Monitoring system based on acid discharging process after mutton slitting

Abstract

The invention relates to the technical field of mutton processing monitoring, and discloses a monitoring system based on an acid discharging process after mutton slitting. The system comprises a data acquisition module, a feature extraction module, a distribution generation module, a state analysis module and a control execution module. The data acquisition module acquires temperature, humidity, gas concentration and slitting area position data in an acid discharge environment in real time, and extracts an initial space-time sequence; a feature extraction module extracts coupling features from the initial space-time sequence; the distribution generation module generates partitions with different fine degrees according to the resolutions of the monitoring data at different positions to obtain multi-resolution feature distribution; the state analysis module analyzes the time sequence of the multi-resolution characteristic distribution to obtain an acid discharge state spectrum; and the control execution module identifies the abnormal indexes of the key nodes based on the acid discharge state spectrum and generates a control instruction to adjust the acid discharge environment parameters. The system realizes multi-parameter synchronous monitoring and accurate control in the acid discharge process, and improves the stability of the acid discharge quality.

Description

Technical Field

[0001] This invention relates to the field of mutton processing monitoring technology, specifically a monitoring system based on the aging process of mutton after cutting. Background Technology

[0002] In the mutton processing industry, the aging process after slicing directly impacts mutton quality. The stability and precise control of environmental parameters during aging are crucial for ensuring the mutton's taste, flavor, and safety. Currently, mutton processing enterprises mostly employ traditional manual inspections or single-parameter monitoring methods in the aging process. These methods have significant limitations. Manual inspections rely on the experience and judgment of staff, which is not only time-consuming and labor-intensive but also prone to subjective biases, making it difficult to monitor the dynamic changes in the aging environment in real time. Single-parameter monitoring equipment typically only collects data on one indicator, such as temperature or humidity, failing to simultaneously monitor multiple key parameters in the aging environment or capture the correlations between different parameters. As consumers' demands for mutton quality continue to rise, traditional monitoring methods are no longer sufficient to meet the needs of modern processing and production. In actual production, parameter distribution varies across different areas within the aging environment. For example, temperature and humidity may fluctuate significantly between areas near the aging equipment's air outlet and corner areas. Using a uniform monitoring resolution to cover the entire aging space would result in insufficient accuracy in monitoring parameters in some critical areas, while data in non-critical areas would be excessively redundant. This would affect monitoring efficiency and fail to provide a reliable basis for accurately assessing the aging status. Furthermore, most existing monitoring systems can only collect and store data, lacking the ability to perform in-depth data analysis. They cannot generate a comprehensive assessment of the aging status based on the monitoring data. When parameter anomalies occur, it is difficult to quickly locate the cause and adjust the aging environment parameters in a timely manner, potentially leading to a decline in mutton aging quality and even food safety risks. These problems hinder the intelligent and refined development of the mutton processing industry, necessitating a aging process monitoring system capable of simultaneous multi-parameter monitoring, multi-resolution analysis, and precise control. Summary of the Invention

[0003] The purpose of this invention is to provide a monitoring system based on the acid removal process after mutton is cut, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides a monitoring system based on the aging process of mutton after slicing, the system comprising: The data acquisition module is used to acquire monitoring data in the acid-removing environment after mutton is cut in real time. The monitoring data includes temperature data, humidity data, gas concentration data, and cutting area location data, and the module extracts the initial spatiotemporal sequence from the monitoring data. The feature extraction module is used to extract coupling features from the initial spatiotemporal sequence; The distribution generation module is used to generate partitions of different levels of detail based on the coupling features and the resolution of monitoring data at different locations in the aging environment after mutton is cut, and to obtain a multi-resolution feature distribution based on the partition generation results; the partitions corresponding to the high-resolution regions are more detailed than those corresponding to the low-resolution regions. The state analysis module is used to analyze the time series of the multi-resolution feature distribution to obtain the acid excretion state spectrum; The control execution module is used to identify abnormal indicators of key nodes based on the acid excretion state spectrum, and generate control commands according to the abnormal indicators to adjust the acid excretion environment parameters.

[0005] Preferably, the data acquisition module acquires real-time monitoring data of the aging environment after the mutton is cut, including: The temperature data, humidity data, gas concentration data, and segmentation area location data are standardized to obtain standardized monitoring data. The initial spatiotemporal sequence is extracted from the standardized monitoring data using a feature extraction unit.

[0006] Preferably, the feature extraction module extracts coupling features from the initial spatiotemporal sequence, including: A preliminary feature set, coupled with temperature data, humidity data, and gas concentration data, is extracted from the initial spatiotemporal sequence; When the feature dimension of the preliminary feature set is greater than a preset dimension threshold, the preliminary feature set is subjected to dimensionality reduction processing to obtain dimensionality-reduced feature data; The time series of the dimensionality-reduced feature data is analyzed, and the dynamic characteristics of the interaction between temperature data, humidity data, and gas concentration data are enhanced to obtain the coupling features.

[0007] Preferably, the distribution generation module, based on the coupling features, generates partitions of varying fineness according to the resolution of monitoring data from different locations in the aging environment after mutton slicing, and obtains a multi-resolution feature distribution based on the partition generation results, including: The aging environment of the mutton after it is cut is divided into multiple sub-regions; Construct multiple processing queues, each corresponding to a sub-region; Based on the resolution of the monitoring data for each sub-region, determine the unit size, number of partitions, and partition density for each sub-region; The multiple processing queues are controlled to generate target partitions for each sub-region according to the unit size, number of partitions, and partition density of each sub-region; Based on the coupling features and the target partitions of each sub-region, the multi-resolution feature distribution is obtained.

