Intelligent factory process optimization method and system based on multi-modal data fusion

Abstract

The invention provides an intelligent factory process optimization method and system based on multi-modal data fusion, and the method comprises the steps: obtaining a multi-modal data set generated in the production process of an intelligent factory, carrying out the multi-modal fusion processing of the multi-modal data set, generating a fusion data set, and carrying out the process feature extraction operation based on the fusion data set. The method comprises the following steps: generating a process optimization feature set, carrying out dynamic optimization strategy matching processing according to the process optimization feature set, generating a process parameter optimization strategy, executing the process parameter optimization strategy, carrying out optimization adjustment on a process execution process of the intelligent factory, and monitoring optimized process execution state data in real time to trigger a strategy iteration updating operation. According to the method, the production quality stability and the process resource utilization efficiency can be synchronously improved.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for intelligent factory process optimization based on multimodal data fusion. Background Art

[0002] With the in-depth application of intelligent manufacturing technologies, process optimization in smart factories has become a key link in improving production efficiency. Current mainstream process optimization technologies discretely process equipment operating data, process parameters, and environmental monitoring information to formulate equipment maintenance strategies, process adjustment plans, and environmental control measures. While such methods can achieve local parameter optimization, they struggle to capture the dynamic correlation between equipment performance degradation and process execution deviations, and are unable to identify the transmission path of environmental interference within the equipment-process system, resulting in equipment-side response lag and a mismatch between environmental compensation measures. Furthermore, existing technologies lack a multi-source data feature fusion mechanism, leading to a conflict of objectives between equipment operation optimization and process parameter adjustment strategies. Environmental compensation measures often lag behind the evolution of actual interference patterns, resulting in wasted process resources and fluctuations in quality control. Building a collaborative optimization system for equipment status, process execution, and environmental interference to achieve dynamic process adjustment with full linkage has become a pressing challenge for the refined management and control of smart factories. Summary of the Invention

[0003] The present invention provides a method and system for intelligent factory process optimization based on multimodal data fusion.

[0004] In a first aspect, an embodiment of the present invention provides a smart factory process optimization method based on multimodal data fusion, comprising: obtaining a multimodal data set generated in the production process of the smart factory, the multimodal data set comprising equipment operation parameter data, process execution process data and environmental monitoring data; performing multimodal fusion processing on the multimodal data set to generate a fused data set; the fused data units in the fused data set comprise dynamic correlation features between the equipment operation parameter data and the process execution process data, and coupling features between the process execution process data and the environmental monitoring data; performing a process feature extraction operation based on the fused data set to generate a process optimization feature set; the process optimization feature set comprises equipment operation stability features, process parameter adaptability features and environmental interference suppression features; performing dynamic optimization strategy matching processing based on the process optimization feature set to generate a process parameter optimization strategy; the process parameter optimization strategy comprises equipment operation parameter adjustment direction, process execution parameter correction direction and environmental compensation parameter adaptation direction; executing the process parameter optimization strategy to optimize and adjust the process execution process of the smart factory, and monitoring the optimized process execution status data in real time to trigger a strategy iteration update operation.

[0005] In a second aspect, an embodiment of the present invention provides a computer system, comprising: a memory storing a computer program; and a processor for loading the computer program to implement the smart factory process optimization method based on multimodal data fusion as described above.

[0006] The intelligent factory process optimization method based on multimodal data fusion provided by the present invention constructs a dynamic mapping relationship between equipment state fluctuations and process accuracy changes through cross-modal feature association of equipment operating parameters, process execution data and environmental monitoring data, and accurately captures the propagation path of environmental interference in the process chain; based on the equipment operation stability characteristics and process parameter adaptation characteristics extracted from the fused data, a two-way collaborative mechanism for dynamic adjustment of process parameters and self-adaptation of environmental compensation parameters is established, so that equipment load changes, process execution deviations and environmental disturbance diffusion form a real-time linkage compensation relationship; through the dynamic matching mechanism of process optimization strategy, equipment operation parameter adjustment, process execution correction and environmental compensation adaptation are transformed into mutually constrained optimization target groups, eliminating the inherent contradiction between equipment-side response lag and environmental-side compensation redundancy in traditional process optimization; combined with real-time feedback of strategy execution effect data, a synchronous update system for process optimization direction and production environment evolution is constructed, so that the process optimization process has the ability to autonomously adapt to equipment performance degradation and environmental interference mode migration, and ultimately achieve the simultaneous improvement of production quality stability and process resource utilization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 This is a flowchart of a smart factory process optimization method based on multimodal data fusion provided by an embodiment of the present invention.

[0008] Figure 2 It is a schematic diagram of the composition of a computer system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0009] See also Figure 1 , Figure 1 A flowchart of a smart factory process optimization method based on multimodal data fusion provided by an embodiment of the present invention. The method can be executed by a computer system and may specifically include the following steps: Step S100: Acquire a multimodal data set generated during the production process of the smart factory, where the multimodal data set includes equipment operating parameter data, process execution data, and environmental monitoring data.

[0010] A multimodal data set is a collection of data of different types and modalities. During the production process of a smart factory, this data can reflect the status and situation of production from multiple dimensions. Equipment operating parameter data describes the operating status of various equipment in a smart factory, such as speed, temperature, pressure, and current. This data can reflect the real-time operating status and performance of the equipment. Process execution data is data generated during the process execution process, covering execution information for each link and step of the process, such as the execution time, execution sequence, and execution accuracy of the process. It reflects the specific process and results of the process execution. Environmental monitoring data is data obtained by monitoring the production environment of a smart factory, including environmental parameters such as temperature, humidity, light, and noise. This data can affect the production process and product quality.

[0011] As an implementation manner, step S100 may specifically include the following steps S110 to S140: Step S110: Collecting original equipment operating parameter data through the sensor network deployed in the smart factory equipment layer, and performing data integrity verification and outlier elimination processing on the original equipment operating parameter data to generate standardized equipment operating parameter data.

[0012] A sensor network is a network system composed of multiple sensor nodes distributed across the equipment layer of a smart factory, collecting real-time equipment operating parameter data. Raw equipment operating parameter data is directly collected by the sensor network and may be incomplete or contain outliers. Data integrity verification checks the data for missing segments. This process analyzes the data's timestamp sequence, identifies discontinuities, and locates the missing segments. Outlier rejection identifies and removes data points that significantly deviate from the normal range. Standardized equipment operating parameter data is processed to be complete, free of outliers, and consistent in dimension.

[0013] As an embodiment, in step S110, data integrity check and outlier elimination processing are performed on the original equipment operating parameter data to generate standardized equipment operating parameter data, which may specifically include the following steps S111 to S114: Step S111: Detect missing data segments in the original device operating parameter data, and complete the missing data segments using a data interpolation algorithm of adjacent timestamps.

[0014] Raw equipment operating parameter data is unprocessed data collected directly by sensors, which may contain missing data, resulting in missing data segments. Missing data segments can affect subsequent data analysis and processing, so they need to be detected and supplemented.

[0015] In actual smart factory production scenarios, if the raw operating parameter data collected by a device's temperature sensor contains missing data segments, analyzing the data's timestamps may reveal missing data between two consecutive timestamps. To complete this missing data segment, a data interpolation algorithm, such as linear interpolation, is used to interpolate data between adjacent timestamps.

[0016] Step S112: Calculate the statistical distribution characteristics of the original equipment operating parameter data, identify abnormal data points that exceed the preset reasonable value range, and use a sliding window mean filtering algorithm to smooth and replace the abnormal data points.

[0017] Statistical distribution characteristics are a set of parameters that describe the distribution of data, such as mean, standard deviation, and variance. By calculating the statistical distribution characteristics of raw equipment operating parameter data, we can understand the data's central tendency and degree of dispersion. The default reasonable value range is a predefined range of data values based on the normal operating range of the equipment. Data points outside this range are considered outliers. The sliding window mean filtering algorithm calculates the mean of the data within a fixed-size window around the data point and replaces the data point with this mean, thereby reducing noise and outliers in the data.

[0018] Taking the pressure data of a certain device as an example, calculate the statistical distribution characteristics of the original device operating parameter data (i.e. pressure data), such as calculating its mean and standard deviation According to the normal operating range of the equipment, the preset reasonable value range is , where k is a preset constant, such as 2 or 3. By traversing the pressure data, data points that exceed the preset reasonable value range are identified as abnormal data points. For these abnormal data points, a sliding window mean filter algorithm is used for smooth replacement. If the size of the sliding window is n, with the abnormal data point as the center, the front and back data points are selected. data points (or take available data points if there are not enough), calculate the mean of these data points, and replace the outlier data points with the mean.

[0019] Step S113: performing dimensional unification processing on the processed equipment operation parameter data, and converting it into equipment operation parameter data of a preset standard unit.

[0020] The processed equipment operating parameter data is data that has undergone data integrity verification and outlier elimination, but these data may have different dimensions. Dimension unification is the process of converting data of different dimensions into data of preset standard units. The preset standard units are unified units set in advance to facilitate data analysis and comparison.

[0021] Step S114: Store the equipment operation parameter data of the standard unit into the equipment operation database, and configure the equipment identifier and collection timestamp for it.

[0022] Equipment operating parameter data in standard units is dimensionally standardized. This data has a unified dimension and format, making it easier to store and manage. The equipment operation database is a database specifically designed to store equipment operating parameter data, providing data storage, query, and management capabilities. A device identifier is a code or number that uniquely identifies a device. This identifier allows you to distinguish operating parameter data from different devices. The collection timestamp records the time the data was collected, reflecting the timeliness and order of the data.

[0023] Step S120: Acquire process execution data through the data interface of the process control system, and perform time series normalization processing on the process execution data to generate process execution data with a unified time reference.

[0024] The process control system is a system used in smart factories to control and manage process execution. It records various data during process execution. The data interface provides a way for the process control system to exchange data with external systems and access process execution data. Process execution data is generated during process execution and includes information such as the execution time, execution sequence, and execution accuracy. Time series normalization is the process of unifying the time base of different process execution data. This unified time base makes different data comparable, facilitating subsequent analysis and processing.

[0025] Step S130: Collect environmental monitoring data through environmental monitoring equipment, and perform spatial interpolation processing on the environmental monitoring data to generate environmental monitoring data covering the entire smart factory.

[0026] Environmental monitoring equipment, such as temperature sensors, humidity sensors, and light sensors, is used to monitor the production environment parameters of smart factories. These sensors are distributed throughout the smart factory and collect real-time environmental monitoring data. Environmental monitoring data reflects the production environment conditions of the smart factory, including parameters such as temperature, humidity, light, and noise. Spatial interpolation, a method for converting discrete environmental monitoring data into continuous spatially distributed data, can be used to obtain environmental parameter values for every location within the smart factory. Kriging interpolation, for example, is used to calculate the spatial distances and correlations between known data points to construct a covariance matrix. Then, based on the covariance matrix and the relationship between the unknown location and the known data points, weight coefficients are calculated. Finally, the temperature values of the known data points are weighted and summed according to the weight coefficients to obtain the predicted temperature value for the unknown location.

[0027] Step S140: Convert the standardized equipment operating parameter data, the process execution data with a unified time base, and the environmental monitoring data covering the entire smart factory into a multimodal data set.

[0028] Data format conversion is the process of converting different types of data into the same format. A multimodal data set consists of standardized equipment operating parameter data, process execution data with a unified time base, and environmental monitoring data covering the entire smart factory. For example, equipment operating parameter data might be in CSV format, process execution data might be in JSON format, and environmental monitoring data might be in XML format. To combine this data into a multimodal data set, data format conversion is required. You can choose a unified data format, such as JSON, and convert data in different formats to JSON.

[0029] Step S200: performing multimodal fusion processing on the multimodal data set to generate a fused data set; the fused data units in the fused data set include dynamic association features between equipment operation parameter data and process execution process data, and coupling features between process execution process data and environmental monitoring data.

[0030] The dynamic correlation between equipment operating parameter data and process execution data reflects the dynamic impact of equipment operating status on process execution. For example, how changes in equipment parameters such as speed and temperature affect process execution accuracy and efficiency. The coupling between process execution data and environmental monitoring data reflects the interference and impact of environmental parameters on process execution. For example, how fluctuations in environmental parameters such as temperature and humidity affect process stability and quality.

