A food production information traceability management method and system

Abstract

The present invention provides a method and system for traceability management of food production information, relating to the technical field of food production monitoring. The method includes: obtaining multiple groups of multi-process production record data of a target baked food, performing a production process characteristic linkage analysis on the target baked food to generate multiple process joint feature matrices; performing pattern clustering based on production process distance perception on the multiple groups of multi-process production record data to generate multiple production anomaly pattern clusters; performing process anomaly analysis and backpropagation traceability on the batch to be analyzed to generate a process anomaly detection result of the batch to be analyzed; performing pattern anomaly traceability on the batch to be analyzed according to the multiple production anomaly pattern clusters to generate a production anomaly pattern matching result of the batch to be analyzed, and fusing to generate a production anomaly traceability analysis result of the batch to be analyzed. The present invention realizes the improvement of the positioning accuracy of the abnormal source process of baked foods.

Description

Technical Field

[0001] The present invention relates to the technical field of food production monitoring, and particularly to a method and system for tracing and managing food production information. Background Art

[0002] In the food production industry, especially for the production process of baked foods such as bread and biscuits, tracing the quality anomalies is a key link to ensure food quality and reduce losses.

[0003] The production process of baked foods involves multiple links and processes, such as raw material preparation, fermentation, mixing, baking, cooling, and packaging, etc. The quality control of each link may have an important impact on subsequent processes. In the process of tracing and analyzing production information, if the mutual influence relationship between processes is ignored, and a scheme that only relies on parameter threshold judgment, such as temperature overrun alarm for anomaly analysis, although it can well identify obvious anomalies, it cannot trace the propagation path of anomalies in the process chain. For example, when a certain batch of bread has a problem of insufficient volume, this method may only detect the temperature anomaly in the baking stage, but ignore the indirect influence that the temperature and humidity changes during the fermentation process may have on the control of baking temperature, resulting in the failure to identify the root cause of insufficient gluten formation in the previous stirring process.

[0004] Moreover, the reasons for anomalies between different batches have certain similarities and trends. If the common characteristics between different production batches can be comprehensively analyzed, it will help to identify potential quality problems, better locate the source factors according to the trend changes of the overall production process parameters, facilitate the optimization of the production process, and thus improve the production quality of baked foods. Summary of the Invention

[0005] The present invention proposes a method and system for tracing and managing food production information, aiming to solve at least one of the technical problems existing in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for tracing and managing food production information, comprising:

[0008] Obtaining multi-process production record data corresponding to multiple historical production batches of target baked foods respectively, performing deviation discretization and trend discretization processing respectively, obtaining deviation characteristic data and trend characteristic data of each group of multi-process production record data, and constructing deviation trend joint characteristics of each production process for multiple groups of multi-process production record data;

[0009] Divide the production record data of multiple groups and multiple processes into normal sample data and abnormal sample data, and conduct a linkage analysis of the production process characteristics of the target baked food based on the combined characteristics of the deviation trends of the normal sample data to generate a process joint feature matrix between any two production processes;

[0010] Perform pattern clustering based on production process distance perception on the production record data of multiple groups and multiple processes according to the combined characteristics of the deviation trends to generate multiple production abnormal pattern clusters;

[0011] Conduct process abnormality analysis and backpropagation traceability on the batch to be analyzed of the target baked food according to the combined characteristics of the deviation trends and the process joint feature matrix, and generate the process abnormality detection result of the batch to be analyzed;

[0012] Perform pattern abnormality traceability on the batch to be analyzed according to multiple production abnormal pattern clusters, generate the production abnormal pattern matching result of the batch to be analyzed, and fuse it with the process abnormality detection result to generate the production abnormal traceability analysis result of the batch to be analyzed.