[0008] Preferably, the distribution generation module controls the multiple processing queues to generate target partitions for each sub-region according to the unit size, number of partitions, and partition density of each sub-region, including: The multiple processing queues are controlled to generate preliminary partition surface information according to the unit size, number of partitions, and partition density of each sub-region; The multiple processing queues are controlled to generate three-dimensional partitions of each sub-region based on the partition surface information and obtain partition volume information; Based on the partition information, obtain the vertices and volume units of each solid partition; and number the vertices and volume units of each solid partition to obtain the unique identifier of each vertex and the unique identifier of each volume unit. The system controls the synchronous communication between the multiple processing queues to map the volume units and unique identifiers between adjacent sub-regions to obtain mapping information. The three-dimensional partitioning of each sub-region is optimized based on the mapping information to obtain the target partitioning of each sub-region.

[0009] Preferably, the state analysis module analyzes the time series of the multi-resolution feature distribution to obtain the acid excretion state spectrum, including: Dynamic characteristic analysis is performed on the time series of the multi-resolution feature distribution to extract key change points; Based on the key change points, the acid removal state spectrum is constructed to represent the dynamic load of the acid removal process.

[0010] Preferably, the control execution module identifies abnormal indicators at key nodes based on the acid discharge state spectrum, including: Determine the state feature vector of the key node; Based on the state feature vector, calculate the state dimension entropy of each key node; Based on the entropy of all state dimensions and the preset process link graph, the core state feature sequence is selected. The abnormal indicators are identified based on the core state feature sequence.

[0011] Preferably, the control execution module selects core state feature sequences based on the entropy of all state dimensions and a preset process link graph, including: For each state-related feature in the state feature vector, determine the state dimension entropy of the dimension in which the state-related feature is located. Extract all feature linkage degrees corresponding to the state-related features from the preset process link graph; The state core entropy of the state-related feature is determined by the state dimension entropy of the dimension in which the state-related feature is located and the corresponding feature linkage degree. All core state features are selected based on the state core entropy of each state-related feature. The core state feature sequence is constructed based on all core state features.

[0012] Preferably, the control execution module generates control instructions based on the abnormal indicators to adjust the acid discharge environment parameters, including: The abnormal indicators are reconstructed to obtain target control data for environmental control and abnormal indication data for dynamic adjustment. Based on the target control data and combined with the preset acid removal strategy, an initial control scheme is generated; Based on the anomaly indication data, anomalies in the real-time monitoring data are identified, and the initial control scheme is dynamically adjusted according to the anomalies to generate an updated control scheme. The control execution module executes the updated control scheme to adjust the temperature data, humidity data, or gas concentration data.

[0013] Preferably, the control execution module reconstructs the abnormal indicators to obtain target control data for environmental control and abnormal indication data for dynamic adjustment, including: The abnormal indicators are converted into a control data matrix; The control data matrix is ​​decomposed to obtain a reconstructed matrix and an indicator matrix; The reconstruction matrix represents the target control data structure after interpolation and denoising; The indicator matrix is ​​used to identify abnormal data points; The reconstructed matrix is ​​used as the target control data, and the indication matrix is ​​used as the anomaly indication data.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This monitoring system, based on the aging process of mutton after slicing, acquires real-time monitoring data on temperature, humidity, gas concentration, and the location of the slicing area through a data acquisition module. It also extracts the initial spatiotemporal sequence, breaking through the limitations of traditional single-parameter monitoring. This system achieves comprehensive and synchronous acquisition of key parameters of the aging environment, providing a more complete reflection of its dynamic changes and avoiding misjudgments of the aging status due to incomplete parameter monitoring. Compared to traditional manual inspection methods, this system does not rely on human experience and can automatically complete data acquisition and preliminary processing, reducing errors caused by human factors. It also significantly improves the efficiency and accuracy of data acquisition, providing a reliable data foundation for subsequent aging status analysis. The feature extraction module extracts coupled features from the initial spatiotemporal sequence, capturing the correlation between different monitoring parameters, such as the combined impact of temperature and humidity changes on gas concentration. This exploration of multi-parameter coupling relationships allows for a deeper analysis of the acid discharge environment, moving beyond independent judgments of single parameters and helping to more accurately grasp the overall trend of the acid discharge process. The distribution generation module generates partitions of varying fineness based on the resolution of monitoring data from different locations, forming a multi-resolution feature distribution. This ensures monitoring accuracy in high-resolution areas, precisely capturing subtle parameter changes in key regions, while avoiding data redundancy in non-critical areas, achieving a rational allocation of monitoring resources and improving data processing efficiency. Compared to traditional uniform-resolution monitoring, this multi-resolution partitioning method better meets the actual monitoring needs of different areas within the acid discharge environment, enabling refined monitoring of key areas within the acid discharge space and providing more targeted data support for subsequent state analysis. The status analysis module analyzes the time series of multi-resolution feature distributions to obtain the aging status spectrum, transforming complex monitoring data into intuitive and comprehensive aging status assessment results. Staff can clearly understand the progress and current status of the aging process through the aging status spectrum, eliminating the need for tedious analysis of massive amounts of raw data. The control execution module identifies abnormal indicators at key nodes based on the aging status spectrum and generates control commands to adjust aging environment parameters, achieving automated control of the aging process. When abnormal parameters occur in the aging environment, the system can quickly locate the abnormal node and its cause, and promptly issue control commands to adjust relevant equipment parameters. For example, when the temperature in a certain area exceeds the normal range, the system automatically adjusts the temperature control equipment in that area to prevent the continuous impact of parameter abnormalities on the aging quality of the mutton. This integrated process from data acquisition and analysis to control execution forms a closed-loop management of the aging process, effectively reducing the impact of parameter abnormalities on the aging effect and ensuring the stability of the mutton's aging quality. Meanwhile, the application of this system can promote the intelligent and refined development of the mutton processing and aging process, optimize the production process, reduce labor and management costs in the production process, improve the production efficiency and market competitiveness of enterprises, and better meet consumers' demand for high-quality mutton products. Attached Figure Description

