Quality safety management method and equipment for industrial production and medium

Abstract

The invention discloses a quality safety management method and device for industrial production and a medium, and relates to the field of industrial Internet of Things, and the method comprises the steps: collecting production data in a production process in real time through a plurality of types of pre-deployed sensors; generating a quality monitoring data set; classifying the quality monitoring data set into a plurality of data subsets in real time based on data types through a rule classification model, and respectively calculating data statistical characteristics of the data subsets in a preset time window; according to the data statistical characteristics, through a pre-trained deep learning model, outputting a quality safety prediction result of a preset future time period; and analyzing the quality safety prediction result, extracting an equipment regulation and control suggestion, generating an equipment control instruction, and issuing the control instruction to corresponding production execution equipment through an industrial bus so as to adjust the operation parameters of the equipment. The stability of the production process is guaranteed, and more intelligent and more reliable management support is provided for industrial production quality safety.

Description

Technical Field

[0001] This application relates to the field of industrial Internet of Things, and specifically relates to a quality and safety management method, device and medium for industrial production. Background Technique

[0002] With the rapid development of modern industry towards the direction of intelligence, digitization and networking, the production process is becoming increasingly complex, and the requirements for product consistency, reliability and traceability have reached an unprecedented height. Especially in key fields related to the national economy and people's livelihood such as manufacturing, food processing, and pharmaceutical production, quality and safety are the lifelines of enterprises. Quality management is no longer a single quality inspection link, but a core value activity that runs through the entire process of design, production and logistics.

[0003] In industrial production, traditional quality management methods mainly rely on manual inspections, regular spot checks and isolated single-point monitoring devices, resulting in a single dimension of data collection, and the inability to achieve the linkage perception of multi-dimensional key elements such as equipment status, product quality indicators and environmental parameters, so that a large number of potential quality and safety hazards are overlooked. The data processing link has weak capabilities, can only perform simple storage and primary statistics, lacks in-depth integration and intelligent analysis of multi-source heterogeneous data, and decision-making relies heavily on manual experience and is lagging. Often, it can only be processed passively after quality problems break out, resulting in waste of resources and rising costs. The automation level of the control link is low, mostly based on a simple trigger mechanism with fixed thresholds, and it is unable to achieve adaptive collaborative control across devices and across links according to global real-time data, resulting in poor control accuracy and low efficiency. In addition, the communication protocols between systems are single, with high data transmission latency, weak security and poor compatibility, forming serious data islands, making it difficult to achieve full-process quality traceability and collaborative management, and ultimately restricting the improvement of production efficiency and quality and safety levels. Summary of the Invention

[0004] To solve the above problems, this application proposes a quality and safety management method for industrial production, including: Real-time collection of production data during the production process through a variety of pre-deployed sensors; the production data includes equipment working state parameters, product quality micro-index parameters, and production environment parameters; Preprocess and fuse the production data to generate a quality monitoring data set, and input it into a rule classification model; Through the rule classification model, based on the data type, the quality monitoring data set is classified into multiple data subsets in real time, and the data statistical features of the data subsets within a preset time window are calculated respectively; According to the data statistical features, through a pre-trained deep learning model, output the quality and safety prediction results for a preset future period; The quality and safety prediction results are analyzed to extract equipment adjustment suggestions, generate equipment control commands, and send the control commands to the corresponding production execution equipment through the industrial bus to adjust the equipment operating parameters.

[0005] On the other hand, this application also proposes a quality and safety management device for industrial production, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform a quality and safety management method for industrial production as described in the above example.

[0006] On the other hand, this application also proposes a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as: a quality and safety management method for industrial production as described in the above example.