[0013] Preferably, conducting a linkage analysis of the production process characteristics of the target baked food to generate a process joint feature matrix between any two production processes includes:

[0014] Determine multiple local combined features of any one production process according to the combined characteristics of the deviation trends. The local combined features include the deviation discrete features and trend discrete features of the production parameters in the deviation characteristic data and trend characteristic data;

[0015] Statistically calculate the frequency parameters corresponding to the multiple local combined features of each production process, and calculate the process joint parameters between any two production processes for each joint feature pair, including determining the joint co-occurrence parameter of each joint feature pair, and determining the downstream reference frequency of the joint feature pair according to the production sequence between the two production processes associated with the joint feature pair. Denote the ratio of the joint co-occurrence parameter of the joint feature pair to the downstream reference frequency as the process joint parameter of the joint feature pair, and generate the process joint feature matrix between the two production processes according to the process joint parameters corresponding to multiple joint feature pairs.

[0016] Preferably, performing pattern clustering based on production process distance perception on the production record data of multiple groups and multiple processes according to the combined characteristics of the deviation trends to generate multiple production abnormal pattern clusters includes:

[0017] Vectorize each group of multi - process production record data according to the combined features of deviation trends, including performing feature scarcity analysis on multiple local combined features of the multi - process production record data in each production process, calculating the global scarcity parameters of each local combined feature, generating the combined weights of the local combined features based on the global scarcity parameters and frequency parameters, and constructing the combined weight feature vector of each group of multi - process production record data;

[0018] Perform pattern clustering based on production process distance perception on multiple groups of multi - process production record data according to multiple combined weight feature vectors, including introducing a production process distance perception weight for correction during the process of calculating the distance between any two groups of multi - process production record data based on the combined weight feature vectors to calculate the comprehensive distance parameter between any two groups of multi - process production record data, and clustering multiple groups of multi - process production record data using a target clustering algorithm to obtain multiple production anomaly pattern clusters.

[0019] Preferably, perform process anomaly analysis and backpropagation traceability on the batch to be analyzed of the target baked food according to the combined features of deviation trends and the process combined feature matrix, including:

[0020] Determine the target production record data of the batch to be analyzed, perform process anomaly analysis on the target production record data according to the frequency parameters corresponding to each local combined feature in the combined features of deviation trends, mark the abnormal processes of the target production record data, determine the multiple process combined feature matrices associated with the abnormal processes, and perform backpropagation traceability on the abnormal processes, calculate the backpropagation matching parameters corresponding to multiple upstream processes for the abnormal processes, and generate the process anomaly traceability analysis result of the batch to be analyzed.

[0021] Preferably, perform backpropagation traceability on the abnormal processes, and calculate the backpropagation matching parameters corresponding to multiple upstream processes for the abnormal processes, including:

[0022] Perform trend transfer analysis on multiple local combined features in the abnormal process according to the combined features of deviation trends, determine the abnormal features in the abnormal process, determine the process combined parameters between the abnormal features and multiple local combined features of the upstream process according to the process combined feature matrix between the abnormal process and the upstream process, and calculate the backpropagation matching parameters of the abnormal process with respect to the upstream process.

[0023] Preferably, use the DBSCAN clustering algorithm to cluster multiple groups of multi - process production record data according to the comprehensive distance parameter to obtain multiple production anomaly pattern clusters.

[0024] A food production information traceability management system, including:

[0025] The data association feature extraction module is used to obtain the multi-process production record data corresponding to the target baked food in multiple historical production batches, perform deviation discretization and trend discretization respectively, obtain the deviation characteristic data and trend characteristic data of each group of multi-process production record data, and construct the deviation trend association features of multiple groups of multi-process production record data for each production process;

[0026] The process association analysis module is used to divide multiple groups of multi-process production record data into normal sample data and abnormal sample data, and perform production process characteristic linkage analysis on the target baked food according to the deviation trend association features of the normal sample data to generate a process association feature matrix between any two production processes;

[0027] The production anomaly pattern clustering module is used to perform pattern clustering based on production process distance perception on multiple groups of multi-process production record data according to the deviation trend association features to generate multiple production anomaly pattern clusters;

[0028] The process anomaly detection module is used to perform process anomaly analysis and backpropagation traceability on the batch to be analyzed of the target baked food according to the deviation trend association features and the process association feature matrix, and generate a process anomaly detection result for the batch to be analyzed;

[0029] The production anomaly traceability analysis module is used to perform pattern anomaly traceability on the batch to be analyzed according to multiple production anomaly pattern clusters, generate a production anomaly pattern matching result for the batch to be analyzed, and fuse it with the process anomaly detection result to generate a production anomaly traceability analysis result for the batch to be analyzed.