[0031] As an implementation manner, step S200 may specifically include the following steps S210 to S250: Step S210: perform timestamp alignment processing on the equipment operation parameter data to generate a time series data sequence; perform spatial registration processing on the process execution process data to generate a spatial data sequence; perform noise filtering processing on the environmental monitoring data to generate an environmental data sequence.

[0032] Timestamp alignment is the process of arranging equipment operating parameter data in chronological order and making the timestamps of different data consistent. A time series data sequence is a sequence of equipment operating parameter data arranged in chronological order after timestamp alignment, which can reflect the changes in equipment operating parameters over time. Spatial registration is the process of matching and calibrating process execution data with spatial positions. A spatial data sequence is a sequence of process execution data associated with spatial positions after spatial registration, which can reflect the conditions of process execution at different spatial positions. Noise filtering is the process of removing noise interference from environmental monitoring data. An environmental data sequence is a sequence of environmental monitoring data that has been subjected to noise filtering to remove noise interference, which can more accurately reflect the true conditions of environmental parameters.

[0033] Step S220: Input the time series data sequence, spatial data sequence and environmental data sequence into a preconfigured multimodal fusion model, and perform cross-modal correlation analysis on the time series fluctuation characteristics of the time series data sequence, the spatial distribution characteristics of the spatial data sequence and the environmental disturbance characteristics of the environmental data sequence through the multimodal fusion model.

[0034] The preconfigured multimodal fusion model is a pre-trained model for processing multimodal data, which can explore the associations and interactions between different types of data. Time series fluctuation characteristics are the characteristics of the fluctuation and change of data in a time series data sequence over time, such as periodic fluctuations and sudden changes in equipment operating parameters. Spatial distribution characteristics are the spatial distribution characteristics of data in a spatial data sequence, such as the distribution and density of process execution in different areas. Environmental disturbance characteristics are the fluctuation and change characteristics of environmental parameters in an environmental data sequence, such as periodic changes and sudden fluctuations in temperature and humidity. Cross-modal association analysis is the process of analyzing the associations and interactions between different types of data. Through cross-modal association analysis, the intrinsic connection between equipment operation, process execution and environmental factors can be discovered.

[0035] Multimodal fusion models can use deep learning architectures, such as a combination of convolutional neural networks (CNNs) and recurrent neural networks (RNNs). CNNs can be used to extract spatial distribution features of spatial data sequences, while RNNs can be used to process temporal fluctuations in temporal data sequences.

[0036] As an implementation manner, step S220 may specifically include the following steps S221 to S226: Step S221: Decompose the time dimension of the time series fluctuation characteristics to generate a set of fluctuation components of the time series fluctuation characteristics at different time scales. The fluctuation component set includes short-term fluctuation components, medium-term fluctuation components and long-term trend components.

[0037] Time dimension decomposition is the process of decomposing time series fluctuation characteristics according to time scales. Through this process, time series fluctuation characteristics can be decomposed into fluctuation components at different time scales. Short-term fluctuation components refer to the fluctuation changes of time series fluctuation characteristics within a shorter time scale, such as the fluctuation of equipment operating parameters within minutes or hours. Medium-term fluctuation components refer to the fluctuation changes of time series fluctuation characteristics within a medium time scale, such as the fluctuation of equipment operating parameters within days or weeks. Long-term trend components refer to the changing trend of time series fluctuation characteristics within a longer time scale, such as the changing trend of equipment operating parameters within months or years.

[0038] The empirical mode decomposition (EMD) algorithm is used to decompose the time series fluctuation characteristics in the time dimension. The basic principle of the EMD algorithm is to decompose the time series data into a series of intrinsic mode functions (IMFs) and a residual component. Each IMF represents a fluctuation component at a different time scale. The specific steps are as follows: find all the maximum and minimum points of the time series fluctuation characteristic data, fit the maximum and minimum points respectively through cubic spline interpolation, and obtain the upper envelope and lower envelope. Calculate the average of the upper and lower envelopes, subtract the average from the original data to obtain a new series. Repeat the above steps until the new series meets the IMF conditions, and obtain an IMF component. Subtract this IMF component from the original data to obtain the residual series. Repeat the above process for the residual series until the residual series is a monotonic function or a constant, and obtain all IMF components and residual components. According to the oscillation frequency of the IMF components, they are divided into short-term fluctuation components, medium-term fluctuation components, and long-term trend components.

[0039] Step S222: performing spatial region segmentation processing on the spatial distribution features to generate multiple spatial region features corresponding to the spatial distribution features, the spatial region features including equipment layout associated region features, process execution intensive region features and environmental interference diffusion region features.

[0040] Spatial region segmentation is the process of dividing spatial distribution features into different spatial regions according to preset rules. Equipment layout-related region features are spatial region features related to equipment layout, such as the impact area around the equipment and the connection area between equipment. Process execution-intensive region features are spatial region features where process execution is relatively concentrated, such as areas where production lines are located and areas with frequent process operations. Environmental interference diffusion region features are regional features where environmental interference factors diffuse spatially, such as areas with significant variations in environmental parameters such as temperature and humidity.

[0041] In smart factories, spatial distribution characteristics can be represented by process execution data and equipment layout information. A density-based spatial clustering algorithm (such as the DBSCAN algorithm) is used to segment spatial distribution characteristics into spatial regions. First, a neighborhood radius and a minimum number of points are defined. For each data point, the number of points within the neighborhood radius is calculated. If the number of points is greater than or equal to the minimum number of points, the data point is considered a core point. Adjacent core points are connected to form a clustering region. For non-core points, if they are within the neighborhood of a core point, they are included in the clustering region; otherwise, they are marked as noise points. Based on the clustering results, the spatial distribution characteristics are divided into different spatial regions. Then, the characteristics of each spatial region are determined based on the equipment layout information and process execution data.

[0042] Step S223: performing disturbance pattern extraction processing on the environmental disturbance feature to generate a plurality of disturbance pattern features corresponding to the environmental disturbance feature, wherein the disturbance pattern features include a periodic disturbance pattern, a sudden disturbance pattern, and a continuous disturbance pattern.

[0043] Disturbance pattern extraction is the process of extracting different disturbance patterns from environmental disturbance features. Periodic disturbance patterns refer to patterns in which environmental disturbance features exhibit periodic changes, such as diurnal temperature fluctuations or periodic vibrations during equipment operation. Sudden disturbance patterns refer to patterns in which environmental disturbance features undergo sudden, significant changes, such as sudden temperature rises or equipment failures. Persistent disturbance patterns refer to patterns in which environmental disturbance features continue to change over an extended period, such as long-term temperature rises or sustained increases in humidity.

[0044] A method combining Fourier transform and anomaly detection algorithms is used to extract disturbance patterns from environmental disturbance characteristics. First, a Fourier transform is performed on the ambient temperature data, converting it from the time domain to the frequency domain. By analyzing the spectral characteristics in the frequency domain, frequency components with obvious periodicity are identified. The time domain signals corresponding to these frequency components are periodic disturbance patterns. For sudden disturbance patterns, the isolation forest algorithm is used for anomaly detection. The basic principle of the isolation forest algorithm is to construct a random binary tree to partition data points into different leaf nodes. Anomalous data points are typically located at shallow locations in the tree. By calculating the path length of the data point, it can be determined whether it is an anomaly. When an anomaly is detected in the ambient temperature data, it is considered that a sudden disturbance has occurred at that moment. For persistent disturbance patterns, trend analysis methods such as linear regression are used. By performing a linear regression on the ambient temperature data, the slope is calculated. If the slope is non-zero and persists for a period of time, it is considered that a persistent disturbance pattern exists.

[0045] Step S224: Perform first modal correlation matching on the short-term fluctuation component in the fluctuation component set and the equipment layout associated area feature in the spatial area feature to generate real-time impact features of equipment operation and spatial layout; perform second modal correlation matching on the medium-term fluctuation component and the process execution intensive area feature to generate dynamic interaction features of process parameters and equipment load; perform third modal correlation matching on the long-term trend component and the environmental interference diffusion area feature to generate environmental interference cumulative effect features.

[0046] First modal correlation matching involves correlating and matching the short-term fluctuation components in the fluctuation component set with the equipment layout-related regional features in the spatial regional features. This matching allows the real-time impact of short-term fluctuations in equipment operation on the equipment layout to be revealed. The real-time impact feature of equipment operation and spatial layout, derived through first modal correlation matching, reflects how short-term fluctuations in equipment operation are affected by the equipment layout, and vice versa. Second modal correlation matching involves correlating and matching the medium-term fluctuation components with the features of densely populated process execution areas. This allows the dynamic interaction between process parameters and equipment loads to be explored. The dynamic interaction feature of process parameters and equipment loads, derived through second modal correlation matching, reflects the interaction between changes in process parameters during process execution and medium-term fluctuations in equipment loads. Third modal correlation matching involves correlating and matching the long-term trend component with the features of diffused environmental interference areas. This matching allows the analysis of the long-term cumulative effects of environmental interference. The cumulative effect feature of environmental interference, derived through third modal correlation matching, reflects the long-term cumulative impact of environmental interference factors on equipment operation and process execution.

[0047] For the first modal correlation matching, the short-term fluctuation component of equipment speed and the characteristics of the equipment layout-related area are used as an example. Statistical characteristics of the short-term fluctuation component within the equipment layout-related area, such as the mean and variance, are calculated. Simultaneously, the topological structure of the equipment layout-related area and the connectivity between devices are analyzed. Correlation analysis is used to identify the correlation between the short-term fluctuation component and the characteristics of the equipment layout-related area. For example, if a certain equipment layout-related area shows large short-term fluctuations in equipment speed and complex equipment connectivity, this indicates that the equipment layout may have an impact on the short-term fluctuations in equipment operation. This generates a real-time impact feature on equipment operation and spatial layout. For the second modal correlation matching, the medium-term fluctuation component of equipment power and the characteristics of the process-intensive area are used as an example. Process parameters within the process-intensive area, such as process speed and process pressure, are analyzed. Through regression analysis, a regression model is established between the medium-term fluctuation component and the process parameters, identifying the dynamic interaction between the process parameters and the medium-term fluctuations in equipment load. For example, if a positive correlation is found between changes in process speed and medium-term fluctuations in equipment power, this indicates that changes in process parameters affect the medium-term fluctuations in equipment load. This generates a dynamic interaction feature between the process parameters and equipment load. For the third modal correlation matching, we use the long-term trend component of ambient temperature and the characteristics of the environmental interference diffusion area as examples. We analyze the changes in environmental parameters such as temperature and humidity within the environmental interference diffusion area. Through integral calculations, we obtain the long-term cumulative effects of environmental interference factors.

[0048] Step S225: Perform time-space coupling analysis on the periodic disturbance pattern and the real-time impact characteristics to generate the equipment operation offset characteristics under periodic environmental interference; perform abnormal event correlation analysis on the sudden disturbance pattern and the dynamic interaction characteristics to generate the process parameter mutation warning characteristics; perform trend superposition analysis on the continuous disturbance pattern and the cumulative effect characteristics of the environmental interference to generate the long-term environmental interference suppression demand characteristics.

[0049] Specifically, for time-space coupling analysis, we use the periodic temperature disturbance pattern and the real-time impact characteristics of equipment operation and spatial layout as examples. The periodic temperature disturbance pattern is discretized in time to obtain temperature disturbance values at different time points. Then, based on the real-time impact characteristics, the operating status of the equipment at different spatial locations is determined. The temperature disturbance values are matched with the equipment operating status in time and space to analyze the impact of periodic temperature disturbances on equipment operation. For example, if a piece of equipment experiences a significant deviation in operating speed when its temperature rises periodically, a device operation deviation characteristic under periodic environmental disturbances is generated. For abnormal event correlation analysis, we use the disturbance pattern of sudden equipment failure and the dynamic interaction characteristics between process parameters and equipment load as examples. When a sudden equipment failure disturbance event is detected, we analyze the changes in process parameters before and after the event. Through correlation analysis, we identify the relationship between the sudden equipment failure and the process parameter changes. For example, if a sudden decrease in process execution speed is observed after a equipment failure, a process parameter mutation warning characteristic is generated. For trend overlay analysis, we use the disturbance pattern of persistent humidity increase and the cumulative effect characteristics of environmental disturbances as examples. The trend of the persistent humidity increase disturbance pattern is superimposed on the characteristics of the cumulative effect of environmental disturbances to analyze the comprehensive impact of long-term humidity increase on equipment operation and process execution.

