents the output of the h-th attention head, and h represents the total number of attention heads;

[0102] Step S3123: Obtain the abnormal probability of each link globally, using the following formula:

[0103] ;

[0104] Where, represents the abnormal probability of the global i-th link, Represents the weight matrix of the global abnormal probability output of the i-th link, Represents the bias parameter of the global abnormal probability output of the i-th link;

[0105] Step S313: Obtain the probability fusion weight, using the following formula:

[0106] ;

[0107] Where, represents the abnormal probability fusion weight of the i-th link, Represents the weight matrix for obtaining the abnormal probability fusion weight, Indicates the bias parameter for obtaining the abnormal probability fusion weight;

[0108] Step S314: Obtain the final abnormality probability of each processing link. The formula used is as follows:

[0109] ;

[0110] Where, represents the final abnormal probability of the i-th link;

[0111] Step S315: Determine if the processing link is abnormal. , then the link is judged as abnormal, indicating that there are potential problems in this link, which will affect the quality of rice after fine processing; if , then the link is judged to be normal, indicating that the link will not affect the quality of rice after fine processing;

[0112] Step S32: anomaly detection model training, used to train the processing anomaly detection model, specifically using a binary cross entropy loss function as the loss function of the model, and using the training data to train the processing anomaly detection model, updating the processing anomaly detection model parameters through the back propagation algorithm and the gradient descent method to obtain a trained processing anomaly detection model;

[0113] Step S33: Processing link anomaly detection is used to obtain the results of the rice processing link process problems. Specifically, the data to be detected is used as a trained processing link process anomaly detection model to obtain the problem link in the rice processing process. The problem link in the rice processing process is specifically the link that affects the quality of the rice after fine processing.

[0114] By performing the above operations, in order to address the technical problem that the existing anomaly detection model applicable to the rice processing process cannot accurately determine which single link has an anomaly, this solution innovatively designs a single-link anomaly detection branch and a global processing process anomaly detection branch to perform anomaly detection on each processing link and the overall production process respectively. This dual detection method can monitor the processing links in real time from both local and global dimensions, accurately identify and locate the source of anomalies, greatly improving the accuracy and comprehensiveness of anomaly detection, thereby effectively ensuring the stability of the production process.

[0115] Example 5, see Figure 1 and Figure 3This embodiment is based on the above embodiment. In step S4, the optimization and refinement of production control parameters are used to optimize the production control parameters in the problematic links in the rice processing process. Specifically, the following steps are included:

[0116] Step S41: Initialize the individual positions of the population, specifically randomly initialize the position of each search individual in the population, and the individual position represents the production control parameter in the production process; the formula used is as follows:

[0117] ;

[0118] Where, represents the initialization position of the dth dimension of the i-th individual, n represents the number of individuals in the population, represents the initialization position of the first dimension in the i-th individual, represents the initialization position of the second dimension in the i-th individual, represents the initialization position of the Dth dimension in the i-th individual, D represents the number of production control parameters in the production link, and each dimension represents each production control parameter in the production link;

[0119] Step S42: Calculate the individual fitness value, specifically calculate the individual fitness value f in the population i ; The quality of rice after fine processing is used as the fitness value of the individual, and the individuals are sorted from best to worst according to the fitness value, and the position of the individual with the highest current global fitness value is obtained. ;

[0120] Step S43: Individual location update, specifically by searching for factors to control the individual location update strategy; the formula used is as follows:

[0121] ;

[0122] ;

[0123] Where, represents the search factor for the t-th iteration, Represents the initialization search factor, t represents the current number of iterations, represents the maximum number of iterations, A constant that adjusts the decay rate of the balance factor. A constant that controls the degree of influence of group fitness on the balance factor. represents the group fitness of the tth iteration, which is the average fitness value of all individuals. represents the number of individuals i in the tth generation population. dimensional positions, represents a random dimension in dimension, represents a random search individual, represents the d-th dimension position of the i-th individual in the t+1 generation population, represents the d-th dimension position of the i-th individual in the t-th generation population, represents the position of the individual with the best fitness value in the population in the tth iteration on dimension d, 、 、 、 and Indicates between A random number uniformly distributed in the range, Represents an even number, It is represented as an odd number;