[0030] The present invention has the following beneficial effects:

[0031] By performing discretization processing on the multi-process production record data, the present invention constructs the deviation trend association features and the process association feature matrix, quantifies the parameter linkage relationship between the upstream and downstream production processes, realizes implicit anomaly detection through the backpropagation calculation of the anomaly contribution degree to identify the weak collaborative offset of parameters between multiple processes; performs pattern recognition on multiple groups of multi-process production record data from the perspective of feature scarcity, and modifies and amplifies the contribution of key process differences through the process perception distance, extracts the change relationship of implicit anomalies in the entire production process from a large amount of sample data, comprehensively performs anomaly traceability detection from different angles, and combines individual anomaly conduction paths and group common characteristics to identify complex anomalies, which can well capture phenomena such as process drift and sudden anomalies in production data and improve the accuracy of anomaly traceability in baked food production. Brief Description of the Drawings

[0032] Figure 1 It is a schematic flowchart of an exemplary food production information traceability management method of the present invention.

[0033] Figure 2 This is a schematic structural diagram of an exemplary food production information traceability management system of the present invention. Detailed implementation manners

[0034] In order to enable those skilled in the art of this technology to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0035] Please refer to Figure 1 , a food production information traceability management method provided by an embodiment of the present invention specifically includes the following contents:

[0036] Step S10: Obtain multi-process production record data corresponding to multiple historical production batches of the target baked food, respectively perform deviation discretization and trend discretization processing on them to obtain deviation characteristic data and trend characteristic data of each group of multi-process production record data, and construct deviation-trend joint characteristics of each group of multi-process production record data for each production process.

[0037] It should be noted that the original data involves multi-process production record data corresponding to different production batches of a certain baked food, including production-related information involved in different production processes, such as time-series record data of production parameters such as temperature, humidity, and rotation speed in the mixing stage. Deviation discretization and trend discretization processing are performed on these data to extract characteristic information of different granularities. Deviation discretization can calculate the mean level of production parameters at different time periods in each production process and the deviation from the expected value, such as a preset reference value or the mean determined through historical data, and discretize it into deviation discretization characteristics such as higher than the normal level, lower than the normal level, and conforming to the normal level, reflecting the fluctuation degree of parameters within each process. Trend discretization can calculate the change trend of production parameters at different time periods in each production process and discretize it into trend discretization characteristics such as rising, falling, or stable trends, reflecting the stability of parameter changes in each process during production. Thus, deviation characteristic data and trend characteristic data of each group of multi-process production record data are obtained, and the deviation discretization characteristics and trend discretization characteristics corresponding to the production parameters in each process are combined to obtain the deviation-trend joint characteristics of each production process.

[0038] Step S20: Divide multiple groups of multi-process production record data into normal sample data and abnormal sample data, and perform production process characteristic linkage analysis on the target baked food according to the deviation-trend joint characteristics of the normal sample data to generate a process joint characteristic matrix between any two production processes.

[0039] It should be noted that based on the above-mentioned jointly extracted deviation trend features, the data of multiple production batches are divided into two categories, namely normal sample data and abnormal sample data, which respectively represent the differences in data changes between the normal production process and the abnormal production process. Then, for the jointly extracted deviation trend features of the normal sample data, a linkage analysis of production process characteristics is carried out, specifically used to identify the relationships between individual production processes, and a model is established by means of the combined influence of the characteristics of these processes on product quality.

[0040] In one of the implementation processes, the process of generating a process joint feature matrix through a linkage analysis of the production process characteristics of the target baked food specifically includes:

[0041] Based on the jointly extracted deviation trend features, multiple local combined features of any production process are determined. The local combined features include the deviation discrete features and trend discrete features of production parameters in the deviation characteristic data and trend characteristic data. Exemplarily, for the mixing stage, the deviation discrete features and trend discrete features of parameters such as mixing speed, material humidity, and material temperature at a specific time period are combined to obtain the local combined features corresponding to the overall multiple production parameters during this time period. For example, the characteristic levels shown by the temperature of the ingredients and the mixing speed of the equipment at the initial stage of mixing.