[0050] Step S226: Integrate the equipment operation offset characteristics, process parameter mutation warning characteristics and long-term environmental interference suppression demand characteristics to generate dynamic correlation characteristics between equipment operation parameter data and process execution process data, and generate coupling characteristics between process execution process data and environmental monitoring data based on the correlation results of periodic disturbance patterns and process execution intensive area characteristics.

[0051] Integration is the process of comprehensively processing and combining the characteristics of equipment operation offsets, process parameter mutation warning characteristics, and long-term environmental interference suppression requirements. Through integration, a comprehensive feature representation can be obtained, reflecting the dynamic correlation between equipment operation parameter data and process execution process data. The dynamic correlation characteristics between equipment operation parameter data and process execution process data are characteristics obtained through integration. It reflects how changes in equipment operation status affect the process execution process, and how changes in the process execution process are fed back to the equipment operation parameters. Based on the correlation results of periodic disturbance patterns and process execution intensive area characteristics, the coupling characteristics between process execution process data and environmental monitoring data are generated. This is to analyze the impact of periodic environmental interference on process execution intensive areas, thereby obtaining the interaction relationship between process execution process data and environmental monitoring data.

[0052] In smart factories, to integrate equipment operation offset characteristics, process parameter mutation warning characteristics, and long-term environmental interference suppression requirements, these characteristics are first standardized to have the same dimension and range. Then, these characteristics are integrated using a weighted summation method. For example, different weights are assigned to equipment operation offset characteristics, process parameter mutation warning characteristics, and long-term environmental interference suppression requirements. Their weighted sum is calculated based on the weights to obtain dynamic correlation characteristics between equipment operation parameter data and process execution process data. To generate coupling characteristics between process execution process data and environmental monitoring data, the correlation between periodic disturbance patterns (such as periodic temperature changes) and characteristics of process execution-intensive areas (such as process execution speed and process execution accuracy) is analyzed. By calculating statistics such as correlation coefficients and covariances, the extent of the impact of periodic environmental interference on process execution-intensive areas is determined.

[0053] Step S230: Based on the results of the cross-modal correlation analysis, construct dynamic correlation features between the equipment operating parameter data and the process execution process data, wherein the dynamic correlation features represent the weight of the impact of changes in the equipment operating parameters on the process execution accuracy.

[0054] Specifically, based on the correlation information between equipment operating parameter data and process execution data obtained through cross-modal correlation analysis, dynamic correlation features between the equipment operating parameter data and process execution data are constructed. For example, a regression analysis method is used to establish a regression model with process execution accuracy as the dependent variable and equipment operating parameters as the independent variables. By fitting historical data, the coefficients of the regression model are obtained, namely the weights of the impact of changes in equipment operating parameters on process execution accuracy. Assume that the equipment operating parameters include equipment speed, equipment temperature, and equipment pressure, and process execution accuracy is represented by the product qualification rate. Regression analysis results in the regression equation: qualification rate = a*speed + b*temperature + c*pressure + d, where a, b, and c are the weights of the impact of changes in equipment speed, equipment temperature, and equipment pressure on process execution accuracy, respectively. In this way, dynamic correlation features are constructed between equipment operating parameter data and process execution data, providing an important reference basis for subsequent process optimization.

[0055] Step S240: Determine the coupling characteristics between the process execution data and the environmental monitoring data based on the noise filtering result of the environmental monitoring data. The coupling characteristics reflect the interference intensity of the environmental parameter fluctuation on the stability of the process execution.

[0056] The noise filtering results of environmental monitoring data are obtained after noise filtering. This removes noise interference from the environmental monitoring data and more accurately reflects the true state of environmental parameters. The coupling characteristics between process execution data and environmental monitoring data reflect the relationship between environmental parameter fluctuations and process execution stability. The interference intensity indicates the degree to which environmental parameter fluctuations affect process execution stability. A greater interference intensity indicates a more severe impact of environmental parameter fluctuations on process execution stability.

[0057] In smart factories, environmental parameter fluctuations are analyzed based on the noise filtering results of environmental monitoring data. For example, the range and frequency of variation of environmental parameters such as temperature, humidity, and noise are analyzed. Simultaneously, process execution stability indicators, such as fluctuations in process execution speed and changes in process execution accuracy, are observed in the process execution data. Correlation analysis is used to determine the correlation between environmental parameter fluctuations and process execution stability. For example, the correlation coefficient between changes in ambient temperature and fluctuations in process execution speed is calculated. A large correlation coefficient indicates that ambient temperature fluctuations have a significant impact on process execution speed stability. Based on this correlation, the coupling characteristics between process execution data and environmental monitoring data are determined, and the degree of impact of environmental parameter fluctuations on process execution stability is quantified using interference intensity. For example, interference intensity can be defined as the ratio of the standard deviation of environmental parameter fluctuations to the standard deviation of process execution stability indicator fluctuations. In this way, the coupling characteristics between process execution data and environmental monitoring data are determined, providing environmental considerations for subsequent process optimization.

[0058] Step S250: Integrate the dynamic association features and the coupling features to generate a fusion data unit in the fusion data set, and store the fusion data unit in the multimodal fusion database according to a preset fusion data storage format.

[0059] Integration is the process of comprehensively processing and combining dynamic correlation features and coupling features. Through integration, a comprehensive feature representation can be obtained to reflect the complex relationship between equipment operation parameter data, process execution process data and environmental monitoring data. The fusion data unit is the basic data unit in the fusion data set. It contains the dynamic correlation features between equipment operation parameter data and process execution process data, as well as the coupling features between process execution process data and environmental monitoring data. The preset fusion data storage format is a pre-defined format for storing fusion data units, which can ensure the consistency and standardization of data. The multimodal fusion database is a database specifically used to store fusion data sets. It can provide data storage, query and management functions.

[0060] For the integration of dynamic correlation features and coupling features, these features are first standardized so that they have the same dimension and range. Then, these features are integrated using splicing or weighted summation methods. For example, the dynamic correlation features and coupling features are spliced into a vector as a fusion data unit. The fusion data unit is formatted according to the preset fusion data storage format, such as JSON format or CSV format. For JSON format, the fusion data unit can be converted into a JSON object, which contains the various attributes and values of the dynamic correlation features and coupling features. Finally, the formatted fusion data unit is stored in the multimodal fusion database.

[0061] Step S300: performing a process feature extraction operation based on the fused data set to generate a process optimization feature set; the process optimization feature set includes equipment operation stability features, process parameter adaptability features, and environmental interference suppression features.

[0062] The process feature extraction operation is the process of extracting features that reflect process characteristics and optimization requirements from the fused data set. The process optimization feature set is the feature set obtained after the process feature extraction operation. It contains equipment operation stability features, process parameter adaptability features, and environmental interference suppression features. Equipment operation stability features reflect the stability of the equipment during operation, such as fluctuations in equipment parameters and the balance of equipment load. Process parameter adaptability features reflect the degree of adaptation of process parameters to environmental interference, such as the offset of process execution accuracy with changes in environmental parameters and the sensitivity of process parameter adjustments. Environmental interference suppression features represent the measures and parameters required to suppress the impact of environmental interference on process execution, such as the environmental noise suppression strength and the temperature fluctuation compensation coefficient.

[0063] As an implementation manner, step S300 may specifically include the following steps S310 to S340: Step S310: Extract the dynamic correlation features between the equipment operation parameter data and the process execution process data from the fused data set, evaluate the equipment operation status based on the dynamic correlation features, and generate equipment operation stability features; the equipment operation stability features include equipment parameter fluctuation tolerance, equipment load balancing and equipment failure prediction indicators.

[0064] Dynamic correlation features reflect the dynamic relationship between equipment operating parameter data and process execution data. They contain information about the impact of equipment operating status on process execution. Equipment operating status assessment is the process of analyzing and evaluating the equipment operating status based on dynamic correlation features. Equipment operating stability features are a set of features that reflect equipment operating stability. The equipment parameter fluctuation tolerance indicates the range of parameter fluctuations that the equipment can tolerate, the equipment load balance indicates the degree of balance of the equipment load across different components or time periods, and the equipment failure prediction index indicates the likelihood of equipment failure.

[0065] As an implementation manner, step S310 may specifically include the following steps S311 to S316: Step S311: Perform operation mode decomposition processing on the dynamic correlation features to generate a multi-dimensional operation sub-feature set between the equipment operation parameter data and the process execution process data. The multi-dimensional operation sub-feature set includes equipment operation mode features, process parameter coordination features and abnormal fluctuation propagation features.

[0066] The operation mode decomposition process can decompose complex dynamic correlation features into multiple operation sub-features with clear meanings. The multi-dimensional operation sub-feature set is a feature set obtained by the operation mode decomposition process, which includes equipment operation mode features, process parameter coordination features and abnormal fluctuation propagation features. The equipment operation mode features reflect the mode and characteristics of the equipment in different operation stages, such as the equipment's start and stop mode, load change mode, etc. The process parameter coordination features reflect the coordination relationship between different process parameters during the process execution, such as the degree of coordination between the process execution speed and the process pressure. The abnormal fluctuation propagation features indicate the propagation of abnormal fluctuations in equipment operating parameters or process execution, such as the propagation range and intensity of abnormal fluctuations caused by equipment failures during the process execution.

[0067] Specifically, cluster analysis algorithms (such as the K-means clustering algorithm) can be used to cluster dynamic correlation features. First, the number of clusters can be determined. For example, dynamic correlation features can be divided into three categories: equipment operation mode, process parameter coordination, and abnormal fluctuation propagation. Then, the distance between each dynamic correlation feature and each cluster center is calculated and assigned to the cluster with the closest distance. For equipment operation mode features, the cluster is analyzed for features related to the equipment operating status, such as the equipment start and stop time and load change trends, to extract equipment operation mode features. For process parameter coordination features, the cluster is analyzed for features related to the relationship between process parameters, such as the correlation between process execution speed and process pressure, to extract process parameter coordination features. For abnormal fluctuation propagation features, the cluster is analyzed for features related to abnormal fluctuations, such as the starting point, propagation path, and intensity of abnormal fluctuations, to extract abnormal fluctuation propagation features.

[0068] Step S312: Perform historical operation mode matching analysis on the equipment operation mode characteristics to extract the equipment operation cycle characteristics associated with the current process execution process. The equipment operation cycle characteristics include equipment start and stop interval characteristics, load change cycle characteristics and parameter adjustment response delay characteristics.

[0069] Specifically, a historical operating mode database can be established to record the operating mode characteristics of the equipment during different process executions. The current equipment operating mode characteristics are then compared with the patterns in the historical operating mode database, and a similarity calculation method (such as cosine similarity) is used to find the most similar historical operating mode. For the equipment operating cycle characteristics associated with the current process execution, the equipment start-stop interval characteristics, load change cycle characteristics, and parameter adjustment response delay characteristics are extracted from the matching historical operating modes. For example, by analyzing the start-stop time records of the equipment in the historical operating mode, the average value and standard deviation of the equipment start-stop interval are calculated as the equipment start-stop interval characteristics. By analyzing the equipment load change curve in the historical operating mode, the load change cycle is found as the load change cycle characteristics. By analyzing the equipment response time to parameter adjustment in the historical operating mode, the average value of the parameter adjustment response delay is calculated as the parameter adjustment response delay characteristic.

[0070] Step S313: perform parameter synchronization evaluation on the process parameter coordination characteristics to generate a coordination index set of the equipment operation parameters and the process execution parameters. The coordination index set includes the parameter adjustment synchronization rate, the instruction response consistency coefficient and the execution deviation compensation efficiency.