[0015] Figure 1 This is a timing diagram of the monitoring system for the aging process of mutton after slicing, as described in this invention. Figure 2 This is a flowchart illustrating the working principle of the data acquisition module. Figure 3 A flowchart illustrating the working principle of the distributed generation module; Figure 4 This is a flowchart illustrating the working principle of the status analysis module. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1 This invention provides a monitoring system for the aging process of mutton after cutting, the system comprising: By integrating multiple modules such as data acquisition, feature extraction, distribution generation, state analysis, and control execution, the system achieves real-time monitoring and intelligent regulation of the aging environment after mutton is cut. The system acquires multi-source monitoring data, including temperature, humidity, gas concentration, and the location of the cutting area, to construct an initial spatiotemporal sequence and extract coupling features. Based on the resolution differences of monitoring data from different locations, it generates multi-resolution feature distributions with varying degrees of detail, and then analyzes the time series to obtain the aging state spectrum. Finally, the system identifies abnormal indicators at key nodes and generates control commands to dynamically adjust the aging environment parameters, ensuring the optimization and stability of the mutton aging process.

[0018] Example 1: See Figure 2 The data acquisition module acquires real-time temperature, humidity, gas concentration, and segmentation area location data through a sensor network deployed in the acid discharge environment. This data undergoes standardization to eliminate dimensional differences and noise interference, resulting in standardized monitoring data. The standardization process uses the Z-score method, ensuring that each data dimension has a mean of zero and a variance of one. The feature extraction unit extracts an initial spatiotemporal sequence from the standardized monitoring data. This sequence includes timestamps, spatial coordinates, and corresponding monitoring values, forming a structured dataset.

[0019] The feature extraction module extracts coupled features from the initial spatiotemporal sequence. First, it extracts a preliminary feature set of coupled temperature, humidity, and gas concentration data from the sequence. This feature set includes the original data and their interaction terms, such as the product of temperature and humidity, and the rate of change of gas concentration over time. When the feature dimension of the preliminary feature set exceeds a preset dimensionality threshold, principal component analysis is used for dimensionality reduction, resulting in dimensionality-reduced feature data that retains the components contributing the main variance. Subsequently, the time series of the dimensionality-reduced feature data is analyzed, and a dynamic time warping algorithm is applied to enhance the dynamic characteristics of the interaction between temperature, humidity, and gas concentration data, ultimately yielding coupled features that reflect the nonlinear relationships and time-varying characteristics between environmental parameters.

[0020] The distribution generation module, based on coupling characteristics, generates partitions of varying fineness according to the resolution of monitoring data from different locations in the aging environment after mutton slicing. First, the aging environment is divided into multiple sub-regions, based on physical spatial layout and sensor distribution density. Multiple processing queues are constructed, each corresponding to a sub-region, employing a parallel computing architecture for efficient processing. Based on the resolution of the monitoring data in each sub-region, the unit size, number of partitions, and partition density of each sub-region are determined; regions with higher resolution have smaller unit sizes, more partitions, and higher partition densities, and vice versa.

[0021] Multiple processing queues are controlled to initially generate partition surface information based on the unit size, number of partitions, and partition density of each sub-region. This information includes the partition boundaries and coordinates in a two-dimensional plane. Subsequently, the processing queues generate 3D partitions for each sub-region based on the partition surface information and obtain partition volume information, which includes volume elements and their attributes in three-dimensional space. Based on the partition volume information, the vertices and volume elements of each 3D partition are obtained and assigned unique identifiers to ensure traceability. Synchronous communication is established between the multiple processing queues, and a message passing interface is used to map volume elements and unique identifiers between adjacent sub-regions, obtaining mapping information. The 3D partitions of each sub-region are optimized based on the mapping information to eliminate boundary inconsistencies and overlaps, ultimately yielding the target partitions for each sub-region. Based on coupling features and target partitions, a multi-resolution feature distribution is generated, which presents the spatial variations of feature values ​​at different resolutions in a grid format.

[0022] The state analysis module analyzes the time series of multi-resolution feature distributions to obtain the acid excretion state spectrum. First, dynamic characteristic analysis is performed on the time series, applying a change point detection algorithm and extracting key change points based on Bayesian inference. These change points correspond to significant transitions or trend reversals in environmental parameters. Based on these key change points, an acid excretion state spectrum is constructed. This spectrum represents the dynamic load of the acid excretion process in matrix form, with rows corresponding to time points, columns corresponding to spatial partitions, and element values ​​being normalized load indices of eigenvalues.