[0124] Step S44: Individual position upgrade, specifically, individual position upgrade is performed based on the current fitness and search status. The formula used is as follows:

[0125] ;

[0126] Where, and Indicates between A random number uniformly distributed in the range, shows the position of the i-th individual in the t+1 generation population after the d-th dimension upgrade;

[0127] Step S45: updating the current optimal solution, specifically recalculating the fitness values ​​of the individuals after position upgrade and the current individual, and comparing them with the fitness of the current optimal solution. If the fitness of the current solution is better than the global optimal solution, then updating the optimal solution;

[0128] Step S46: Obtain the optimal individual position, specifically by i When the fitness threshold is higher than the set fitness threshold and the maximum number of iterations is reached, the search is terminated and the individual global optimal position is obtained. The individual global optimal position specifically refers to the optimal production control parameters of the rice processing problem link.

[0129] By performing the above operations, in order to address the technical problem that the existing production control parameter acquisition algorithm may fall into a local optimal solution, resulting in the obtained production control parameters not being optimal, this solution improves the algorithm for obtaining production control parameters through an optimization strategy of search factors and individual position upgrades, thereby effectively breaking through the local optimal solution and obtaining the optimal production control parameters for the rice processing problem link, thereby improving the quality of rice fine processing, further enhancing the intelligent control of rice fine processing, and ensuring the maximization of production efficiency and product quality.

[0130] Example 6, see Figure 1This embodiment is based on the above embodiment. In step S5, the rice processing quality control management readjusts various production control parameters in the rice fine processing process according to the optimal production control parameters of the rice processing problem link, ensuring the refinement and standardization of the rice fine processing process and improving the processing quality of rice.

[0131] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0132] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that various changes, modifications, substitutions, and alterations can be made to the embodiments without departing from the principles and spirit of the invention.

[0133] The present invention and its embodiments are described above. This description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by this and, without departing from the purpose of the present invention, designs structures and embodiments similar to this technical solution without inventiveness, they shall fall within the scope of protection of the present invention.

Claims

1. A rice processing quality control method based on big data, characterized by: The method comprises the following steps: Step S1: obtaining raw data of rice fine processing, and obtaining raw data of rice fine processing by collecting data; Step S2: Optimizing the original data, which is used to optimize the original data of rice fine processing to obtain optimized data of rice fine processing; Step S3: Rice processing process anomaly detection, which is used to identify production links with problems in the rice processing process. Specifically, a processing process anomaly detection model is established through a double-branch structure and the anomaly detection model is trained. Finally, the data to be detected is used as input data of the trained processing process anomaly detection model to obtain the problematic links in the rice processing process; the steps include: Step S31: Establishing a processing process anomaly detection model for constructing a processing process anomaly detection model to obtain a processing process anomaly detection model; Step S311: Single-link anomaly detection branch design is used to monitor a single link in the rice processing process, thereby identifying the specific link that causes poor quality of rice after fine processing; Step S312: Designing a global processing anomaly detection branch, which is used to analyze each link in the rice processing process from a global perspective and identify the link that causes quality problems in the rice after processing; Step S313: Obtain the probability fusion weight, using the following formula: ； Where, represents the abnormal probability fusion weight of the i-th link, Represents the weight matrix for obtaining the abnormal probability fusion weight, Indicates the bias parameter for obtaining the abnormal probability fusion weight; Step S314: Obtain the final abnormality probability of each processing link. The formula used is as follows: ； Where, represents the final abnormal probability of the i-th link; Step S315: Determine if the processing link is abnormal. , then the link is judged to be abnormal, indicating that there are potential problems in this link, which will affect the quality of rice after fine processing; if , then the link is judged to be normal, indicating that the link will not affect the quality of rice after fine processing; Step S32: anomaly detection model training, used to train the processing anomaly detection model, specifically using a binary cross entropy loss function as the loss function of the model, and using the training data to train the processing anomaly detection model, updating the processing anomaly detection model parameters through the back propagation algorithm and the gradient descent method to obtain a trained processing anomaly detection model; Step S33: Processing link anomaly detection, used to obtain the results of the rice processing link process problems, specifically using the data to be detected as input data of the trained processing link process anomaly detection model to obtain the problem link in the rice processing process, wherein the problem link in the rice processing process is specifically the link that affects the quality of the rice after fine processing; Step S4: Optimizing and refining production control parameters, which are used to optimize the production control parameters in the problematic link in the rice processing process, by improving the algorithm for obtaining the production control parameters through an optimization strategy of search factors and individual position upgrades, and obtaining the optimal production control parameters for the problematic link in rice processing; Step S5: rice processing quality control management, readjusting the production control parameters of each link according to the optimal production control parameters of the rice processing problem link.