[0071] Specifically, in smart factories, process parameter coordination features are evaluated for parameter synchronization. For parameter adjustment synchronization, the adjustment time and magnitude of equipment operating parameters and process execution parameters are analyzed. The ratio of the number of simultaneous adjustments of equipment operating parameters and process execution parameters to the total number of adjustments is calculated as the parameter adjustment synchronization rate. For execution deviation compensation efficiency, the equipment's compensation for execution deviations during process execution is analyzed. The speed and effectiveness of the equipment's compensation for execution deviations are calculated as the execution deviation compensation efficiency.

[0072] Step S314: Perform fluctuation intensity classification processing on the abnormal fluctuation propagation characteristics to generate the abnormal fluctuation intensity level of equipment operation and the interference level of process execution. The fluctuation intensity level includes the amplitude of equipment parameter mutation, abnormal duration span and fluctuation diffusion range.

[0073] Specifically, the abnormal fluctuation propagation characteristics are graded for fluctuation intensity. The criteria for fluctuation intensity grading are determined. For example, the amplitude of equipment parameter mutations can be categorized into small, medium, and large, corresponding to different amplitude ranges. The duration of abnormality can be categorized into short, medium, and long, corresponding to different time ranges. The range of fluctuation diffusion can be categorized into local, regional, and global, corresponding to different diffusion ranges. Then, based on the amplitude of equipment parameter mutations, the duration of abnormality, and the range of fluctuation diffusion in the abnormal fluctuation propagation characteristics, the abnormal fluctuations are classified into different intensity levels.

[0074] Step S315: Input the equipment operation cycle characteristics, coordination index set and fluctuation intensity level into the equipment health assessment model, and jointly calculate the comprehensive influence weights of the periodic stability of the equipment operation cycle characteristics, the parameter matching degree of the coordination index set and the fluctuation intensity level through the equipment health assessment model.

[0075] Specifically, in smart factories, the equipment health assessment model can adopt a neural network architecture, such as a multilayer perceptron (MLP). This model consists of an input layer, hidden layers, and an output layer. The input layer receives as input data equipment operation cycle characteristics (equipment start-stop interval characteristics, load change cycle characteristics, and parameter adjustment response delay characteristics), a set of coordination indicators (parameter adjustment synchronization rate, command response consistency coefficient, and execution deviation compensation efficiency), and fluctuation intensity levels (equipment parameter mutation amplitude, abnormal duration span, and fluctuation diffusion range). The hidden layer extracts and transforms the input data, mining potential relationships within the data through connections and activation functions between neurons. The output layer outputs the equipment health assessment score. When training the equipment health assessment model, historical equipment operation data is used as the training set. First, the equipment operation cycle characteristics, coordination indicators, and fluctuation intensity levels in the training set are normalized to the same scale and range. Then, each sample is labeled with a device health status label. The backpropagation algorithm is used to adjust the model weights and biases to ensure that the model output is as close to the actual health status label as possible. During the training process, appropriate model parameters, such as the number of neurons in the hidden layer and the learning rate, are selected through methods such as cross-validation to improve the generalization ability of the model.

[0076] Step S316: Generate equipment operation stability characteristics based on the results of the joint calculation. The equipment operation stability characteristics include equipment parameter fluctuation tolerance, equipment load balance, and equipment failure prediction index.

[0077] The tolerance of equipment parameter fluctuations is determined by the ratio of the parameter adjustment synchronization rate to the abnormality duration span. The equipment load balance is determined by the product of the load change cycle characteristics and the instruction response consistency coefficient. The equipment failure prediction index is determined by the inverse proportional relationship between the fluctuation diffusion range and the execution deviation compensation efficiency.

[0078] The result of the joint calculation is the comprehensive score and related weight information obtained by jointly calculating the equipment operation cycle characteristics, coordination index set and fluctuation intensity level through the equipment health assessment model.

[0079] Step S320: Based on the coupling characteristics between the process execution data and the environmental monitoring data, the adaptability analysis of the process parameters and the environmental interference is performed to generate the process parameter adaptability characteristics; the process parameter adaptability characteristics include the offset of the process execution accuracy with the change of environmental parameters, the process parameter adjustment sensitivity and the environmental compensation parameter response efficiency.

[0080] The coupling characteristics between process execution data and environmental monitoring data reflect the impact of environmental parameter fluctuations on process execution. Compatibility analysis of process parameters and environmental disturbances studies how process parameters adapt to environmental disturbances to ensure process execution stability and accuracy. Process parameter compatibility features are a set of features derived from this compatibility analysis and are used to assess the compatibility of process parameters with environmental disturbances.

[0081] As an implementation manner, step S320 may specifically include the following steps S321 to S326: Step S321: performing environmental interference pattern separation processing on the coupling features, extracting the interference pattern feature set associated with the process execution data in the environmental monitoring data, the interference pattern feature set including temperature fluctuation interference features, humidity change interference features and noise propagation interference features.

[0082] In smart factories, coupling features are used to separate environmental interference patterns. Independent component analysis (ICA) algorithms can be used to preprocess the coupling features, such as removing the mean and whitening, to improve the algorithm's efficiency and accuracy. Then, through iterative calculations, a set of separation matrices is found that maximizes the independence of the separated components. During this iterative process, criteria such as maximizing negative entropy are used to measure the independence of the components.

[0083] To extract the temperature fluctuation interference feature, the temperature data in the environmental monitoring data is analyzed, and statistical features such as the temperature variation amplitude and frequency are calculated. For example, the difference between the maximum and minimum temperature values over a period of time is calculated as the temperature variation amplitude, and the number of temperature changes within a set time interval is counted as the temperature variation frequency. For the humidity variation interference feature, the humidity data in the environmental monitoring data is similarly analyzed to extract features such as the humidity increase or decrease trend and the range of variation. For example, by performing time series analysis on the humidity data, it is determined whether the humidity is trending upward or downward; the maximum and minimum humidity values are calculated to determine the humidity variation range. For the noise propagation interference feature, the noise data in the environmental monitoring data is spectrally analyzed to determine the noise intensity and frequency distribution. For example, a fast Fourier transform (FFT) is used to convert the noise signal from the time domain to the frequency domain, and the energy distribution of different frequency components is analyzed to obtain the frequency distribution features of the noise.

[0084] Step S322: performing timing alignment processing on the temperature fluctuation interference feature and the temperature-sensitive process parameters in the process execution data to generate a temperature interference-process parameter timing correlation feature, and analyzing the parameter offset and response delay characteristics of the temperature-sensitive process parameters under temperature fluctuation interference.

[0085] Specifically, first check the timestamps of the temperature fluctuation interference feature and the temperature-sensitive process parameters to ensure they have the same time resolution. If the time resolutions differ, adjust them using interpolation or resampling. For example, if the time interval of the temperature fluctuation interference feature is 1 minute, while the time interval of the temperature-sensitive process parameter is 2 minutes, interpolate the time interval of the temperature-sensitive process parameter to also be 1 minute. Then, associate the temperature fluctuation interference feature with the temperature-sensitive process parameter based on the timestamps one-to-one to generate the temperature interference-process parameter time series correlation feature.

[0086] Analyze the parameter offset and response delay characteristics of temperature-sensitive process parameters under temperature fluctuations. For parameter offset, use the normal operating value of the temperature-sensitive process parameter as a benchmark and calculate the difference between the actual value of the process parameter and the benchmark value when the temperature fluctuation occurs. For response delay, record the time when the temperature fluctuation occurs and the time when the temperature-sensitive process parameter begins to change. Calculate the time difference between the two as the response delay characteristic.

[0087] Step S323: Match the humidity change interference characteristics with the process execution space distribution, identify the coverage relationship between the humidity interference diffusion path and the process execution area, generate the humidity interference-process area impact characteristics, and extract the stability deviation and compensation hysteresis characteristics of the process parameters within the humidity interference diffusion range.

[0088] The humidity disturbance diffusion path can be determined using numerical simulation or experimental measurement. For example, computational fluid dynamics (CFD) software can be used to simulate the diffusion of humidity in space to determine the humidity disturbance diffusion path. The humidity disturbance diffusion path is then matched with the spatial distribution of the process execution area to identify their coverage relationship. For example, the humidity disturbance diffusion path is determined to cover key locations within the process execution area, as well as the area and extent of coverage. A humidity disturbance-process area impact signature is generated, including information such as the intensity and scope of the humidity disturbance's impact on the process execution area. For example, the average humidity change within the process execution area can be calculated as an indicator of the intensity of the impact, and the proportion of the process execution area covered by the humidity disturbance diffusion path can be calculated as an indicator of the scope of the impact. The stability deviation and compensation hysteresis characteristics of the process parameters within the humidity disturbance diffusion range are extracted. For the stability deviation, the process parameters are monitored for changes within the humidity disturbance diffusion range. The degree of deviation under humidity disturbance is calculated based on the normal stable value of the process parameters. For the compensation hysteresis characteristic, the time when the humidity disturbance occurs and the time when the process parameter compensation adjustment begins are recorded, and the time difference between the two is calculated as the compensation hysteresis characteristic.

[0089] Step S324: performing frequency domain correlation analysis on the noise propagation interference characteristics and the vibration parameters of the process execution equipment to generate noise-equipment vibration coupling characteristics, and determining the frequency band offset amplitude and amplitude abnormal growth rate of the vibration parameters affected by noise interference.

[0090] Specifically, the noise propagation interference signature and the vibration parameters of the process equipment are first converted from the time domain to the frequency domain. A common method is the Fast Fourier Transform (FFT). The noise and vibration signals are each subjected to an FFT transform to obtain their spectra. Next, the relationship between the noise and vibration spectra is analyzed. For example, the correlation coefficient between the two is calculated to identify the frequencies at which the correlation is highest. A noise-equipment vibration coupling signature is generated, including information such as the coupling strength between noise and vibration at different frequencies and the coupling frequency range. For example, a high correlation coefficient between the noise and vibration spectra within a certain frequency range indicates a strong coupling relationship between noise and vibration within this frequency range. The frequency band shift amplitude and the abnormal amplitude growth rate of the vibration parameters affected by noise are determined. For the frequency band shift amplitude, the primary frequency components of the vibration parameters before and after the noise interference are compared. For example, in the absence of noise interference, the primary frequency of the vibration parameter is f1. Under noise interference, the primary frequency changes to f2. Therefore, the frequency band shift amplitude is |f2 - f1|. For the abnormal amplitude growth rate, the rate of increase in the vibration parameter amplitude under noise interference is calculated. For example, within a period of time after the noise interference begins, the amplitude of the vibration parameter increases from A1 to A2, and the time interval is t. Then the abnormal amplitude growth rate is (A2-A1)÷t.

[0091] Step S325: Based on the parameter offset, response delay characteristics, stability deviation, compensation hysteresis characteristics, frequency band offset amplitude and amplitude abnormal growth rate, a comprehensive impact index set of process parameters affected by environmental interference is constructed.

[0092] Specifically, the parameter offset, response delay characteristics, stability deviation, compensation lag characteristics, frequency band offset amplitude, and amplitude abnormal growth rate are first standardized to have the same dimension and range. For example, the z-score normalization method is used to convert the data of each indicator into standard normal distribution data with a mean of 0 and a standard deviation of 1.

[0093] Then, assign a weight to each indicator based on the importance of different environmental interference factors to process execution. For example, for temperature fluctuation interference, parameter offset and response delay characteristics may be more important and can be given a higher weight; for humidity change interference, stability deviation and compensation hysteresis characteristics may be more critical and can be given a higher weight.

[0094] Finally, the standardized index value is multiplied by the corresponding weight and then summed to obtain the comprehensive impact index of the process parameters affected by environmental interference.

[0095] Step S326: Evaluate the process parameter adaptability based on the comprehensive influencing indicator set to generate process parameter adaptation characteristics. The process parameter adaptation characteristics include the offset of process execution accuracy with changes in environmental parameters, process parameter adjustment sensitivity, and environmental compensation parameter response efficiency.

[0096] The offset is determined by the weighted sum of the parameter offset and the stability deviation, the adjustment sensitivity is determined by the inverse product of the response delay characteristic and the compensation lag characteristic, and the response efficiency is determined by the ratio of the frequency band offset amplitude to the amplitude abnormality growth rate.