[0023] The control execution module identifies abnormal indicators of key nodes based on the acid discharge state spectrum. First, it determines the state feature vector of each key node, which consists of the load value and its temporal derivative of the corresponding node in the state spectrum. Based on the state feature vector, it calculates the state dimension entropy of each key node and uses the Shannon entropy formula to assess the uncertainty of each dimension. Based on all state dimension entropies and a predefined process link graph, it selects core state feature sequences. The process link graph is a predefined graph structure where nodes represent features and edges represent causal relationships between features. Based on the core state feature sequences, it identifies abnormal indicators, which are feature values ​​or entropy values ​​exceeding preset thresholds.

[0024] The control execution module selects core state feature sequences based on the entropy of all state dimensions and a preset process link graph. For each state-related feature in the state feature vector, the state dimension entropy of the dimension in which the state-related feature is located is determined. This entropy value reflects the degree of fluctuation of the feature over time. All feature link degrees corresponding to the state-related features are extracted from the preset process link graph. The link degree is calculated based on the sum of the number of adjacent nodes and edge weights of the feature in the graph. The state core entropy of the state-related feature is determined by the state dimension entropy of the dimension in which the state-related feature is located and the corresponding link degrees of all features. The state core entropy is the weighted product of the entropy value and the link degree. All core state features are selected based on the state core entropy of each state-related feature, with the selection criterion being that the state core entropy is higher than a preset threshold. A core state feature sequence is constructed based on all core state features. This sequence is arranged in chronological order and used for subsequent anomaly indicator identification.

[0025] The control execution module generates control commands based on abnormal indicators to adjust the acid discharge environment parameters. First, the abnormal indicators are reconstructed, converting them into a control data matrix. Rows in the matrix represent time points, and columns represent different control parameters. The control data matrix is ​​then decomposed using singular value decomposition to obtain a reconstructed matrix and an indicator matrix. The reconstructed matrix represents the target control data structure after interpolation and denoising, while the indicator matrix identifies abnormal data points and their locations. The reconstructed matrix serves as the target control data, and the indicator matrix serves as the abnormality indication data.

[0026] Based on the target control data and a preset acid removal strategy, an initial control scheme is generated. This preset strategy includes target ranges for temperature, humidity, and gas concentration, along with control logic. Based on anomaly indication data, anomalies in the real-time monitoring data are identified; these anomalies are identified by non-zero elements in the indication matrix. The initial control scheme is dynamically adjusted according to the anomalies, generating an updated control scheme. Adjustment methods include PID control algorithms or fuzzy logic control to compensate for abnormal deviations. The control execution module executes the updated control scheme, adjusting temperature, humidity, or gas concentration data via actuators to achieve closed-loop control of the acid removal environment.

[0027] Example 2: See Figure 3The distributed generation module, based on coupling characteristics, generates partitions of varying fineness according to the resolution of monitoring data from different locations in the aging environment after mutton cutting. In practice, the entire aging environment space is first divided into regions based on physical structure characteristics and sensor deployment density. For example, in an aging workshop measuring 30 meters long, 20 meters wide, and 4 meters high, the space is divided into six sub-regions based on the conveyor belt direction and suspension area distribution: the inlet pre-treatment area, the main aging areas A / B / C, the quality inspection area, and the outlet temporary storage area. Each sub-region is equipped with an independent data processing queue, which runs on distributed computing nodes, forming a parallel processing architecture.

[0028] The monitoring data resolution varies significantly across sub-regions. The main acid removal zone A employs a high-density sensor array, with five temperature and humidity sensors and three gas concentration probes per square meter, sampling once every 10 seconds, classifying it as a high-resolution area. In contrast, the outlet storage zone has only one basic sensor per 5 square meters, sampling once per minute, classifying it as a low-resolution area. Based on these resolution differences, the system dynamically determines the grid parameters for each sub-region: the main acid removal zone A uses a 0.2m × 0.2m × 0.3m unit size, dividing it into approximately 15,000 basic units; the outlet storage zone uses a 1m × 1m × 1m unit size, generating only approximately 240 basic units. The density also exhibits a gradient, with high-resolution areas reaching a unit density of 125 units per cubic meter, while low-resolution areas have only 1 unit per cubic meter.

[0029] Each processing queue performs partition generation according to the assigned unit size parameters. The processing queue for the main acid removal zone A first generates a two-dimensional grid in the XY plane, with each grid point recording its spatial coordinates and associated with the average historical monitoring data for that location. Then, it is layered and expanded along the Z-axis to form a three-dimensional grid with a height of 0.3 meters, ultimately generating partition volume information containing location coordinates, volume parameters, and environmental characteristic values. Each volume unit is assigned a three-dimensional coordinate code; for example, a unit in zone A is identified as "A-35-28-04," indicating that the unit is located in column 35, row 28, and height 4 of zone A. All vertices use a globally unique encoding rule; for example, vertex "AV-1289" corresponds to vertex number 1289 in zone A.