[0042] Then, the frequency parameters corresponding to the multiple local combined features of each production process are statistically analyzed, and based on the frequency parameters corresponding to different local combined features, the process joint parameters between any two production processes for each joint feature pair are calculated.

[0043] In this process, the joint co-occurrence parameter of each joint feature pair is first determined. Taking any two processes as an example, such as the mixing stage and the fermentation stage, a parameter correlation feature of a local combined feature corresponding to each of them, such as the parameter correlation feature of the late mixing stage and the mid-fermentation stage, constitutes a feature pair. Then, based on the frequency parameters of the local combined features, the joint co-occurrence parameter, that is, the frequency of co-occurrence, of this feature combination pair is statistically analyzed in the multi-process production record data corresponding to multiple production batches. According to the production order between the two production processes associated with the joint feature pair, the downstream reference frequency of the joint feature pair is determined, that is, the frequency parameter corresponding to the local combined feature of the downstream process among the two production processes involved in the joint feature pair. The ratio of the joint co-occurrence parameter of the joint feature pair to the downstream reference frequency is recorded as the process joint parameter of the joint feature pair, which characterizes whether the downstream process is affected by the upstream process. Through the process joint parameters of multiple joint feature pairs, the relationship strength and interaction between processes are quantified, and finally a process joint feature matrix between two production processes is constructed to reflect the association characteristics between the upstream and downstream processes.

[0044] Step S30: Based on the combined features of deviation trends, perform pattern clustering on multiple groups of multi-process production record data with production process distance perception to generate multiple production anomaly pattern clusters.

[0045] It should be noted that in the process of performing pattern clustering on multiple groups of production record data using the combined features of deviation trends, specifically, the normal samples and abnormal samples are jointly analyzed. The normal samples constitute the main distribution of the data, and the abnormal samples exist as outliers or sub-clusters deviating from the main distribution. Considering that some anomalies may be manifested as the deviation of the combination of normal parameters, such as normal temperature but abnormal humidity and rotation speed, it is necessary to identify them by comparing with normal samples to discover hidden abnormal models. At the same time, the distribution change of normal samples, such as gradual deviation, may indicate subsequent production risks. Therefore, it is observed in a unified data space to achieve the drift detection of production processes. In the process of analyzing the distance between different production batches, the production process distance perception is introduced to amplify the impact of key production processes on the whole, and at the same time, considering the timing characteristics of the process flow, the matching situation between different batches is analyzed. Finally, multiple groups of multi-process production record data are clustered to generate multiple production anomaly pattern clusters.

[0046] In one implementation process, the above-mentioned pattern clustering of multiple groups of multi-process production record data based on the combined features of deviation trends with production process distance perception to generate multiple production anomaly pattern clusters includes the following:

[0047] Perform vectorization processing on each group of multi-process production record data according to the combined features of deviation trends to construct a combined weight feature vector for each group of multi-process production record data.

[0048] Specifically, first perform feature scarcity analysis on multiple local combined features of multi-process production record data in each production process, and calculate the global scarcity parameter of each local combined feature: In the formula, represents a local combined feature in one production process, represents the total number of batches, that is, the total number of production batches involved in multiple groups of multi-process production record data, represents the number of batches in which the local combined feature appears in this production process among multiple production batches.

[0049] After obtaining the global scarcity parameter of the local combined feature through the above method, the joint weight of the local combined feature is generated based on the global scarcity parameter and the frequency parameter. Specifically, the feature frequency of the local combined feature is calculated through the frequency parameter, that is, the ratio between the frequency parameter of the local combined feature and the total frequency of multiple local combined features in this production process. Then, the product of the feature frequency of the local combined feature and the global scarcity parameter is used as the joint weight of the local combined feature. During the calculation process of the joint weight, a large number of conventional combined features are suppressed and some key combined features are enhanced. Specifically, considering that normal samples in the multi-process production record data are the main body of the data, and some hidden abnormalities and local drift scenarios of parameters are mined. Finally, the local combined features with low joint weights represent some hidden abnormal trend features in the data, and are more likely to be key indicators related to abnormal production. A low joint weight indicates that the combined feature is common and conforms to the overall normal change law in the actual production process. The joint weight feature vector constructed in this way can well highlight the hidden abnormal patterns in a large number of historical sample data, such as highlighting the abnormal parameter deviation phenomenon and abnormal change trend in the production process.