[0097] The offset in process execution accuracy due to changes in environmental parameters is determined by the weighted sum of the parameter offset and the stability deviation. The process parameter adjustment sensitivity is determined by the inverse product of the response delay characteristic and the compensation hysteresis characteristic. Let t1 be the response delay characteristic and t2 be the compensation hysteresis characteristic. Adjustment sensitivity = 1 ÷ (t1 × t2). The shorter the response delay and compensation hysteresis characteristics, the higher the process parameter adjustment sensitivity, indicating that the process parameters can respond more quickly to environmental disturbances and adjust accordingly.

[0098] The response efficiency of the environmental compensation parameter is determined by the ratio of the frequency band offset amplitude to the amplitude anomaly growth rate. Assume the frequency band offset amplitude is f and the amplitude anomaly growth rate is a. The response efficiency = f ÷ a. The smaller the frequency band offset amplitude and the greater the amplitude anomaly growth rate, the lower the environmental compensation parameter response efficiency, indicating that the environmental compensation parameter is less effective in compensating for environmental interference.

[0099] Step S330: Perform environmental interference pattern recognition on the environmental monitoring data to generate environmental interference suppression features; the environmental interference suppression features include environmental noise suppression strength, temperature fluctuation compensation coefficient and humidity change buffer threshold.

[0100] Environmental interference pattern recognition is the process of identifying different types of environmental interference patterns from environmental monitoring data, such as periodic interference patterns, sudden interference patterns, and continuous interference patterns. Environmental interference suppression features are a set of features that must be determined to mitigate the impact of environmental interference on process execution. These include environmental noise suppression strength, temperature fluctuation compensation coefficient, and humidity change buffer threshold.

[0101] As an implementation manner, step S330 may specifically include the following steps S331 to S336: Step S331: Perform interference source separation processing on the environmental monitoring data to extract the independent interference source feature set associated with the process execution process in the environmental monitoring data. The independent interference source feature set includes temperature interference source fluctuation characteristics, humidity interference source diffusion characteristics and noise interference source propagation characteristics.

[0102] Specifically, in smart factories, blind source separation (BSS) algorithms, such as the FastICA algorithm, can be used to isolate interference sources from environmental monitoring data. The basic idea of this algorithm is to isolate independent interference sources from mixed environmental monitoring data without knowing the interference source and mixing matrix. First, the environmental monitoring data is preprocessed, such as by removing the mean and whitening, to improve the algorithm's efficiency and accuracy. Then, through iterative calculations, a set of separation matrices is found that maximizes the independence of the separated signals. During this iterative process, criteria such as maximizing negative entropy are used to measure the independence of the signals.

[0103] To identify the fluctuation characteristics of the extracted temperature interference source, the separated temperature interference source signal is analyzed to calculate statistical characteristics such as the temperature variation amplitude and frequency. For example, the difference between the maximum and minimum temperature values over a period of time is calculated as the temperature variation amplitude, and the number of temperature changes within a set time interval is counted as the temperature variation frequency. To identify the diffusion characteristics of the humidity interference source, the separated humidity interference source signal is analyzed to extract characteristics such as the humidity propagation path and diffusion speed. For example, by performing spatial analysis on the humidity data, the propagation direction and range of the humidity are determined, and the diffusion distance of the humidity within a set time period is calculated as the diffusion speed. To identify the propagation characteristics of the noise interference source, the separated noise interference source signal is analyzed to determine characteristics such as the noise propagation direction and intensity. For example, using microphone array technology, the noise intensity in different directions is measured to determine the noise propagation direction, and the noise sound pressure level is calculated as the propagation intensity.

[0104] Step S332: periodically analyzing the fluctuation characteristics of the temperature interference source, identifying the steady-state fluctuation component and the transient fluctuation component in the fluctuation characteristics of the temperature interference source, and determining the correlation between the duration of the steady-state fluctuation component and the fluctuation amplitude.

[0105] Specifically, in smart factories, the periodicity analysis of the fluctuation characteristics of temperature interference sources can be performed using methods such as Fourier transform (FT) or wavelet transform (WT). The Fourier transform converts the fluctuation characteristics of temperature interference sources from the time domain to the frequency domain, analyzing their frequency components to identify frequency components with significant periodicity. The wavelet transform analyzes the fluctuation characteristics of temperature interference sources in both the time and frequency domains, making it more suitable for processing non-stationary signals.

[0106] First, perform a Fourier transform on the temperature interference source's fluctuation characteristics to obtain its spectrum. Analyze the amplitudes of the various frequency components in the spectrum to identify those with larger amplitudes. The time-domain signals corresponding to these frequency components are the steady-state fluctuation components. Transient fluctuation components can be obtained by subtracting the steady-state fluctuation components from the original temperature interference source's fluctuation characteristics.

[0107] Then, determine the correlation between the duration and amplitude of the steady-state fluctuation component. Statistical analysis methods can be used to calculate the duration and amplitude of the steady-state fluctuation component. For example, divide the duration of the steady-state fluctuation component into different intervals and calculate the mean and standard deviation of the amplitude within each interval. Observe the relationship between duration and amplitude by drawing a scatter plot or fitting a curve. If the longer the duration, the larger the amplitude, indicating a positive correlation between the two; if there is no obvious pattern between duration and amplitude, indicating an insignificant relationship.

[0108] Step S333: Tracing the spatial propagation path of the humidity interference source diffusion feature, generating a humidity diffusion gradient distribution feature corresponding to the humidity interference source diffusion feature, and extracting the diffusion path coverage and humidity concentration accumulation rate overlapping with the process execution area in the humidity diffusion gradient distribution feature.

[0109] Specifically, in smart factories, the diffusion characteristics of humidity interference sources are tracked through spatial propagation paths. This can be achieved through numerical simulation or experimental measurement. Numerical simulation uses computational fluid dynamics (CFD) software to establish a mathematical model of humidity diffusion and simulate the process of humidity propagation in space. Experimental measurement utilizes a humidity sensor network to monitor humidity changes at different locations in real time, and analyze the data to determine the humidity propagation path.

[0110] Generate a humidity diffusion gradient distribution signature corresponding to the humidity interference source diffusion signature. For numerical simulation, calculate the humidity gradient at different locations based on the simulated humidity distribution data to generate the humidity diffusion gradient distribution signature. For experimental measurement, interpolate and differentiate the humidity sensor data to obtain the humidity diffusion gradient distribution signature.

[0111] Extract the diffusion path coverage and humidity concentration accumulation rate of the humidity diffusion gradient distribution feature that overlaps with the process execution area. First, determine the spatial extent of the process execution area. Then, find the portion of the humidity diffusion gradient distribution feature that overlaps with the process execution area and calculate the area or volume of the overlapping portion as the diffusion path coverage. For the humidity concentration accumulation rate, calculate the rate of increase of humidity concentration over time along the diffusion path that overlaps with the process execution area.

[0112] Step S334: Perform frequency domain energy distribution analysis on the noise interference source propagation characteristics, divide the noise interference source propagation characteristics into low-frequency noise energy distribution intervals and high-frequency noise energy distribution intervals, and determine the interference contribution of the noise energy in each distribution interval to the vibration of the process equipment.

[0113] Frequency-domain energy distribution analysis converts the propagation characteristics of a noise source from the time domain to the frequency domain, analyzing the distribution of its energy across different frequencies. The low-frequency noise energy distribution range and the high-frequency noise energy distribution range are divided into two intervals based on the noise's frequency range. Low-frequency noise has a frequency range of 0-1000Hz, while high-frequency noise has a frequency range above 1000Hz.

[0114] To analyze the propagation characteristics of noise interference sources, perform a frequency domain energy distribution analysis. First, perform a Fourier transform on the propagation characteristics of the noise interference source, converting them from the time domain to the frequency domain to obtain the noise spectrum. Then, based on the energy distribution within the spectrum, divide the low-frequency noise energy distribution range into the high-frequency noise energy distribution range. Based on the noise frequency range and energy distribution curve, determine an appropriate frequency threshold to divide the spectrum into low-frequency and high-frequency parts.

[0115] Determine the interference contribution of noise energy within each distribution interval to the vibration of process equipment. This can be determined using either experimental measurement or numerical simulation. The experimental measurement method involves placing vibration and noise sensors around the process equipment to simultaneously measure the noise and equipment vibration signals. Correlation analysis is performed on the measured data to calculate the correlation coefficients between low-frequency and high-frequency noise energy and the equipment vibration signal, respectively, as the interference contribution of noise energy within each distribution interval to the vibration of process equipment. Numerical simulation methods can establish a noise-equipment vibration coupling model to simulate the impact of noise of different frequencies on equipment vibration. The energy transfer efficiency of noise energy within each distribution interval to equipment vibration is then calculated as the interference contribution.

[0116] Step S335: The duration of the steady-state fluctuation component, the correlation between the fluctuation amplitude, the diffusion path coverage, the humidity concentration accumulation rate, the low-frequency noise energy distribution interval and the high-frequency noise energy distribution interval are input into the interference suppression evaluation model, and the temperature compensation requirement coefficient of the temperature interference source fluctuation characteristic, the humidity buffer requirement strength of the humidity interference source diffusion characteristic and the noise suppression requirement level of the noise interference source propagation characteristic are quantitatively calculated through the interference suppression evaluation model.

[0117] Specifically, the interference suppression assessment model can employ an artificial neural network (ANN) model, consisting of an input layer, hidden layers, and an output layer. The input layer receives input data such as the duration of the steady-state fluctuation component, the correlation between fluctuation amplitudes, the coverage of diffusion paths, the accumulation rate of humidity concentration, and the low-frequency and high-frequency noise energy distribution ranges. The hidden layer extracts and transforms features from the input data, mining potential relationships within the data through connections between neurons and activation functions. The output layer outputs the temperature compensation requirement coefficient, humidity buffering requirement strength, and noise suppression requirement level. When training the interference suppression assessment model, historical environmental interference data and the corresponding interference suppression requirement data serve as the training set. First, the input data in the training set is normalized to have the same scale and range. Then, each sample is labeled with the true value of the temperature compensation requirement coefficient, humidity buffering requirement strength, and noise suppression requirement level. The backpropagation algorithm is used to adjust the model weights and biases so that the model output is as close to the true value as possible. During the training process, appropriate model parameters, such as the number of neurons in the hidden layer and the learning rate, are selected through methods such as cross-validation to improve the model's generalization ability.

[0118] Step S336: Generate environmental interference suppression features based on the quantization calculation results. The environmental interference suppression features include environmental noise suppression strength, temperature fluctuation compensation coefficient, and humidity change buffer threshold.

[0119] Among them, the environmental noise suppression intensity is determined by the sum of the interference contributions of the low-frequency noise energy distribution interval and the high-frequency noise energy distribution interval, the temperature fluctuation compensation coefficient is determined by the product of the duration of the steady-state fluctuation component and the correlation between the fluctuation amplitude, and the humidity change buffer threshold is determined by the ratio of the diffusion path coverage range to the humidity concentration accumulation rate.

[0120] Step S340: Unify the characteristic dimensions of the equipment operation stability characteristics, process parameter adaptation characteristics, and environmental interference suppression characteristics to generate a process optimization feature set with the same characteristic dimensions, and input the process optimization feature set into the process optimization strategy generation module.

[0121] Specifically, the equipment operation stability features, process parameter adaptability features, and environmental interference suppression features are first standardized to have the same scale and range. Then, based on their importance and relevance, features are selected and combined, removing redundant and irrelevant features to ensure they have the same feature dimension.

[0122] After generating a set of process optimization features with the same feature dimension, it is input into the process optimization strategy generation module. The process optimization strategy generation module can use methods such as rule reasoning and machine learning to generate a process parameter optimization strategy based on the process optimization feature set. For example, for the equipment operation stability feature, if the tolerance for equipment parameter fluctuations is low, it means that the equipment operation is unstable. The process optimization strategy generation module can generate strategies such as adjusting equipment operation parameters and strengthening equipment maintenance. For the process parameter adaptability feature, if the offset of process execution accuracy with changes in environmental parameters is large, it means that the process parameters are poorly adaptable to environmental interference. The process optimization strategy generation module can generate strategies such as adjusting process parameters and adding environmental compensation measures. For the environmental interference suppression feature, if the environmental noise suppression intensity is high, it means that the noise interference is serious. The process optimization strategy generation module can generate strategies such as installing sound insulation equipment and optimizing equipment layout.