[0007] The quality and safety management method for industrial production proposed in this application can bring the following beneficial effects: By pre-deploying various types of sensors, real-time collection of production data across all dimensions is achieved, simultaneously covering three core production elements: equipment operating status, micro-indicators of product quality, and production environment parameters. Targeted preprocessing and fusion are used to generate high-quality monitoring datasets. Combined with rule-based classification models for accurate data classification and statistical feature extraction, production data is transformed from fragmented raw information into structured, high-value analytical data. This allows the data to move beyond being limited to a single link or isolated parameter, deeply connecting key nodes throughout the entire production process, and providing detailed insights into the potential impact between different production elements. This ensures that every piece of production data provides effective support for quality and safety management, improving the comprehensiveness and detail of control.

[0008] Furthermore, by leveraging deep learning models to deeply analyze statistical features, proactive predictions of quality and safety risks can be achieved. This allows for the early detection of potential quality fluctuations or equipment anomalies before risks materialize, providing ample room for adjustment in production management. Simultaneously, by analyzing the prediction results, control suggestions are automatically extracted and control commands are generated. These commands are then precisely distributed to the corresponding production execution equipment via the industrial bus, eliminating the need for manual judgment in adjusting equipment parameters. This significantly reduces human error and delays, forming a complete closed-loop quality and safety management system. This continuously optimizes equipment operation, ensures the stability of the production process, and provides more intelligent and reliable management support for industrial production quality and safety. Attached Figure Description

[0009] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating a quality and safety management method for industrial production as described in an embodiment of this application. Figure 2 This is a schematic diagram of a quality and safety management device for industrial production, as described in an embodiment of this application. Detailed Implementation

[0010] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0012] like Figure 1 As shown in the embodiment of this application, a quality and safety management method for industrial production is provided, including: S101: Real-time collection of production data during the production process through pre-deployed various types of sensors; the production data includes equipment operating status parameters, product quality micro-indicator parameters, and production environment parameters.

[0013] Specifically, various types of sensors are pre-deployed according to the core control needs of different industrial sectors. For example, in a machinery manufacturing workshop, vibration sensors need to be installed on machine tool bearings to monitor vibration frequency and amplitude, and temperature sensors need to be deployed on motor housings to collect operating temperature and temperature rise rate. In a food processing production line, near-infrared sensors need to be installed next to the product conveyor belt to detect food moisture content, fat percentage and other component data in real time, targeting microscopic indicators of product quality. In a chemical production workshop, gas sensors need to be installed in different areas of the workshop to monitor the concentration of toxic and harmful gases, and temperature and humidity sensors need to record parameters such as ambient temperature fluctuations and relative humidity, targeting production environment parameters.

[0014] It should be noted that sensor deployment must cover key equipment components, product production chains, and key environmental areas to avoid monitoring blind spots. At the same time, wireless (such as ZigBee, LoRa) or wired (such as industrial Ethernet) transmission methods should be used to transmit the collected real-time data to the data processing terminal to ensure that the data can reflect the production dynamics in real time and solve the problem of fragmented data on equipment, products, and environment in traditional monitoring.

[0015] S102: The production data is preprocessed and fused to generate a quality monitoring dataset, which is then input into the rule classification model.

[0016] Specifically, data preprocessing is performed. For outliers that may appear in the sensors, the 3σ principle or Grubbs test is used to remove them. For missing values ​​caused by transmission delays (such as gaps in temperature and humidity data caused by fluctuations in the workshop network), linear interpolation or nearest-neighbor mean imputation is used to supplement them. For the differences in data formats of different sensors, they are uniformly converted into formats that conform to industry standards (such as according to the units and precision specified in the "Guidelines for Industrial Data Standardization").

[0017] Data fusion is performed using multi-source heterogeneous data fusion algorithms, such as static fusion algorithms based on weighted averages or dynamic fusion algorithms based on Kalman filtering, to integrate equipment, product, and environmental data into the same timeline and data framework. For example, using "production batch ID and timestamp" as indexes, machine tool vibration data, food composition data, and workshop temperature and humidity data at a specific moment can be linked to generate a structured quality monitoring dataset.