[0050] Finally, based on multiple joint weight feature vectors, pattern clustering based on production process distance perception is performed on multiple groups of multi-process production record data.

[0051] Specifically, during the process of calculating the distance between any two groups of multi-process production record data based on the joint weight feature vector, a production process distance perception weight is introduced for correction to calculate the comprehensive distance parameter between any two groups of multi-process production record data. For the production process distance perception weight: In the formula represents the production batch and the production batch the production process distance perception weight between them, represents the production batch and the production batch at the th production process, the matching parameter between the joint weight feature vectors, represents the th production process, the maximum length of the joint weight feature vector, represents the th production process, the importance weight, represents the number of production processes.

[0052] It should be noted that the production process distance perception weight is used to measure the differences between different production batches, including amplifying the contribution degree of key production processes to the whole. Specifically, it is optimized through the importance weight of the production process. The importance weight can be reasonably set based on the frequency of anomalies caused by specific production processes in historical data or combined with expert experience. The matching parameter is used to measure the matching degree of the combined features between the joint weight feature vectors. For example, due to the differences in production batches, an additional local combined feature appears in a certain production process in one batch, resulting in a difference in the number of combined features in the corresponding joint weight feature vector. Data filling is required to maintain a consistent sequence degree. The matching parameter can specifically be the number of common local combined features between two parameters. By correcting the conventional distance (such as the Euclidean distance) between two joint weight feature vectors through the production process distance perception weight, the comprehensive distance parameter between the multi-process production record data calculated can better conform to the production reality and more comprehensively represent the differences between different production batches.

[0053] Finally, according to the comprehensive distance parameter, a target clustering algorithm is adopted, such as a density-based or distance-based algorithm. In this embodiment, the DBSCAN clustering algorithm is taken as an example. Using the comprehensive distance parameter as the distance between the multi-process production record data, clustering of multiple groups of multi-process production record data is achieved through the DBSCAN clustering algorithm to obtain multiple production anomaly pattern clusters, and each cluster represents one or more similar production anomaly patterns. For example, a certain production anomaly pattern cluster corresponds to the production batches with anomalies caused by too low stirring humidity, and by suppressing common signals and amplifying hidden anomaly signals, the normal patterns with similar change trends to the anomaly pattern are fused into the cluster, which can assist in locating the root process and identifying the abnormal change pattern of the overall production process caused by a certain root anomaly point.

[0054] Step S40: According to the deviation trend joint feature and the process joint feature matrix, perform process anomaly analysis and backpropagation traceability on the batch to be analyzed of the target baked food, and generate the process anomaly detection result of the batch to be analyzed.

[0055] It should be noted that by analyzing the target production record data of the batch to be analyzed of the target baked food, including first performing process anomaly analysis according to the deviation trend joint feature to identify the deviation or change trend of the data from the normal level, determining the abnormal processes existing in the batch to be analyzed, and then analyzing its correlation with multiple upstream production processes according to the process joint feature matrix, and through backpropagation traceability, locating the root cause of the anomaly, and generating the process anomaly detection result of the batch to be analyzed.

[0056] In this process, according to the frequency parameters corresponding to each local combined feature in the deviation trend combined feature, process anomaly analysis is performed on the target production record data, and some special local combined features in the target production record data are identified, that is, the combined features with a low probability of occurrence in the normal sample data, and the corresponding production processes are marked as abnormal processes.

[0057] Then, determine multiple process combined feature matrices associated with the abnormal process, that is, the process combined feature matrices between the abnormal process and multiple upstream processes. Perform backpropagation tracing on the abnormal process according to the process combined feature matrix, including calculating the backpropagation matching parameters between the abnormal process and each upstream process, and measuring the influence degree of the upstream process on the abnormal process through the backpropagation matching parameters, so as to generate the process anomaly tracing analysis result of the batch to be analyzed.