[0123] Step S400: Perform dynamic optimization strategy matching processing according to the process optimization feature set to generate a process parameter optimization strategy; the process parameter optimization strategy includes the equipment operation parameter adjustment direction, the process execution parameter correction direction and the environment compensation parameter adaptation direction.

[0124] Dynamic optimization strategy matching is the process of matching the most appropriate optimization strategy from a preset optimization strategy library based on a set of process optimization features. Process parameter optimization strategies are a series of strategies developed to optimize the process execution process of smart factories, including the direction of equipment operating parameter adjustment, the direction of process execution parameter correction, and the direction of environmental compensation parameter adaptation. The equipment operating parameter adjustment direction refers to the direction and magnitude of adjustments to the equipment's operating parameters (such as speed, temperature, and pressure); the process execution parameter correction direction refers to the direction and magnitude of corrections to parameters during process execution (such as process execution speed, process pressure, and process time); and the environmental compensation parameter adaptation direction refers to the direction and magnitude of adaptive adjustments to environmental compensation parameters (such as temperature compensation parameters, humidity buffer parameters, and noise suppression parameters).

[0125] As an implementation manner, step S400 may specifically include the following steps S410 to S450: Step S410: Evaluate the health of the equipment operation based on the equipment operation stability characteristics and determine the equipment operation parameter adjustment direction; the equipment operation parameter adjustment direction includes the equipment load balancing adjustment amount, the equipment operation rate correction value and the equipment maintenance cycle optimization suggestion.

[0126] Specifically, in smart factories, equipment operational health is assessed based on operational stability characteristics. A multi-index comprehensive assessment approach can be used to comprehensively consider the equipment parameter fluctuation tolerance, equipment load balance, and equipment failure prediction indicators within these characteristics. For example, different weights can be assigned to each of these indicators to calculate the equipment operational health score. The weights can be adjusted based on the equipment type, process requirements, and historical data.

[0127] Based on the results of the equipment health assessment, the direction of equipment operating parameter adjustments is determined. Regarding equipment load balancing adjustments, if the equipment load balancing is low, it indicates that the load distribution among the equipment components is uneven. It is necessary to reduce the load on the components with higher loads and increase the load on the components with lower loads. Equipment load balancing can be achieved by adjusting the equipment's workload distribution and optimizing the equipment's operational processes. For example, if a motor in a piece of equipment is overloaded while other motors are underloaded, some workload can be shifted from that motor to the other motors.

[0128] If the equipment's operating speed is unstable or unreasonable, the equipment's operating speed needs to be corrected. This can be done by adjusting the speed of the equipment's drive motor and controlling the equipment's acceleration and deceleration times based on the equipment's process requirements and operating conditions. For example, if the equipment's operating speed is too fast, resulting in reduced product quality, the operating speed can be appropriately reduced; if the equipment's operating speed is too slow, resulting in low production efficiency, the operating speed can be appropriately increased.

[0129] Step S420: performing process execution accuracy optimization analysis based on the process parameter adaptability characteristics to determine the process execution parameter correction direction; the process execution parameter correction direction includes the process execution speed adjustment amount, the process parameter fault tolerance threshold correction value and the process execution path optimization coefficient.

[0130] Specifically, in smart factories, process execution accuracy optimization analysis is performed based on the process parameter adaptation characteristics. The process parameter adaptation characteristics include the offset of process execution accuracy with changes in environmental parameters, the sensitivity of process parameter adjustment, and the response efficiency of environmental compensation parameters. The relationship between these characteristics and process execution accuracy is analyzed to find out the key factors affecting process execution accuracy. For example, if the offset of process execution accuracy with changes in environmental parameters is large, it means that the environmental parameters have a greater impact on process execution accuracy, and corresponding measures need to be taken to reduce environmental interference; if the process parameter adjustment sensitivity is low, it means that the process parameters have a weak response to environmental changes, and the adjustment sensitivity of the process parameters needs to be improved.

[0131] Based on the results of the process execution accuracy optimization analysis, determine the direction for process execution parameter correction. Regarding process execution speed adjustments, if the process execution speed is too fast or too slow, it will affect process execution accuracy. Therefore, the process execution speed needs to be adjusted based on the process requirements and actual conditions. For example, if the process execution accuracy requirements are high and the current process execution speed is too fast, resulting in unstable product quality, the process execution speed can be appropriately reduced; if the process execution speed is too slow, resulting in low production efficiency, the process execution speed can be appropriately increased.

[0132] Regarding the process parameter fault tolerance threshold correction value, if the process parameter fault tolerance threshold is set improperly, it will lead to excessive waste or defective products during process execution. The process parameter fault tolerance threshold needs to be corrected according to the process requirements and actual conditions. Regarding the process execution path optimization coefficient, if the process execution path is unreasonable, it will lead to problems such as excessive process execution time and excessive energy consumption. The process execution path needs to be optimized according to the process requirements and actual conditions. For example, a path planning algorithm (such as the A* algorithm, the Dijkstra algorithm, etc.) can be used to replan the process execution path, find the optimal process execution path, and calculate the corresponding process execution path optimization coefficient.

[0133] Step S430: Performing environmental compensation parameter adaptation analysis based on the environmental interference suppression characteristics to determine the environmental compensation parameter adaptation direction; the environmental compensation parameter adaptation direction includes the temperature compensation parameter adjustment amount, the humidity buffer parameter correction value, and the noise suppression strength optimization coefficient.

[0134] Specifically, in smart factories, environmental compensation parameter adaptation analysis is performed based on the environmental interference suppression characteristics. Environmental interference suppression characteristics include environmental noise suppression strength, temperature fluctuation compensation coefficient, and humidity change buffer threshold. The relationship between these characteristics and environmental compensation parameters is analyzed to identify the key factors affecting the effectiveness of environmental compensation. For example, if the environmental noise suppression strength is high, it means that the noise interference is serious and the effectiveness of noise suppression measures needs to be improved; if the temperature fluctuation compensation coefficient is large, it means that temperature fluctuations have a significant impact on the process execution process and the intensity of temperature compensation needs to be increased; if the humidity change buffer threshold is small, it means that humidity changes have a significant impact on the process execution process and humidity buffer measures need to be strengthened.

[0135] Based on the results of the environmental compensation parameter adaptation analysis, determine the direction of environmental compensation parameter adaptation. Regarding the temperature compensation parameter adjustment amount, if temperature fluctuations significantly impact process execution, adjust the temperature compensation parameters based on the temperature fluctuation compensation coefficient. For example, if the temperature fluctuation compensation coefficient indicates that the temperature compensation amount needs to be increased, increase the power of the heating equipment or reduce the power of the cooling equipment to increase the temperature compensation strength.

[0136] For humidity buffer parameter corrections, if humidity changes significantly impact process execution, the humidity buffer parameter needs to be adjusted based on the humidity change buffer threshold. For example, if the humidity change buffer threshold is low, it indicates that humidity buffer measures need to be strengthened. This can be achieved by increasing the operating time of the dehumidifier or increasing the power of the humidifier to improve the humidity buffer effect.

[0137] Regarding the noise suppression intensity optimization coefficient, if the ambient noise suppression intensity is high, it means that the current noise suppression measures are ineffective and need to be optimized based on the actual situation. For example, the thickness of the sound insulation material can be increased, the layout of the sound insulation equipment can be optimized, or more efficient noise reduction equipment can be replaced to improve the noise suppression effect.

[0138] Step S440: sorting the equipment operation parameter adjustment direction, process execution parameter correction direction and environment compensation parameter adaptation direction by strategy priority to generate a strategy execution sequence.

[0139] For example, multi-attribute decision-making methods such as the analytic hierarchy process (AHP) or fuzzy comprehensive evaluation can be employed. First, factors influencing the strategy's priority are determined, such as the impact on process execution accuracy, equipment operational stability, environmental interference suppression, implementation cost, and implementation difficulty. Each factor is then assigned a specific weight, which can be adjusted based on process requirements, equipment characteristics, and historical experience.

[0140] Each strategy (device operating parameter adjustment, process execution parameter correction, and environmental compensation parameter adaptation) is scored based on its impact on each factor. For example, if the equipment operating parameter adjustment has a significant impact on equipment operational stability but moderate implementation cost and difficulty, a higher score is assigned for equipment operational stability and a moderate score for implementation cost and difficulty. Next, a comprehensive score is calculated for each strategy based on its scores on each factor and the weights assigned to each factor. A higher comprehensive score indicates a higher priority for the strategy. For example, using the Analytic Hierarchy Process (AHP), a judgment matrix is constructed and eigenvectors are calculated to determine the comprehensive score for each strategy. The comprehensive scores for the equipment operating parameter adjustment, process execution parameter correction, and environmental compensation parameter adaptation directions are ranked based on the comprehensive scores to generate a strategy execution sequence.

[0141] Step S450: constructing a process parameter optimization strategy based on the strategy execution sequence, and mapping the process parameter optimization strategy to the strategy execution interface of the process control system of the smart factory.

[0142] Building a process parameter optimization strategy based on a strategy execution sequence involves integrating and refining the equipment operating parameter adjustment direction, process execution parameter correction direction, and environmental compensation parameter adaptation direction according to the strategy execution sequence to form a complete and executable process parameter optimization strategy. The process control system's strategy execution interface is an interface provided by the smart factory process control system for receiving and executing process parameter optimization strategies. By mapping the process parameter optimization strategy to this interface, automatic adjustment and optimization of process parameters can be achieved.

[0143] In a smart factory, the direction of equipment operating parameter adjustment, process execution parameter correction, and environmental compensation parameter adaptation is refined and integrated according to the strategy execution sequence. For equipment operating parameter adjustment, the specific adjustment parameters and adjustment ranges are clearly defined. For example, if the equipment operating parameter adjustment direction includes equipment load balancing adjustment, equipment operating rate correction value, and equipment maintenance cycle optimization suggestions, it is necessary to determine the specific equipment loads to be adjusted, the specific adjustment values, the equipment operating rate correction value, and the optimized equipment maintenance cycle duration.

[0144] The direction of process execution parameter correction also requires specific correction parameters and correction ranges. For example, the process execution speed adjustment should be increased or decreased, the process parameter tolerance threshold correction value should be increased or decreased, and the specific application of the process execution path optimization coefficient to the process execution path adjustment should be determined.

[0145] For environmental compensation parameter adaptation, determine the specific compensation parameter adjustment values and how to apply the optimization coefficients. For example, the temperature compensation parameter adjustment amount should be increased or decreased by a specific amount, the humidity buffer parameter correction value should be increased or decreased by a specific amount, and how the noise suppression strength optimization coefficient should be used to adjust the power or layout of the sound insulation equipment.

[0146] The refined and integrated equipment operating parameter adjustment directions, process execution parameter correction directions, and environmental compensation parameter adaptation directions are combined into a complete process parameter optimization strategy. This strategy includes detailed information such as adjustment parameters, adjustment ranges, and execution sequence, and is an actionable process optimization solution. The process parameter optimization strategy is mapped to the policy execution interface of the smart factory's process control system. Different process control systems may have different policy execution interfaces, such as network interfaces (such as HTTP and MQTT) and database interfaces.

[0147] Step S500: Execute the process parameter optimization strategy, optimize and adjust the process execution process of the smart factory, and monitor the optimized process execution status data in real time to trigger the strategy iterative update operation.

[0148] Optimizing and adjusting the process execution of a smart factory involves implementing process parameter optimization strategies to improve process execution stability, accuracy, and efficiency. Real-time monitoring of optimized process execution status data involves the continuous collection and analysis of process-related data, such as equipment operating status data, process execution accuracy data, and environmental control response data. Triggering a strategy iteration update operation involves re-executing multimodal fusion processing and subsequent steps to generate an updated process parameter optimization strategy when the monitored process execution status data indicates that the process optimization effect has not met expectations.

[0149] As an implementation manner, step S500 may specifically include the following steps S510 to S550: Step S510: The equipment operating parameter adjustment direction in the process parameter optimization strategy is sent to the equipment control unit through the strategy execution interface of the process control system, triggering a real-time adjustment operation of the equipment operating parameters.