2. The rice processing quality control method based on big data according to claim 1, wherein: In step S311, the single-link anomaly detection branch design specifically includes the following steps: Step S3111: Extract local features of a single processing step. Specifically, use depthwise separable convolution and pointwise convolution to extract local features of each processing step. The formula used is as follows: ； ； Where, It represents the feature map obtained after the depth convolution operation of the cth channel in the i-th processing link. Indicates the input data of the cth channel in the i-th processing link, represents the depth convolution kernel of the cth channel in the i-th processing link, represents the convolution operation, It represents the output single-link local features after the point-by-point convolution operation in the i-th processing link, Represents the point-by-point convolution kernel for the i-th processing link, which is used to fuse features from different channels; Step S3112: Obtain weighted single-link features to highlight the important production control parameters of each link. Specifically, global average pooling and channel weighting mechanisms are introduced to calculate the weighted features of each processing link. The formula used is as follows: ； ； Where, represents the global characteristics of the i-th processing link, represents the eigenvalue in the feature map of the i-th processing link, H and W represent the height and width in the feature map respectively, and w represent the height and width coordinates of the feature map respectively, represents the weighted feature of the i-th link, represents the Sigmoid activation function, and represents the weight matrix; Step S3113: Obtain the single-link abnormality probability. Specifically, the abnormality probability of each processing link is calculated by combining the fully connected layer. The formula used is as follows: ； Where, represents the abnormal probability of the i-th link, Represents the weight matrix of the abnormal probability output of the i-th link, Represents the bias parameter of the abnormal probability output of the i-th link.

3. The rice processing quality control method based on big data according to claim 1, wherein: In step S312: the global machining process abnormality detection branch design specifically includes the following steps: Step S3121: Capture the temporal dependencies between processing links. Specifically, the forward and backward dependencies between processing links can be captured through the Bi-LSTM bidirectional structure. The formula used is as follows: ； Where, Indicates the unit running function, Indicates the hidden state at the current moment, Represents the input data at the current moment, Indicates the hidden state in the forward direction at the previous moment, Indicates the hidden state in the reverse direction at the next moment; Step S3122: Calculate the weighted features of the processing links. Specifically, the multi-head self-attention mechanism is used to adaptively assign weights to different processing links, and finally obtain the weighted features of the processing links. The formula used is as follows: ； Where H represents the temporal dependency between processing links, represents the r-th query matrix, represents the rth bond matrix, represents the r-th value matrix, represents the weight matrix mapped to the query matrix, represents the weight matrix mapped to the key matrix, represents the weight matrix mapped to the value matrix, represents the output of the rth attention head, represents the dimension of the key matrix, T represents the transpose operation, represents the weighted characteristics of the processing link, represents the concatenation function, Represents the weight matrix of the weighted feature output of the processing link, represents the output of the first attention head, represents the output of the second attention head, represents the output of the h-th attention head, and h represents the total number of attention heads; Step S3123: Obtain the abnormal probability of each link globally, using the following formula: ； Where, represents the abnormal probability of the global i-th link, Represents the weight matrix of the global abnormal probability output of the i-th link, Represents the bias parameter of the global abnormal probability output of the i-th link.