[0030] After adjacent sub-regions complete their respective partitioning, the system initiates a boundary coordination mechanism. The main acid removal zone A and main acid removal zone B share a 0.5-meter overlap. The processing queues of the two zones exchange boundary unit information via a message interface. For example, boundary unit "AB-87" in zone A and unit "BA-12" in zone B establish a mapping relationship. The system detects a 15% volume overlap between the two units and immediately initiates an optimization algorithm: First, it calculates the environmental feature gradient of the overlapping area, re-divides the boundary based on the gradient change trend, and assigns the overlapping portion to the region with more significant feature changes; then, it adjusts the unit size so that the size difference between units on both sides of the boundary does not exceed 20%; finally, it updates the unit number to ensure that the boundary unit receives dual identification across regions.

[0031] After boundary optimization, the system maps coupled feature values ​​to each partition unit. High-resolution areas employ a fine-mapping mode, with each 0.2-meter square unit independently carrying a feature value. Low-resolution areas use a feature value regional averaging method, integrating monitoring data within a 1-cubic-meter unit into a single feature value. The resulting multi-resolution feature distribution exhibits spatial gradation: in areas with densely hung mutton, the feature distribution map can display parameter fluctuations in the 0.1-cubic-meter microenvironment surrounding a single piece of mutton; while in the passageway area, the feature distribution only retains the macroscopic state within a 5-cubic-meter range. This multi-scale representation satisfies the fine-grained monitoring needs of key areas while reducing the computational load on non-key areas.

[0032] The entire partitioning process adopts a pipeline operation mode. When the data acquisition module updates the monitoring data, the processing queue only performs local recalculation on the affected areas. For example, when a sensor detects abnormal temperature fluctuations, the system automatically locates the sub-region to which the sensor belongs, triggering only the real-time update of the corresponding processing queue, while maintaining the stability of the partitioning structure of other regions. This dynamic update mechanism enables the multi-resolution feature distribution to reflect environmental changes in a timely manner, while maintaining the overall operating efficiency of the system. The final generated feature distribution data is stored in the form of a three-dimensional mesh database, with each mesh cell containing a complete set of attributes such as spatial coordinates, resolution level, feature value, and timestamp.

[0033] Example 3: See Figure 4The state analysis module receives multi-resolution feature distribution data from the distribution generation module. This data is organized in time series form, with each time point corresponding to a set of spatial partition feature values. In specific implementation, this module first performs dynamic characteristic analysis on the time series. For the high-resolution partition data (10-second time intervals) of the main acid-removing zone A, a sliding window mechanism is used to process the feature value changes over 72 consecutive hours. The window size is 360 time points (corresponding to 1 hour), and 60 points are slid in each window (10 minutes). Within each window, the mean, variance, and first difference of the feature values ​​are calculated to form a dynamic feature vector. The change point detection algorithm is based on the Mahalanobis distance change rate of the feature vector. When the distance change rate of three consecutive windows exceeds three standard deviations of the historical baseline, it is marked as a potential key change point. Subsequently, a backward verification mechanism is used to check the similarity of the feature distribution of the five windows before and after the point. If the similarity is lower than a preset threshold, it is confirmed as a key change point. For example, during the 28th hour of the acid removal process, a temperature characteristic value of a 0.2 cubic meter area around a certain suspension unit was detected to jump by 0.8 standard deviations within 120 seconds, while the humidity characteristic showed an inverse trend. This point was identified as a key change point reflecting the denaturation process of mutton protein.

[0034] When constructing the acid discharge state spectrum based on key change points, a three-dimensional tensor data structure is used. The first dimension of the tensor represents the time axis, which is divided into several intervals with key change points as nodes. For example, 17 key change points are identified within a 72-hour period, forming 18 time intervals. The second dimension represents spatial partitioning, preserving multi-resolution characteristics: each 0.2-meter unit in high-resolution areas is an independent partition, while low-resolution areas are merged into 5-cubic-meter unit groups. The third dimension carries the characteristic load index, which is obtained through composite calculation: first, the range of temperature, humidity, and gas concentration characteristic values ​​in each partition is normalized; then, the integral mean of each characteristic within the time interval is calculated; finally, the load value in the 0-1 interval is generated through weighted fusion. The state spectrum is presented as a dynamic load matrix. For example, within the time interval [28.3h, 29.1h], the unit "A-35-28-04" in the main acid discharge zone A records a load value of 0.78, while the load value of the unit group "EX-Z01" in the outlet temporary storage zone at the same time is only 0.21. This method of expression can capture changes in the local microenvironment while maintaining the visualization of the overall process.

[0035] When the control execution module analyzes the acid discharge state spectrum, it first locates key nodes. Spatially, it selects the top 10% of partitions with the largest load variance as key nodes, and temporally, it selects the center point of the time interval containing the key change points. A state feature vector is constructed for each key node, containing five dimensions: current load value, moving average of load values ​​from the previous three time intervals, and load change acceleration. For each dimension of the vector, the state dimension entropy is calculated using the following formula:

[0036] in: Represents the entropy value of a specific dimension. This represents the number of discrete state intervals (default value is 10). Is the feature value of this dimension falling into the first... The probability of an interval, The base of the logarithm is 2. For example, in 200 consecutive observation points, the temperature feature dimension of a certain node shows a distribution of 7 dense intervals, with an entropy value of 2.81; while the humidity feature dimension is evenly distributed, with an entropy value of 3.32. High entropy values ​​indicate that the feature fluctuates violently, while low entropy values ​​indicate that the state is stable.