[0018] Input the dataset into a pre-defined rule-based classification model. The model needs to set classification rules in advance based on industrial production business logic and data characteristics, such as dividing the data into categories like "equipment operation", "product quality", and "environmental parameters", to prepare for subsequent targeted analysis.

[0019] In this embodiment, the specific process of preprocessing and fusing production data is as follows: based on the source sensor type of the production data, the corresponding preprocessing engine is called for preprocessing. When the production data comes from a vibration sensor, the production data is denoised using a wavelet transform algorithm. When the production data comes from a near-infrared spectral sensor, the spectral parameters in the production data are corrected for scattering using a normal variable transformation algorithm. Based on the acquisition timestamp, the preprocessed production data is time-series aligned. The time-series aligned production data is then fused to generate a quality monitoring dataset.

[0020] It's important to note that, based on the sensor type from which the production data originates, corresponding preprocessing tools are used. For example, vibration data collected by vibration sensors is prone to mechanical noise or electromagnetic interference, so wavelet transform algorithms are used to filter out this useless noise, retaining only the effective data that truly reflects the equipment's operating conditions. Similarly, when near-infrared spectroscopy sensors detect product components, the spectral signal can be affected by scattering from sample particles, leading to data deviations. Normal variable transformation algorithms are used to correct this scattering interference, making the product component parameters more accurate. After preprocessing individual sensor data, all data are aligned chronologically using the acquisition timestamp as a unified standard. For instance, equipment vibration data and product near-infrared component data from the same moment are matched to avoid data misalignment caused by acquisition time differences. The time-aligned equipment, product, and environmental data are then integrated to form a structured and accurate quality monitoring dataset, laying a solid foundation for subsequent classification and analysis.

[0021] S103: Using the rule-based classification model, the quality monitoring dataset is classified into multiple data subsets in real time based on the data type, and the statistical characteristics of each data subset within a preset time window are calculated.

[0022] Specifically, through a rule-based classification model, the data points in the quality monitoring dataset are traversed, the physical unit attribute of each data point is extracted, and the physical unit attribute is matched with the mapping relationship in the preset classification rule library. The preset classification rule library stores the mapping relationship between the physical unit attribute of production data and the data category, as well as the mapping relationship between production data and the production process code. Based on the matching results, the data points are classified into the corresponding data subsets, and the corresponding production process code identifier is added to the classified data points. The data itself includes a subset of equipment operation data, a subset of product quality data, or a subset of production environment data.

[0023] The rule-based classification model iterates through each data point in the quality monitoring dataset, extracting the physical unit attribute that reflects the data type from each data point. It then calls a pre-defined classification rule library, which stores two types of core mapping relationships: one is the correspondence between the physical unit attribute of production data and the data category (such as equipment, product, and environment-related categories), and the other is the correspondence between production data and production process codes. The model compares and matches the extracted physical unit attribute with these mapping relationships in the rule library. Based on the successful matching, the model divides each data point into a corresponding exclusive data subset, namely, the equipment operation data subset, the product quality data subset, or the production environment data subset. At the same time, it adds the corresponding production process code identifier to the classified data points, thereby realizing the classification management and process association of data, laying the foundation for subsequent targeted analysis.

[0024] It should be noted that the rule-based classification model classifies the quality monitoring dataset in real time according to preset classification rules. For example, "machine tool vibration frequency and motor temperature" are classified into the equipment operation data subset, "food moisture content and component deviation rate" are classified into the product quality data subset, and "workshop temperature and gas concentration" are classified into the environmental parameter data subset. The classification process must be real-time to avoid data accumulation affecting subsequent analysis.