[0058] In this process, first determine multiple local combined features in the abnormal process. Perform trend transfer analysis on each local combined feature according to the deviation trend combined feature to determine the abnormal features in the abnormal process. Specifically, count the occurrence frequency of each local combined feature in the abnormal process, and compare the occurrence frequency with the reference frequency determined by the deviation trend combined feature of the normal sample data. If it is significantly higher than the normal level, it indicates that this local combined feature is an abnormal feature, that is, there is a relatively significant anomaly compared with the historical pattern. Then calculate the backpropagation matching parameters between the abnormal process and each upstream process according to the abnormal features. Specifically, through the corresponding process combined feature matrix, count the cumulative value corresponding to the process combined parameters of the abnormal features and multiple combined features in the upstream process as the backpropagation matching parameter between the abnormal feature and one of the upstream processes, which is used to characterize the overall association degree between the abnormal feature and the upstream process. If in a certain upstream process, some local combined features frequently appear with this abnormal feature, that is, these local combined features in the upstream process will cause this abnormal feature in the abnormal process to appear with a high probability, it indicates that this upstream process is the source process of this abnormal feature, and this abnormal feature appears abnormally in the abnormal process and has a large correlation with a certain upstream process. Calculate the backpropagation matching parameters between the abnormal process and each upstream process in this way to quantitatively characterize the influence of different upstream processes on the abnormal process.

[0059] Step S50: Perform pattern anomaly tracing on the batch to be analyzed according to multiple production anomaly pattern clusters, generate the production anomaly pattern matching result of the batch to be analyzed, and fuse it with the process anomaly detection result to generate the production anomaly tracing analysis result of the batch to be analyzed.

[0060] It should be noted that, based on the target production record data of the batch to be analyzed and combined with the results of vectorizing the production record data of multiple groups and multiple processes in the foregoing steps, a combined weight feature vector corresponding to the target production record data of the batch to be analyzed is generated. Then, the combined weight feature vector corresponding to the target production record data is matched with multiple production anomaly pattern clusters, including calculating the cluster center distance between the combined weight feature vector corresponding to the target production record data and each production anomaly pattern cluster to determine the production anomaly pattern cluster with the closest distance, and obtaining the production anomaly pattern matching result of the batch to be analyzed.

[0061] In the above process, the process anomaly detection result is generated from the perspective of the upstream and downstream influence between production processes. In this process, due to weak signals, the deviation degree or change trend of some parameters may not be very significant, resulting in the possible failure to accurately lock the source process. Instead, a production process between the real source process and the abnormal process may be taken as the target, or the appearance of the abnormal process may be caused by the combined action of small anomalies in multiple processes. The production anomaly pattern clusters analyze the parameter change laws between the overall processes in the entire production process, and at the same time integrate the influence laws brought by the source processes in the abnormal samples, which can better lock the root cause process. The production anomaly pattern matching result can be fused with the process anomaly detection result, and the combined influence characteristics between independent processes can be corrected by combining historical group laws. By comprehensively considering the matching situation between each production process and the abnormal process, it is possible to more comprehensively locate and explain the quality problems in production, achieve high-precision source process positioning, and provide the associated influence relationship data of multiple processes between the source and the current process, facilitating the optimization of the production process and improving the production quality of baked foods.

[0062] Based on the same inventive concept, an embodiment of the present invention provides a food production information traceability management system. Please refer to Figure 2 , and this system includes:

[0063] A data joint feature extraction module, configured to obtain the multi-process production record data corresponding to the target baked food in multiple historical production batches respectively, perform deviation discretization and trend discretization processing respectively, obtain the deviation characteristic data and trend characteristic data of each group of multi-process production record data, and construct the deviation trend joint features of multiple groups of multi-process production record data for each production process;

[0064] A process joint analysis module, configured to divide multiple groups of multi-process production record data into normal sample data and abnormal sample data, perform production process characteristic linkage analysis on the target baked food according to the deviation trend joint features of the normal sample data, and generate a process joint feature matrix between any two production processes;