[0037] The preset process link graph is stored in a directed graph structure, containing 120 feature nodes and 356 weighted edges. The "Temperature-Humidity Interaction Feature" node in the graph has high centrality, connecting 18 downstream nodes. When analyzing a key node, the link degree of all features in its state feature vector is extracted and analyzed within the process link graph. The link degree calculation includes three factors: the number of directly adjacent nodes, the sum of edge weights, and the weighted sum of the inverse distances to indirectly connected nodes. For example, the "Carbon Dioxide Concentration Change Rate" feature directly connects 5 nodes in the graph, has a total edge weight of 8.7, and indirectly affects 12 second-degree nodes, resulting in a link degree of 15.3.

[0038] The state core entropy is obtained through a non-linear combination of entropy value and connectivity: ,in For connectivity. When filtering core state features, set a dynamic threshold. ( The mean of the core entropy of all current features. (Standard deviation). The core entropy of the "temperature-gas concentration covariance feature" at a certain node reaches 18.7 ( ), exceeding the threshold The selected core sequence is then constructed. The final core state feature sequence is arranged in chronological order. When three consecutive time points in the sequence contain the same type of core feature, an anomaly detection mechanism is triggered.

[0039] Anomaly identification employs a dual criterion: first, the core feature value exceeds the historical fluctuation range (more than three standard deviations); second, the feature combination violates the constraints of the process link diagram. For example, when the "surface humidity gradient" feature value reaches 0.92 (baseline 0.65±0.12), and simultaneously deviates in the opposite direction from the associated "lactic acid evaporation rate" feature, the system determines that there is an anomaly in acid discharge obstruction. All anomalies are recorded as quadruplets (timestamp, spatial coordinates, feature type, deviation degree) and transmitted to the control command generation unit. This implementation method establishes a mapping relationship between process state and anomalies by combining spatiotemporal analysis, while preserving multi-resolution characteristics.

[0040] Example 4: The state analysis module analyzes the time series of multi-resolution feature distributions to obtain the acid removal state spectrum. This state spectrum represents the dynamic load of the acid removal process in matrix form, with rows corresponding to time points, columns corresponding to spatial partitions, and element values ​​being normalized load indices of feature values. The control execution module identifies abnormal indicators of key nodes based on the acid removal state spectrum. First, the state feature vector of the key nodes is determined, which consists of the load value of the corresponding node in the state spectrum and its temporal derivative. Based on the state feature vector, the state dimension entropy of each key node is calculated, and the Shannon entropy formula is used to evaluate the uncertainty of each dimension feature. In the specific implementation process, the system selects five representative key monitoring nodes in the acid removal workshop for analysis. These nodes are located at the four corners and the center of the main acid removal area A, with each node corresponding to a 0.2 cubic meter monitoring unit. The state feature vector contains six dimensions: temperature load value, humidity load value, carbon dioxide concentration load value, and the first-order temporal derivative of these three parameters. The system collects data continuously for 240 minutes, recording the feature values ​​of each dimension once per minute.

[0041] For each state-related feature, the system calculates its state dimension entropy. Taking temperature as an example, the temperature load value is divided into 10 equally wide intervals, and the distribution probability of 240 data points in each interval is statistically analyzed. The entropy value is then calculated using the Shannon entropy formula. Humidity and gas concentration features are calculated using the same method. The entropy value calculation for time-series derivative features uses a dynamic interval division method, adjusting the interval boundaries according to the actual distribution range of the derivative values.

[0042] The preset process link graph contains 28 feature nodes and 67 weighted edges. Nodes in the graph include basic environmental parameters, derived parameters, and process state indicators. Edge weights represent the strength of the association between features, ranging from 0.1 to 1.0. When analyzing a feature related to a certain state, the system extracts the link degree of all features corresponding to that feature from the graph. The link degree calculation considers direct and indirect connections; the edge weight of a direct connection is multiplied by 1.0, a first-degree indirect connection by 0.5, and a second-degree indirect connection by 0.2.

[0043] The core state entropy is calculated by multiplying the state dimension entropy and the feature connectivity. The system sets the threshold for core state entropy to 1.5 times the average value of all features. When the core state entropy of a feature exceeds the threshold, that feature is marked as a core state feature. All core state features are arranged in chronological order to form a core state feature sequence.

[0044] Table 1: Analysis of the State Characteristics of Key Nodes.

[0045]

[0046] Based on the analysis results, temperature load, CO2 concentration load, temperature change rate, and CO2 concentration change rate were identified as core state features. The state core entropy of these features all exceeded the threshold of 11.25 (average 7.5). The system arranged these core features in chronological order to construct a core state feature sequence. This sequence reflects the most important state change features during acid removal, providing a basis for anomaly detection.

[0047] In subsequent analysis, the system continuously monitors changes in the core state characteristic sequence. When an abnormal pattern appears in the sequence, such as a sudden deviation of a core characteristic value from the normal range, or inconsistent changes in multiple core characteristics, the system will trigger an early warning mechanism. The establishment of the core state characteristic sequence allows the system to focus on the most critical process parameters, improving the accuracy and timeliness of anomaly detection. The entire implementation process employs a rolling update mechanism, recalculating the core state entropy every 10 minutes and dynamically adjusting the core state characteristic sequence. This dynamic update ensures that the system can adapt to changes in environmental conditions during acid discharge, maintaining the sensitivity and reliability of monitoring. The final generated core state characteristic sequence is interfaced with the anomaly indicator identification module, providing a decision-making basis for the generation of control commands.