[0025] Furthermore, for each data subset, a corresponding sliding time window is created based on a preset fixed time length. The movement event of the sliding time window is triggered, and the latest data point of the sliding time window is retained. The values ​​of the latest data points in the same data subset are summed to calculate the average value of the sliding time window, thus obtaining the average value feature of the data subset. The latest data points are traversed to determine the maximum and minimum value data points. Based on the corresponding maximum and minimum values, the range within the sliding time window is calculated to obtain the range feature of the data subset. The previous data point corresponding to the previous sliding time window in the data subset is obtained. Based on the current data point and the previous data point in the current sliding time window, the change in data points is determined. Based on the change in data points, the average trend of the data subset is calculated.

[0026] For each data subset, a corresponding sliding time window is created based on a preset fixed time length to build a time range framework for feature calculation. The sliding time window movement event is triggered, and the window automatically retains the latest data points. After summing the values ​​of these latest data points in the same data subset, the average value is calculated by combining the number of data points in the window, thereby obtaining the average feature of the data subset.

[0027] The algorithm iterates through the latest data points within the window, identifying the largest and smallest values. These two values ​​are then used to calculate the range of the data within the window, thus obtaining the range characteristics of the data subset. Next, the algorithm retrieves the previous data point corresponding to the previous sliding time window from the data subset. This is compared to the current data point in the current sliding time window, and the change in data point between the two is calculated. Based on this change, further calculations are performed to obtain the average trend of the data subset, thereby fully extracting the statistical characteristics of the data subset in the time dimension.

[0028] It should be noted that a preset time window is set, and statistical characteristics are calculated for each data subset within the corresponding time window. For the equipment operation data subset, the mean, variance, and peak values ​​of vibration frequency, as well as the maximum and minimum values ​​and temperature rise rate of motor temperature, are calculated. For the product quality data subset, the component qualification rate and the standard deviation of deviation values ​​are calculated. For the environmental parameter data subset, the fluctuation range of temperature and humidity and the average value of gas concentration are calculated. These statistical characteristics can effectively filter out instantaneous noise, accurately reflect the overall state of a certain production stage, and provide high-quality predictive input for deep learning models.

[0029] S104: Based on the statistical characteristics of the data, output the quality and safety prediction results for a preset future time period through a pre-trained deep learning model.

[0030] Specifically, based on the average value, range, and average trend within multiple consecutive sliding time windows, multiple sets of feature vectors are generated. These feature vectors, corresponding to subsets of equipment operation data, product quality data, and production environment data, are then input into a deep learning model in parallel. The deep learning model analyzes the time-dimensional evolution patterns of these feature vectors and outputs anomaly probability values ​​for a future preset time period based on these patterns. These anomaly probability values ​​include a first probability value for equipment failure, a second probability value for key product indicators exceeding acceptable ranges, and a third probability value for environmental parameters exceeding process requirements. The anomaly probability values ​​are compared with their corresponding preset anomaly thresholds, and quality and safety prediction results are generated based on the comparison results.

[0031] Based on the average value, range, and average trend extracted from multiple consecutive sliding time windows, these features reflecting the temporal patterns of the data are combined into multiple feature vectors, transforming scattered temporal information into structured input that can be processed by the deep learning model. Multiple feature vectors corresponding to three data subsets—equipment operation, product quality, and production environment—are simultaneously and in parallel input into the deep learning model, ensuring that the model can synchronously capture the temporal correlations of the three dimensions of data. The deep learning model deeply analyzes the evolution patterns of these feature vectors in the time dimension (such as the rising, falling, or fluctuating patterns of feature values), and calculates and outputs three types of anomaly probability values ​​within a preset future time period based on these patterns. These correspond to the likelihood of equipment failure, exceeding key product indicators, and exceeding process requirements for environmental parameters, respectively. Each anomaly probability value is compared with a pre-set corresponding anomaly threshold. Based on the comparison results (such as whether the probability value exceeds the threshold and the degree of exceedance), a comprehensive judgment is made on whether there is a quality and safety risk, thereby generating the final quality and safety prediction result.