4. The rice processing quality control method based on big data according to claim 1, wherein: In step S4, the optimization and refinement of production control parameters is used to optimize the production control parameters in the problematic links in the rice processing process; specifically, the following steps are included: Step S41: Initialize the individual positions of the population, specifically randomly initialize the position of each search individual in the population, and the individual position represents the production control parameter in the production process; the formula used is as follows: ； Where, represents the initialization position of the dth dimension of the i-th individual, n represents the number of individuals in the population, represents the initialization position of the first dimension in the i-th individual, represents the initialization position of the second dimension in the i-th individual, represents the initialization position of the Dth dimension in the i-th individual, D represents the number of production control parameters in the production link, and each dimension represents each production control parameter in the production link; Step S42: Calculate the individual fitness value, specifically calculate the individual fitness value f in the population i ; The quality of rice after fine processing is used as the fitness value of the individual, and the individuals are sorted from best to worst according to the fitness value, and the position of the individual with the highest current global fitness value is obtained. ; Step S43: Individual location update, specifically by searching for factors to control the individual location update strategy; the formula used is as follows: ； ； Where, represents the search factor for the t-th iteration, Represents the initialization search factor, t represents the current number of iterations, represents the maximum number of iterations, A constant that adjusts the decay rate of the balance factor. A constant that controls the degree of influence of group fitness on the balance factor. represents the group fitness of the tth iteration, which is the average fitness value of all individuals. represents the number of individuals i in the tth generation population. dimensional positions, represents a random dimension in dimension, represents a random search individual, represents the d-th dimension position of the i-th individual in the t+1 generation population, represents the d-th dimension position of the i-th individual in the t-th generation population, represents the position of the individual with the best fitness value in the population in the tth iteration on dimension d, 、 、 、 and Indicates between A random number uniformly distributed in the range, Represents an even number, It is represented as an odd number; Step S44: Individual position upgrade, specifically, individual position upgrade is performed based on the current fitness and search status. The formula used is as follows: ； Where, and Indicates between A random number uniformly distributed in the range, shows the position of the i-th individual in the t+1 generation population after the d-th dimension upgrade; Step S45: updating the current optimal solution, specifically recalculating the fitness values ​​of the individuals after position upgrade and the current individual, and comparing them with the fitness of the current optimal solution. If the fitness of the current solution is better than the global optimal solution, then updating the optimal solution; Step S46: Obtain the optimal individual position, specifically by i When the fitness threshold is higher than the set fitness threshold and the maximum number of iterations is reached, the search is terminated and the individual global optimal position is obtained. The individual global optimal position specifically refers to the optimal production control parameters of the rice processing problem link.

5. The rice processing quality control method based on big data according to claim 1, wherein: In step S5, the rice processing quality control management readjusts the production control parameters of various links in the rice fine processing process according to the optimal production control parameters of the rice processing problem link, ensures the refinement and standardization of the rice fine processing process, and improves the processing quality of rice.

6. The rice fine processing quality control method based on big data according to claim 1, wherein: In step S1, the raw data of rice fine processing is obtained by collecting the raw data from the management system of the rice fine processing factory; the raw data of rice fine processing includes data of fine production links, quality data of rice after fine processing, fine processing environment data and fine production control parameter data.

7. The rice fine processing quality control method based on big data according to claim 1, wherein: In step S2, the raw data is optimized, specifically performing data cleaning optimization, data standardization optimization and feature optimization on the raw data of rice fine processing to obtain rice fine processing optimized data, and dividing training data and data to be tested; the data cleaning optimization is used to eliminate invalid and inaccurate data, specifically performing missing value processing, data outlier processing and data duplicate value deletion processing on the data; the data standardization optimization is specifically performing standardization processing on the data using the maximum and minimum normalization method; the feature optimization is specifically performing statistical feature selection on the data to screen out features related to quality impact.

Images