[0048] Example 5: The control execution module receives abnormal indicator data from the status analysis module. This data includes timestamps, spatial location identifiers, abnormal feature types, and deviation information. In the specific implementation process, the system first performs structural transformation on the abnormal indicators, integrating discrete abnormal reports into a control data matrix. The row dimension of this matrix corresponds to continuous time points, with a five-minute interval; the column dimension contains three types of control parameters: temperature compensation, humidity adjustment, and gas concentration correction. For example, when an abnormal lactic acid volatilization is detected in unit "A-35-28-04" of the main acid discharge area, the system generates a matrix row containing the location identifier of that unit, with the temperature compensation column filled with -0.5℃ (negative values ​​indicate that cooling is required), and the gas concentration column filled with +200ppm (positive values ​​indicate that ventilation is required).

[0049] The control data matrix undergoes decomposition and reconstruction. Signal separation technology is used to break down the original matrix into two components: the reconstruction matrix carries smoothed trend-based control commands, fills data gaps using interpolation algorithms, and filters out instantaneous fluctuation noise; the indicator matrix retains the original abnormal feature point location markers and uses binary identifiers to mark abnormal areas requiring special attention. For example, when a temperature anomaly occurs in a region at three consecutive time points, a gradual adjustment curve is generated at the corresponding location in the reconstruction matrix, while the corresponding location in the indicator matrix retains its anomaly marker status.

[0050] Based on the reconstruction matrix, an initial control scheme is generated. The system calls a preset aging strategy knowledge base, which contains the ideal environmental parameter ranges and adjustment rules for different parts of mutton and different aging stages. For the temperature compensation of -0.5℃ in the reconstruction matrix, the strategy base is consulted to derive the corresponding measure: when the target temperature drop is less than 1℃, the speed of the air cooler unit is adjusted first, rather than starting the compressor. Based on this, the initial command is generated: reduce the speed of air cooler unit No. 3 by 15%, and simultaneously open the ventilation louvers in zone B at a 30-degree angle. The initial scheme forms a structured operation sequence, including equipment number, action type, execution parameters, and other elements.

[0051] The anomaly indicator matrix drives a dynamic adjustment mechanism. The system monitors environmental parameter changes in the marked areas of the indicator matrix in real time. If an anomaly-marked area fails to return to normal after two consecutive detections, a solution upgrade procedure is triggered. For example, if an anomaly marker for gas concentration persists in a certain unit, the system automatically increases the monitoring frequency for that area to 10 seconds per instance, while simultaneously analyzing data from related areas. If similar anomalies are found in three adjacent units, additional measures are implemented based on the initial plan: the backup ventilation unit is activated, and the adjustment range is increased from 30% to 45%. The adjustment process employs fuzzy control logic, responding in stages based on factors such as the duration and scope of the anomaly.

[0052] The updated control scheme is converted into execution commands through the device interface layer. Temperature regulation commands drive twelve sets of temperature control actuators distributed on the ceiling, achieving ±0.3℃ precision control by adjusting the refrigerant valve opening. Humidity control commands activate the atomizing humidifier array, adjusting the ambient humidity in 0.5% RH gradients. Gas concentration correction operates the variable frequency ventilation system, changing the air composition in 50 ppm increments. After all commands are executed, the system collects feedback data in real time. When the environmental parameters deviate from the target value for more than 10 minutes, a new control cycle is automatically initiated.

[0053] A complete audit trail is established during execution, recording the scheme version number, decision basis matrix fingerprint, and execution equipment status snapshot each time a control command is generated. Execution results generate an effectiveness evaluation report, including parameter adjustment curves, target achievement indicators, and equipment energy consumption increments. These records are stored in association with the original anomaly indicators, forming a closed-loop evidence chain from anomaly detection to control feedback. In typical application scenarios, the system's average response time for handling a localized temperature anomaly is 90 seconds, and the average adjustment cycle from command issuance to environmental parameter stabilization is 8 minutes. All control operations achieve precise control of localized anomaly areas while maintaining overall stability of the acid discharge environment.

[0054] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0055] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A monitoring system based on the aging process of mutton after cutting, characterized in that, include: The data acquisition module is used to acquire monitoring data in the acid-removing environment after mutton is cut in real time. The monitoring data includes temperature data, humidity data, gas concentration data, and cutting area location data, and the module extracts the initial spatiotemporal sequence from the monitoring data. The feature extraction module is used to extract coupling features from the initial spatiotemporal sequence; The distribution generation module is used to generate partitions of different levels of detail based on the coupling features and the resolution of monitoring data at different locations in the aging environment after mutton is cut, and to obtain a multi-resolution feature distribution based on the partition generation results; the partitions corresponding to the high-resolution regions are more detailed than those corresponding to the low-resolution regions. The state analysis module is used to analyze the time series of the multi-resolution feature distribution to obtain the acid excretion state spectrum; The control execution module is used to identify abnormal indicators of key nodes based on the acid excretion state spectrum, and generate control commands according to the abnormal indicators to adjust the acid excretion environment parameters.