[0032] Furthermore, the abnormal probability value is compared with the corresponding preset abnormal threshold. When the abnormal probability value is higher than the corresponding preset abnormal threshold, the risk type corresponding to the abnormal probability value is determined to be an abnormal state. Based on the risk type, equipment control suggestions are generated. Based on the risk type, the corresponding abnormal probability value, the time interval information of the future period, and the equipment control suggestions, a quality and safety prediction result is generated.

[0033] It should be noted that the specific process of model training is as follows: collect historical production data in this industrial field, including normal production data, such as vibration data during stable equipment operation, composition data when products are qualified, and data on quality anomaly cases, such as product scrap records caused by equipment failure and quality problem data caused by environmental exceedances, and label the data; use a Long Short-Term Memory (LSTM) network or a gated recurrent unit (GRU) to build a deep learning model, divide the labeled historical data into training and testing sets, input them into the deep learning model for training, and adjust parameters such as the learning rate and number of iterations to enable the model to learn the correlation patterns between data statistical features and quality and safety risks until the model's prediction accuracy reaches the standards for industrial applications.

[0034] S105: Analyze the quality and safety prediction results, extract equipment adjustment suggestions, generate equipment control commands, and send the control commands to the corresponding production execution equipment through the industrial bus to adjust the equipment operating parameters.

[0035] Specifically, the quality and safety prediction results are broken down and analyzed to extract specific risk types, identify the core categories of current potential quality and safety problems, and take differentiated treatment measures according to different risk types.

[0036] If the extracted risk type is equipment failure risk, the system will rely on a pre-set equipment risk control instruction mapping table to directly obtain equipment control instructions that can cope with the failure risk by matching the risk type with the entries in the table. If the risk type is product qualification risk, the system will automatically generate a standardized quality inspection report based on the risk details in the quality and safety prediction results (such as indicators that may exceed the standard, risk-related links, etc.) and send the report to the management terminal of the quality inspectors to ensure that the quality inspectors can keep abreast of product qualification risk information and carry out subsequent processing.

[0037] Furthermore, by adapting to the stable transmission characteristics of industrial production scenarios through industrial bus, control commands are transmitted to the target equipment through its dedicated communication link. After receiving the command, the target production execution equipment will automatically parse the parameter adjustment requirements contained in the command, and then adjust its own operating parameters (such as speed, temperature, pressure, etc.) in real time according to the values ​​or standards set by the command. Ultimately, it can achieve precise control over the operating status of production equipment and ensure that the production process meets quality and safety requirements.

[0038] In addition, after the equipment continues to operate based on the adjusted parameters, the control commands and the adjusted equipment parameters are recorded, the offline inspection results of the products based on the preset production cycle are collected, the deviation between the offline inspection results of the products and the expected standard is calculated, and training samples are constructed based on the control commands, the adjusted equipment parameters and the deviation, and stored in the historical training database. Based on the preset time interval, the deep learning model is incrementally trained periodically using the samples in the historical database.

[0039] This application achieves real-time, multi-dimensional acquisition of production data through the pre-deployment of various types of sensors, simultaneously covering three core production elements: equipment operating status, micro-indicators of product quality, and production environment parameters. It also pre-processes and integrates data to generate high-quality monitoring datasets. Combined with a rule-based classification model for precise data classification and statistical feature extraction, this transforms production data from fragmented raw information into structured, high-value analytical data. This allows the data to move beyond single links or isolated parameters, deeply connecting to key nodes throughout the entire production process and providing detailed insights into the potential impacts between different production elements. This ensures that every piece of production data provides effective support for quality and safety management, enhancing the comprehensiveness and detail of control.

[0040] Furthermore, by leveraging deep learning models to deeply analyze statistical features, proactive predictions of quality and safety risks can be achieved. This allows for the early detection of potential quality fluctuations or equipment anomalies before risks materialize, providing ample room for adjustment in production management. Simultaneously, by analyzing the prediction results, control suggestions are automatically extracted and control commands are generated. These commands are then precisely distributed to the corresponding production execution equipment via the industrial bus, eliminating the need for manual judgment in adjusting equipment parameters. This significantly reduces human error and delays, forming a complete closed-loop quality and safety management system. This continuously optimizes equipment operation, ensures the stability of the production process, and provides more intelligent and reliable management support for industrial production quality and safety.