[0065] Including determining multiple local combined features of any production process according to the combined features of deviation trends, where the local combined features include the deviation discrete features and trend discrete features of production parameters in the deviation characteristic data and trend characteristic data;

[0066] Statistically calculate the frequency parameters corresponding to the multiple local combined features of each production process, and calculate the process combined parameters between any two production processes for each combined feature pair, including determining the combined co-occurrence parameter of each combined feature pair, and according to the production sequence between the two production processes associated with the combined feature pair, determining the downstream reference frequency of the combined feature pair, and recording the ratio of the combined co-occurrence parameter of the combined feature pair to the downstream reference frequency as the process combined parameter of the combined feature pair, and generating a process combined feature matrix between two production processes according to the process combined parameters corresponding to multiple combined feature pairs.

[0067] The production anomaly pattern clustering module is used to perform pattern clustering based on production process distance perception on multiple groups of multi-process production record data according to the combined features of deviation trends, and generate multiple production anomaly pattern clusters;

[0068] Including vectorizing each group of multi-process production record data according to the combined features of deviation trends, including performing feature scarcity analysis on the multiple local combined features of the multi-process production record data in each production process, calculating the global scarcity parameter of each local combined feature, generating the combined weight of the local combined feature according to the global scarcity parameter and the frequency parameter, and constructing the combined weight feature vector of each group of multi-process production record data;

[0069] Performing pattern clustering based on production process distance perception on multiple groups of multi-process production record data according to multiple combined weight feature vectors, including introducing a production process distance perception weight for correction during the process of calculating the distance between any two groups of multi-process production record data according to the combined weight feature vector to calculate the comprehensive distance parameter between any two groups of multi-process production record data, and clustering the multiple groups of multi-process production record data using the target clustering algorithm to obtain multiple production anomaly pattern clusters.

[0070] The process anomaly detection module is used to perform process anomaly analysis and backpropagation traceability on the batch to be analyzed of the target baked food according to the combined features of deviation trends and the process combined feature matrix, and generate the process anomaly detection result of the batch to be analyzed;

[0071] The production anomaly traceability analysis module is used to perform pattern anomaly traceability on the batch to be analyzed according to multiple production anomaly pattern clusters, generate the production anomaly pattern matching result of the batch to be analyzed, and fuse it with the process anomaly detection result to generate the production anomaly traceability analysis result of the batch to be analyzed.

[0072] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The parts not described in detail in this specification belong to the prior art well-known to those skilled in the art.

Claims

1. A food production information traceability management method, characterized in that, Including: Obtain the multi-process production record data corresponding to multiple historical production batches of the target baked food respectively, perform deviation discretization and trend discretization processing respectively, obtain the deviation characteristic data and trend characteristic data of each group of multi-process production record data, and construct the deviation trend joint characteristics of each group of multi-process production record data for each production process; Divide multiple groups of multi-process production record data into normal sample data and abnormal sample data. According to the deviation trend joint characteristics of the normal sample data, perform production process characteristic linkage analysis on the target baked food, and generate a process joint characteristic matrix between any two production processes, including determining multiple local combination characteristics of any one production process according to the deviation trend joint characteristics. The local combination characteristics include the deviation discretization characteristics and trend discretization characteristics of production parameters in the deviation characteristic data and trend characteristic data; Statistically calculate the frequency parameters corresponding to multiple local combination characteristics of each production process, and calculate the process joint parameters between any two production processes for each joint feature pair, including determining the joint co-occurrence parameter of each joint feature pair, and according to the production sequence between the two production processes associated with the joint feature pair, determining the downstream reference frequency of the joint feature pair. Denote the ratio of the joint co-occurrence parameter of the joint feature pair to the downstream reference frequency as the process joint parameter of the joint feature pair, and generate a process joint characteristic matrix between two production processes according to the process joint parameters corresponding to multiple joint feature pairs; Perform pattern clustering based on production process distance perception on multiple groups of multi-process production record data according to the deviation trend joint characteristics, and generate multiple production anomaly pattern clusters, including vectorizing each group of multi-process production record data according to the deviation trend joint characteristics, including performing feature scarcity analysis on multiple local combination characteristics of the multi-process production record data in each production process, calculating the global scarcity parameter of each local combination characteristic, generating the joint weight of the local combination characteristic according to the global scarcity parameter and the frequency parameter, and constructing the joint weight feature vector of each group of multi-process production record data; Perform pattern clustering based on production process distance perception on multiple groups of multi-process production record data according to multiple joint weight feature vectors, including introducing a production process distance perception weight for correction during the process of calculating the distance between any two groups of multi-process production record data according to the joint weight feature vectors to calculate the comprehensive distance parameter between any two groups of multi-process production record data, and clustering multiple groups of multi-process production record data using the target clustering algorithm according to the comprehensive distance parameter to obtain multiple production anomaly pattern clusters; Perform process anomaly analysis and backpropagation traceability on the batch to be analyzed of the target baked food according to the deviation trend joint characteristics and the process joint characteristic matrix, and generate the process anomaly detection result of the batch to be analyzed; Perform pattern anomaly traceability on the batch to be analyzed according to multiple production anomaly pattern clusters, generate the production anomaly pattern matching result of the batch to be analyzed, and fuse it with the process anomaly detection result to generate the production anomaly traceability analysis result of the batch to be analyzed.