2. The monitoring system based on the aging process after mutton cutting as described in claim 1, characterized in that, The data acquisition module acquires real-time monitoring data of the aging environment after the mutton is cut, including: The temperature data, humidity data, gas concentration data, and segmentation area location data are standardized to obtain standardized monitoring data. The initial spatiotemporal sequence is extracted from the standardized monitoring data using a feature extraction unit.

3. The monitoring system based on the aging process after mutton cutting as described in claim 1, characterized in that, The feature extraction module extracts coupling features from the initial spatiotemporal sequence, including: A preliminary feature set, coupled with temperature data, humidity data, and gas concentration data, is extracted from the initial spatiotemporal sequence; When the feature dimension of the preliminary feature set is greater than a preset dimension threshold, the preliminary feature set is subjected to dimensionality reduction processing to obtain dimensionality-reduced feature data; The time series of the dimensionality-reduced feature data is analyzed, and the dynamic characteristics of the interaction between temperature data, humidity data, and gas concentration data are enhanced to obtain the coupling features.

4. The monitoring system based on the aging process after mutton cutting as described in claim 1, characterized in that, Based on the coupling features, the distribution generation module generates partitions of varying fineness according to the resolution of monitoring data from different locations in the aging environment after mutton slicing. The resulting multi-resolution feature distribution is obtained from the partition generation results, including: The aging environment of the mutton after it is cut is divided into multiple sub-regions; Construct multiple processing queues, each corresponding to a sub-region; Based on the resolution of the monitoring data for each sub-region, determine the unit size, number of partitions, and partition density for each sub-region; The multiple processing queues are controlled to generate target partitions for each sub-region according to the unit size, number of partitions, and partition density of each sub-region; Based on the coupling features and the target partitions of each sub-region, the multi-resolution feature distribution is obtained.

5. A monitoring system based on the aging process after mutton cutting as described in claim 4, characterized in that, The distribution generation module controls the multiple processing queues to generate target partitions for each sub-region according to the unit size, number of partitions, and partition density of each sub-region, including: The multiple processing queues are controlled to generate preliminary partition surface information according to the unit size, number of partitions, and partition density of each sub-region; The multiple processing queues are controlled to generate three-dimensional partitions of each sub-region based on the partition surface information and obtain partition volume information; Based on the partition information, obtain the vertices and volume units of each solid partition; and number the vertices and volume units of each solid partition to obtain a unique identifier for each vertex and a unique identifier for each volume unit. The system controls the synchronous communication between the multiple processing queues to map the volume units and unique identifiers between adjacent sub-regions to obtain mapping information. The three-dimensional partitioning of each sub-region is optimized based on the mapping information to obtain the target partitioning of each sub-region.

6. The monitoring system based on the aging process after mutton cutting as described in claim 1, characterized in that, The state analysis module analyzes the time series of the multi-resolution feature distribution to obtain the acid excretion state spectrum, including: Dynamic characteristic analysis is performed on the time series of the multi-resolution feature distribution to extract key change points; Based on the key change points, the acid removal state spectrum is constructed to represent the dynamic load of the acid removal process.

7. A monitoring system based on the aging process after mutton cutting as described in claim 1, characterized in that, The control execution module identifies abnormal indicators at key nodes based on the acid discharge state spectrum, including: Determine the state feature vector of the key node; Calculate the state dimension entropy of each key node based on the state feature vector. Based on the entropy of all state dimensions and the preset process link graph, the core state feature sequence is selected. The abnormal indicators are identified based on the core state feature sequence.

8. A monitoring system based on the aging process after mutton cutting as described in claim 7, characterized in that, The control execution module, based on the entropy of all state dimensions and a preset process link graph, selects core state feature sequences, including: For each state-related feature in the state feature vector, determine the state dimension entropy of the dimension in which the state-related feature is located. Extract all feature linkage degrees corresponding to the state-related features from the preset process link graph; The state core entropy of the state-related feature is determined by the state dimension entropy of the dimension in which the state-related feature is located and the corresponding feature linkage degree. All core state features are selected based on the state core entropy of each state-related feature. The core state feature sequence is constructed based on all core state features.

9. A monitoring system based on the aging process after mutton cutting as described in claim 1, characterized in that, The control execution module generates control commands based on the abnormal indicators to adjust the acid discharge environment parameters, including: The abnormal indicators are reconstructed to obtain target control data for environmental control and abnormal indication data for dynamic adjustment. Based on the target control data and combined with the preset acid removal strategy, an initial control scheme is generated; Based on the anomaly indication data, anomalies in the real-time monitoring data are identified, and the initial control scheme is dynamically adjusted according to the anomalies to generate an updated control scheme. The control execution module executes the updated control scheme to adjust the temperature data, humidity data, or gas concentration data.

10. A monitoring system based on the aging process after mutton cutting as described in claim 9, characterized in that, The control execution module reconstructs the abnormal indicators to obtain target control data for environmental control and abnormal indication data for dynamic adjustment, including: The abnormal indicators are converted into a control data matrix; The control data matrix is ​​decomposed to obtain a reconstructed matrix and an indicator matrix; The reconstruction matrix represents the target control data structure after interpolation and denoising; The indicator matrix is ​​used to identify abnormal data points; The reconstructed matrix is ​​used as the target control data, and the indication matrix is ​​used as the anomaly indication data.

Images