[0150] The process control system's strategy execution interface is the interface through which the process control system exchanges data with external systems. Through this interface, information from the process parameter optimization strategy is transmitted to the corresponding execution unit. The equipment control unit is responsible for controlling equipment operating parameters. It receives instructions from the process control system and adjusts the equipment's operating parameters accordingly. Real-time equipment operating parameter adjustment is the process of adjusting equipment operating parameters immediately based on received instructions during process execution.

[0151] In smart factories, the equipment operating parameter adjustment directions in process parameter optimization strategies include specific adjustment parameters and adjustment ranges. For example, these adjustments might include adjusting equipment speed, temperature, pressure, and other parameters. These adjustment directions are sent to the equipment control unit in a predefined data format (such as JSON or XML) via the process control system's policy execution interface. Upon receiving the instructions, the equipment control unit parses the adjustment parameters and adjustment ranges contained in the instructions and makes real-time adjustments to the equipment's operating parameters based on the device's control logic.

[0152] Step S520: Send the process execution parameter correction direction to the process execution unit to trigger the dynamic correction operation of the process execution parameter.

[0153] The process execution unit is responsible for executing the process, performing operations according to preset process parameters. The process execution parameter correction direction is used to modify parameters during the process to optimize the process execution results. Dynamic process execution parameter correction is the process of adjusting process execution parameters in real time based on received correction directions during process execution.

[0154] In smart factories, the process parameter optimization strategy includes specific correction parameters and their magnitudes. For example, these corrections can include adjustments to process speed, pressure, and time. These corrections are sent to the process execution unit via the process control system's strategy execution interface. Upon receiving these instructions, the process execution unit dynamically adjusts the process parameters based on the corrections.

[0155] Step S530: Send the environmental compensation parameter adaptation formula downward to the environmental control unit to trigger the environmental compensation parameter adaptive adjustment operation.

[0156] The environmental control unit is responsible for regulating the smart factory's environmental parameters, adjusting them based on received instructions. Environmental compensation parameter adaptation targets the impact of environmental interference on process execution. Adaptive adjustment of environmental compensation parameters involves automatically adjusting them based on received adaptation directions during process execution.

[0157] In smart factories, the environmental compensation parameter adaptation directions within process parameter optimization strategies include specific adaptation parameters and adjustment ranges. For example, these include adjusting temperature compensation parameters, humidity buffer parameters, and noise suppression parameters. These adaptation directions are sent to the environmental control unit via the process control system's strategy execution interface. Upon receiving the instructions, the environmental control unit adaptively adjusts the environmental compensation parameters based on the adaptation directions.

[0158] Step S540: Collect the optimized and adjusted equipment operation status data, process execution accuracy data and environmental control response data in real time to generate a strategy execution effect monitoring data set.

[0159] The equipment operating status data after optimization and adjustment refers to the actual operating status data of the equipment after the process parameter optimization strategy is executed, such as the equipment's speed, temperature, pressure, vibration and other parameters. The process execution accuracy data refers to the quality-related data of the product during the process execution, such as the product's dimensional accuracy, shape accuracy, surface roughness and other parameters. The environmental control response data refers to the actual changes in environmental parameters after the environmental compensation parameters are adjusted, such as the change values of environmental parameters such as temperature, humidity, and noise. The strategy execution effect monitoring data set is a data set composed of the equipment operating status data after optimization and adjustment, the process execution accuracy data and the environmental control response data, and is used to evaluate the execution effect of the process parameter optimization strategy.

[0160] In smart factories, various sensors installed on equipment collect real-time data on optimized and adjusted equipment operating status. For example, speed sensors collect equipment speed, temperature sensors collect equipment temperature, and pressure sensors collect equipment pressure. These sensors transmit the collected data to the data acquisition system in real time.

[0161] Process execution accuracy data is collected through quality inspection equipment or process monitoring systems. For example, a coordinate measuring machine (CMM) can be used to measure product dimensional accuracy, while a surface roughness tester can be used to measure surface roughness. The quality inspection equipment transmits the detected data to the data acquisition system in real time.

[0162] Environmental control response data is collected through environmental monitoring equipment. For example, temperature sensors collect ambient temperature, humidity sensors collect ambient humidity, and noise sensors collect ambient noise. The environmental monitoring equipment transmits the collected data to the data acquisition system in real time.

[0163] The data acquisition system integrates and processes the collected optimized and adjusted equipment operating status data, process execution accuracy data, and environmental control response data to generate a data set for monitoring the effectiveness of strategy execution. This data set can be stored in a database, file, or other format for subsequent analysis and evaluation.

[0164] Step S550: Evaluate the strategy effect on the strategy execution effect monitoring data set and generate a strategy optimization index; if the strategy optimization index is lower than the preset optimization threshold, trigger the strategy iteration update operation, re-execute the multimodal fusion processing and subsequent steps to generate an updated process parameter optimization strategy.

[0165] As an implementation manner, step S550 may specifically include the following steps S551 to S555: Step S551: extracting device operation status data from the strategy execution effect monitoring data set, and calculating the device load balancing deviation, device operation rate stability coefficient, and device failure occurrence rate change after the device operation parameters are adjusted.

[0166] Specifically, in smart factories, equipment operating status data is extracted from the policy execution effect monitoring data set. This data may include parameters such as the load current, power consumption, and rotational speed of each equipment component. To determine the equipment load balancing deviation, first calculate the average load of each equipment component. For example, for an equipment system consisting of multiple motors, calculate the load current of each motor and find the average load current of all motors. Then, calculate the absolute value of the difference between the load current of each motor and the average load current, sum the absolute values of these differences, and divide them by the number of motors to obtain the equipment load balancing deviation. The smaller the deviation, the more balanced the equipment load distribution.

[0167] For the device operating rate stability coefficient, calculate the standard deviation of the device operating rate. The standard deviation reflects the degree of data dispersion; smaller standard deviations indicate more stable device operating rates. For example, collect device operating rate data over a period of time and calculate the standard deviation of this data to use as the device operating rate stability coefficient.

[0168] For the change in equipment failure rate, the number of equipment failures in a period of time before and after the execution of the process parameter optimization strategy is counted.

[0169] Step S552: extracting process execution accuracy data from the strategy execution effect monitoring data set, and calculating the process execution speed fluctuation rate, process parameter fault tolerance threshold adaptability, and process execution path optimization efficiency after the process execution parameters are corrected.

[0170] Process execution accuracy data contains information related to product quality during process execution. Process execution speed fluctuation rate measures the degree of fluctuation in process execution speed and reflects the stability of process speed during execution. Process parameter tolerance threshold adaptability measures the degree of match between the process parameter tolerance threshold and actual process execution, reflecting the adaptability of process parameters to certain fluctuations. Process execution path optimization efficiency measures the effectiveness of process execution path optimization and reflects the effectiveness of the optimized process execution path in reducing time and costs.

[0171] In smart factories, process execution accuracy data is extracted from the strategy execution effect monitoring data set. This data may include parameters such as product dimensional accuracy, shape accuracy, surface roughness, as well as information such as process execution speed and process time. For process execution speed fluctuation, the ratio of the standard deviation of the process execution speed to the average speed is calculated. For example, the process execution speed data over a period of time is collected and the standard deviation of this data is calculated. and average , then the process execution speed fluctuation rate is The smaller the fluctuation rate, the more stable the process execution speed.

[0172] For process parameter tolerance threshold adaptability, analyze the relationship between process parameter fluctuations during actual execution and the tolerance threshold. For example, for a specific process parameter, calculate its actual fluctuation range over a period of time and compare it with the preset tolerance threshold. If the actual fluctuation range is mostly within the tolerance threshold, it indicates that the process parameter tolerance threshold adaptability is high; if the actual fluctuation range frequently exceeds the tolerance threshold, it indicates that the process parameter tolerance threshold needs to be adjusted. The process parameter tolerance threshold adaptability can be quantified by calculating the proportion of the actual fluctuation range that is within the tolerance threshold.

[0173] For the optimization efficiency of process execution path, compare the time and cost of process execution before and after optimization. Assume that the time of process execution before optimization is , the cost is ; The time for the optimized process execution is , the cost is Then the process execution path optimization efficiency can be optimized by time optimization efficiency and cost-optimization efficiency The higher the efficiency value, the better the process execution path optimization effect.

[0174] Step S553: extracting environmental control response data from the strategy execution effect monitoring data set, and calculating the temperature fluctuation suppression rate, humidity buffer response time and noise suppression effect improvement after the environmental compensation parameters are adjusted.

[0175] Environmental control response data includes the actual changes in environmental parameters after adjusting the environmental compensation parameters. Temperature fluctuation suppression rate measures the degree to which ambient temperature fluctuations are suppressed, reflecting the improvement in temperature stability achieved by adjusting the temperature compensation parameters. Humidity buffer response time measures the time it takes for ambient humidity to return to a stable state after a disturbance, reflecting the responsiveness of humidity buffer parameter adjustments. Noise suppression improvement measures the degree to which ambient noise is suppressed, reflecting the effectiveness of noise suppression parameter adjustments in reducing noise levels.

[0176] In smart factories, environmental control response data is extracted from the strategy execution effect monitoring data set. This data may include real-time measurements of parameters such as ambient temperature, humidity, and noise. For the temperature fluctuation suppression rate, the temperature fluctuation amplitude before and after the environmental compensation parameter adjustment is first calculated. Assume that the temperature fluctuation amplitude before adjustment is , the adjusted temperature fluctuation range is , then the temperature fluctuation suppression rate is The higher the suppression rate, the better the effect of temperature compensation parameter adjustment on suppressing temperature fluctuations.

[0177] The humidity buffer response time measures the time it takes for the humidity to return to the preset stable value after a disturbance (such as a sudden increase or decrease). For example, if the humidity suddenly increases, the humidity buffer device is activated, and the time from the humidity starting to change to the humidity returning to a stable value is recorded as the humidity buffer response time. The shorter the response time, the faster the humidity buffer parameter adjustment.

[0178] Step S554: construct a strategy execution effect evaluation matrix based on the equipment load balancing deviation, process execution speed fluctuation rate, and temperature fluctuation suppression rate.

[0179] The strategy execution effectiveness evaluation matrix is used to comprehensively assess the effectiveness of process parameter optimization strategies. It integrates multiple evaluation indicators, such as equipment load balancing deviation, process execution speed fluctuation, and temperature fluctuation suppression, to more comprehensively analyze strategy execution effectiveness. In a smart factory, the strategy execution effectiveness evaluation matrix is constructed based on the equipment load balancing deviation, process execution speed fluctuation, and temperature fluctuation suppression calculated in steps S551, S552, and S553. Each row of the matrix represents an evaluation indicator, and each column represents a different process execution stage or time period.

[0180] Step S555: Perform weighted calculation on each evaluation dimension in the strategy execution effect evaluation matrix through the preset weight allocation strategy to generate a comprehensive strategy optimization index; compare the comprehensive strategy optimization index with the preset optimization threshold, and if it is lower than the optimization threshold, generate a strategy iteration trigger signal.

[0181] The preset weight allocation strategy is a pre-set strategy for allocating weights to each evaluation dimension in the strategy execution effect evaluation matrix. These weights reflect the importance of each evaluation dimension in the comprehensive evaluation.

[0182] For example, according to the preset weight allocation strategy, weights are assigned to evaluation dimensions such as equipment load balancing deviation, process execution speed fluctuation rate, and temperature fluctuation suppression rate. For each column of data in the strategy execution effect evaluation matrix, the value of each evaluation dimension is multiplied by the corresponding weight, and then summed to obtain the comprehensive strategy optimization index corresponding to the column. The calculated comprehensive strategy optimization index is compared with the preset optimization threshold. If the comprehensive strategy optimization index is lower than the preset optimization threshold, it means that the execution effect of the process parameter optimization strategy has not met expectations and needs to be optimized and adjusted. At this time, a strategy iteration trigger signal is generated. After receiving the signal, the system re-executes the multimodal fusion processing and subsequent steps, collects new multimodal data, and generates an updated process parameter optimization strategy to further improve the effect and quality of process execution.