2. The food production information traceability management method according to claim 1, characterized in that, Based on the deviation trend combined features and the process combined feature matrix, perform process anomaly analysis and backpropagation traceability on the batch to be analyzed of the target baked food, including: Determine the target production record data of the batch to be analyzed. According to the frequency parameters corresponding to each local combined feature in the deviation trend combined features, perform process anomaly analysis on the target production record data, mark the abnormal processes in the target production record data, determine the multiple process combined feature matrices associated with the abnormal processes, and perform backpropagation traceability on the abnormal processes. Calculate the backpropagation matching parameters corresponding to multiple upstream processes for the abnormal processes, and generate the process anomaly traceability analysis result of the batch to be analyzed.

3. The food production information traceability management method according to claim 2, characterized in that, Perform backpropagation traceability on the abnormal processes, and calculate the backpropagation matching parameters corresponding to multiple upstream processes for the abnormal processes, including: Perform trend transfer analysis on multiple local combined features in the abnormal processes according to the deviation trend combined features, determine the abnormal features in the abnormal processes, determine the process combined parameters between the abnormal features and multiple local combined features of the upstream processes according to the process combined feature matrix between the abnormal processes and the upstream processes, and calculate the backpropagation matching parameters of the abnormal processes with respect to the upstream processes.

4. A food production information traceability management method according to claim 1, characterized in that, Use the DBSCAN clustering algorithm to cluster multiple sets of multi-process production record data according to the comprehensive distance parameter to obtain multiple production anomaly pattern clusters.

5. A food production information traceability management system, characterized in that, The system is used to implement a food production information traceability management method described in any one of the above claims 1-4, including: A data combined feature extraction module, configured to obtain multi-process production record data corresponding to multiple historical production batches of the target baked food, perform deviation discretization and trend discretization processing respectively, obtain the deviation characteristic data and trend characteristic data of each set of multi-process production record data, and construct deviation trend combined features of multiple sets of multi-process production record data for each production process; A process combined analysis module, configured to divide multiple sets of multi-process production record data into normal sample data and abnormal sample data, perform production process characteristic linkage analysis on the target baked food according to the deviation trend combined features of the normal sample data, and generate a process combined feature matrix between any two production processes; A production anomaly pattern clustering module, configured to perform pattern clustering based on production process distance perception on multiple sets of multi-process production record data according to the deviation trend combined features, and generate multiple production anomaly pattern clusters; A process anomaly detection module, configured to perform process anomaly analysis and backpropagation traceability on the batch to be analyzed of the target baked food according to the deviation trend combined features and the process combined feature matrix, and generate a process anomaly detection result of the batch to be analyzed; A production anomaly traceability analysis module, configured to perform pattern anomaly traceability on the batch to be analyzed according to multiple production anomaly pattern clusters, generate a production anomaly pattern matching result of the batch to be analyzed, and fuse it with the process anomaly detection result to generate a production anomaly traceability analysis result of the batch to be analyzed.

Images