[0183] It is understandable that the various algorithms involved in the above-mentioned introductions to the embodiments of the present invention can be learned from the relevant content in the prior art. In order to save space, they will not be expanded too much in the embodiments of this application. In addition, when implementing the scheme of this application, those skilled in the art can supplement the details according to the common knowledge in this field. For example, according to the common knowledge in this field, normalization can be used to eliminate dimensional conflicts before feature fusion, interpolation can be used to eliminate dimensional differences, and historical data, experience or business scenario requirements can be combined to reasonably set thresholds, and models can be trained based on general model training methods, etc. This application will no longer provide redundant introductions to the overly detailed implementation process.

[0184] See also Figure 2 , Figure 2 A structural diagram of a computer system provided in an embodiment of the present invention, the computer system at least includes a processor 101, a communication interface 102 and a memory 103. The processor 101, the communication interface 102 and the memory 103 can be connected via a bus or other means. The processor 101 (or central processing unit (CPU)) is the computing core and control core of the computer system, which can parse various instructions in the computer system and process various data of the computer system. The memory 103 (Memory) is a memory device in the computer system for storing programs and data. The memory 103 provides a storage space, which stores the operating system of the computer system, and the present invention does not limit this. In one embodiment, the processor 101 executes the intelligent factory process optimization method based on multimodal data fusion provided above in the embodiment of the present invention by running the computer program in the memory 103.

Claims

1. A smart factory process optimization method based on multimodal data fusion, characterized in that: include: Acquire a multimodal data set generated during the production process of the smart factory, wherein the multimodal data set includes equipment operating parameter data, process execution process data, and environmental monitoring data; Performing multimodal fusion processing on the multimodal data set to generate a fused data set; the fused data units in the fused data set include dynamic association features between equipment operation parameter data and process execution process data, and coupling features between process execution process data and environmental monitoring data; Performing a process feature extraction operation based on the fused data set to generate a process optimization feature set; The process optimization feature set includes equipment operation stability features, process parameter adaptability features and environmental interference suppression features; Perform dynamic optimization strategy matching processing according to the process optimization feature set to generate a process parameter optimization strategy; The process parameter optimization strategy includes the direction of equipment operation parameter adjustment, process execution parameter correction and environmental compensation parameter adaptation; The process parameter optimization strategy is executed to optimize and adjust the process execution process of the smart factory, and the optimized process execution status data is monitored in real time to trigger the strategy iterative update operation.

2. The method according to claim 1, wherein The performing multimodal fusion processing on the multimodal data set to generate a fused data set includes: Performing timestamp alignment processing on the equipment operation parameter data to generate a time series data sequence; performing spatial registration processing on the process execution data to generate a spatial data sequence; performing noise filtering processing on the environmental monitoring data to generate an environmental data sequence; Inputting the time series data sequence, the spatial data sequence, and the environmental data sequence into a preconfigured multimodal fusion model, and performing cross-modal correlation analysis on the time series fluctuation characteristics of the time series data sequence, the spatial distribution characteristics of the spatial data sequence, and the environmental disturbance characteristics of the environmental data sequence through the multimodal fusion model; Based on the results of the cross-modal correlation analysis, constructing a dynamic correlation feature between the equipment operating parameter data and the process execution process data, wherein the dynamic correlation feature represents the influence weight of the equipment operating parameter change on the process execution accuracy; determining, based on a noise filtering result of the environmental monitoring data, a coupling feature between the process execution data and the environmental monitoring data, wherein the coupling feature reflects the interference intensity of environmental parameter fluctuations on process execution stability; The dynamic association feature and the coupling feature are integrated to generate a fusion data unit in the fusion data set, and the fusion data unit is stored in a multimodal fusion database according to a preset fusion data storage format.

3. The method according to claim 1, wherein The performing of a process feature extraction operation based on the fused data set to generate a process optimization feature set includes: Extracting dynamic correlation features between equipment operating parameter data and process execution process data from the fused data set, evaluating equipment operating status based on the dynamic correlation features, and generating equipment operating stability features; the equipment operating stability features include equipment parameter fluctuation tolerance, equipment load balance, and equipment failure prediction indicators; Based on the coupling characteristics between the process execution data and the environmental monitoring data, a compatibility analysis of the process parameters and the environmental interference is performed to generate a process parameter compatibility characteristic; the process parameter compatibility characteristic includes an offset of the process execution accuracy with changes in the environmental parameters, a sensitivity of the process parameter adjustment, and a response efficiency of the environmental compensation parameter; Performing environmental interference pattern recognition on the environmental monitoring data to generate environmental interference suppression features; the environmental interference suppression features include environmental noise suppression strength, temperature fluctuation compensation coefficient, and humidity change buffer threshold; The equipment operation stability characteristics, the process parameter adaptability characteristics and the environmental interference suppression characteristics are unified in feature dimension to generate a process optimization feature set with the same feature dimension, and the process optimization feature set is input into a process optimization strategy generation module.

4. The method according to claim 3, wherein The performing dynamic optimization strategy matching processing according to the process optimization feature set to generate a process parameter optimization strategy includes: Conducting an equipment operation health assessment based on the equipment operation stability characteristics to determine the equipment operation parameter adjustment direction; the equipment operation parameter adjustment direction includes the equipment load balancing adjustment amount, the equipment operation rate correction value, and equipment maintenance cycle optimization suggestions; Performing process execution accuracy optimization analysis based on the process parameter adaptability characteristics to determine the process execution parameter correction direction; the process execution parameter correction direction includes the process execution speed adjustment amount, the process parameter fault tolerance threshold correction value, and the process execution path optimization coefficient; Performing environmental compensation parameter adaptation analysis based on the environmental interference suppression characteristics to determine the environmental compensation parameter adaptation direction; the environmental compensation parameter adaptation direction includes a temperature compensation parameter adjustment amount, a humidity buffer parameter correction value, and a noise suppression strength optimization coefficient; Prioritize the equipment operation parameter adjustment direction, the process execution parameter correction direction, and the environmental compensation parameter adaptation direction to generate a strategy execution sequence; The process parameter optimization strategy is constructed based on the strategy execution sequence, and the process parameter optimization strategy is mapped to the strategy execution interface of the process control system of the smart factory.

5. The method according to claim 4, wherein The process parameter optimization strategy is executed to optimize and adjust the process execution process of the smart factory, and the optimized process execution status data is monitored in real time to trigger the strategy iteration update operation, including: The equipment operating parameter adjustment direction in the process parameter optimization strategy is sent to the equipment control unit through the strategy execution interface of the process control system, triggering the real-time adjustment operation of the equipment operating parameters; Sending the process execution parameter correction direction to the process execution unit to trigger a dynamic correction operation of the process execution parameter; Sending the environmental compensation parameter adaptation formula downward to the environmental control unit to trigger the environmental compensation parameter adaptive adjustment operation; Collect optimized and adjusted equipment operation status data, process execution accuracy data, and environmental control response data in real time to generate a set of strategy execution effect monitoring data; A strategy effect evaluation is performed on the strategy execution effect monitoring data set to generate a strategy optimization index; if the strategy optimization index is lower than a preset optimization threshold, a strategy iteration update operation is triggered, and the multimodal fusion processing and subsequent steps are re-executed to generate an updated process parameter optimization strategy.

6. The method according to claim 5, wherein The step of performing a strategy effect evaluation on the strategy execution effect monitoring data set to generate a strategy optimization index includes: Extracting device operation status data from the strategy execution effect monitoring data set, and calculating the device load balancing deviation, device operation rate stability coefficient, and device failure rate change after the device operation parameters are adjusted; Extracting process execution accuracy data from the strategy execution effect monitoring data set, and calculating the process execution speed fluctuation rate, process parameter fault tolerance threshold adaptability, and process execution path optimization efficiency after the process execution parameters are corrected; Extracting environmental control response data from the strategy execution effect monitoring data set, and calculating the temperature fluctuation suppression rate, humidity buffer response time, and noise suppression effect improvement after the environmental compensation parameters are adjusted; Constructing a strategy execution effect evaluation matrix based on the equipment load balancing deviation, the process execution speed fluctuation rate, and the temperature fluctuation suppression rate; A weighted calculation is performed on each evaluation dimension in the strategy execution effect evaluation matrix through a preset weight allocation strategy to generate a comprehensive strategy optimization index; the comprehensive strategy optimization index is compared with the preset optimization threshold, and if it is lower than the optimization threshold, a strategy iteration trigger signal is generated.

7. The method according to claim 1, wherein The obtaining of a multimodal data set generated during the production process of the smart factory includes: The sensor network deployed at the equipment layer of the smart factory collects raw equipment operating parameter data, performs data integrity verification and outlier removal on the raw equipment operating parameter data, and generates standardized equipment operating parameter data; Acquiring process execution data through a data interface of a process control system, and performing time series normalization processing on the process execution data to generate process execution data with a unified time reference; Collect environmental monitoring data through environmental monitoring equipment, and perform spatial interpolation processing on the environmental monitoring data to generate environmental monitoring data covering the entire area of the smart factory; The standardized equipment operation parameter data, the process execution data with a unified time base, and the environmental monitoring data covering the entire smart factory are converted into data formats to generate the multimodal data set.

8. The method according to claim 7, wherein The data integrity check and outlier elimination processing are performed on the original equipment operating parameter data to generate standardized equipment operating parameter data, including: Detecting missing data segments in the original device operating parameter data, and completing the missing data segments using a data interpolation algorithm of adjacent timestamps; Calculating the statistical distribution characteristics of the original equipment operating parameter data, identifying abnormal data points that exceed a preset reasonable value range, and using a sliding window mean filtering algorithm to smooth and replace the abnormal data points; Perform dimensional unification on the processed equipment operating parameter data and convert it into equipment operating parameter data in preset standard units; The equipment operation parameter data of the standard unit is stored in the equipment operation database, and is configured with an equipment identifier and a collection timestamp.

9. The method according to claim 2, wherein The performing cross-modal correlation analysis on the temporal fluctuation characteristics of the temporal data sequence, the spatial distribution characteristics of the spatial data sequence, and the environmental disturbance characteristics of the environmental data sequence by the multimodal fusion model includes: Performing time dimension decomposition processing on the time series fluctuation feature to generate a set of fluctuation components of the time series fluctuation feature at different time scales, wherein the set of fluctuation components includes a short-term fluctuation component, a medium-term fluctuation component, and a long-term trend component; Performing spatial region segmentation processing on the spatial distribution feature to generate a plurality of spatial region features corresponding to the spatial distribution feature, the spatial region features including equipment layout associated region features, process execution intensive region features, and environmental interference diffusion region features; Performing disturbance pattern extraction processing on the environmental disturbance feature to generate a plurality of disturbance pattern features corresponding to the environmental disturbance feature, wherein the disturbance pattern features include a periodic disturbance pattern, a sudden disturbance pattern, and a continuous disturbance pattern; Performing a first modal correlation matching on the short-term fluctuation component in the fluctuation component set and the equipment layout associated regional feature in the spatial regional feature to generate a real-time impact feature of equipment operation and spatial layout; performing a second modal correlation matching on the medium-term fluctuation component and the process execution intensive regional feature to generate a dynamic interaction feature of process parameters and equipment load; performing a third modal correlation matching on the long-term trend component and the environmental interference diffusion regional feature to generate an environmental interference cumulative effect feature; The periodic disturbance pattern and the real-time impact feature are subjected to time-space coupling analysis to generate equipment operation deviation features under periodic environmental interference; the sudden disturbance pattern and the dynamic interaction feature are subjected to abnormal event correlation analysis to generate process parameter mutation warning features; the continuous disturbance pattern and the environmental interference cumulative effect feature are subjected to trend superposition analysis to generate long-term environmental interference suppression demand features; The equipment operation offset characteristics, the process parameter mutation warning characteristics and the long-term environmental interference suppression demand characteristics are integrated to generate dynamic correlation characteristics between the equipment operation parameter data and the process execution process data, and based on the correlation results of the periodic disturbance pattern and the process execution intensive area characteristics, generate coupling characteristics between the process execution process data and the environmental monitoring data.

10. A computer system, characterized in that: include: a memory storing a computer program; A processor, configured to load the computer program to implement the smart factory process optimization method based on multimodal data fusion as described in any one of claims 1 to 9.

Images