[0041] like Figure 2 As shown in the embodiments of this application, a quality and safety management device for industrial production is also proposed, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a quality and safety management method for industrial production as described in any of the above embodiments.

[0042] This application also provides a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured as: a quality and safety management method for industrial production as described in any of the above embodiments.

[0043] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device and medium embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the description of the method embodiments.

[0044] The devices and media provided in this application are one-to-one with the methods. Therefore, the devices and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be repeated here.

[0045] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0046] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0047] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0048] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0049] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0050] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0051] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0052] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0053] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A quality and safety management method for industrial production, characterized in that, include: Production data is collected in real time during the production process through pre-deployed various types of sensors; The production data includes equipment operating status parameters, product quality micro-indicator parameters, and production environment parameters. The production data is preprocessed and fused to generate a quality monitoring dataset, which is then input into a rule-based classification model. Based on the data type, the quality monitoring dataset is classified into multiple data subsets in real time using the rule-based classification model, and the statistical characteristics of each data subset are calculated within a preset time window. Based on the statistical characteristics of the data, a pre-trained deep learning model is used to output a quality and safety prediction result for a preset future time period. The quality and safety prediction results are analyzed to extract equipment adjustment suggestions, generate equipment control commands, and send the control commands to the corresponding production execution equipment through the industrial bus to adjust the equipment operating parameters.

2. The quality and safety management method for industrial production according to claim 1, characterized in that, The step of classifying the quality monitoring dataset into multiple data subsets in real time based on data type using the rule-based classification model specifically includes: Using the rule-based classification model, the data points in the quality monitoring dataset are traversed to extract the physical unit attribute of each data point. The physical unit attributes are matched with the mapping relationships in the preset classification rule base; the preset classification rule base stores the mapping relationships between the physical unit attributes of production data and data categories, as well as the mapping relationships between the production data and production process codes; Based on the matching results, the data points are classified into corresponding data subsets, and corresponding production process codes are added to the classified data points; the data itself includes a subset of equipment operation data, a subset of product quality data, or a subset of production environment data.

3. The quality and safety management method for industrial production according to claim 2, characterized in that, The calculation of the statistical characteristics of the data subsets within a preset time window specifically includes: For each subset of data, a corresponding sliding time window is created based on a preset fixed time length; Trigger the movement event of the sliding time window, retain the latest data point of the sliding time window, sum the values ​​of the latest data points in the same data subset, calculate the average value of the sliding time window, and obtain the average value feature of the data subset; Traverse the latest data points to determine the maximum and minimum data points. Based on the corresponding maximum and minimum values, calculate the range within the sliding time window to obtain the range characteristics of the data subset. Obtain the previous data point corresponding to the previous sliding time window in the data subset. Based on the current data point in the current sliding time window and the previous data point, determine the change in the data point. Based on the change in the data point, calculate the average trend of the data subset.

4. The quality and safety management method for industrial production according to claim 3, characterized in that, The step of outputting a quality and safety prediction result for a preset future time period based on the statistical characteristics of the data and through a pre-trained deep learning model specifically includes: Based on the average value feature, the range feature, and the average change trend within multiple consecutive sliding time windows, multiple sets of feature vectors are generated. Multiple sets of feature vectors corresponding to the equipment operation data subset, the product quality data subset, and the production environment data subset are input in parallel into the deep learning model; The deep learning model analyzes the time-dimensional evolution pattern of the multiple sets of feature vectors and outputs the anomaly probability value within a preset future time period based on the time-dimensional evolution pattern. The anomaly probability value includes a first probability value of equipment failure, a second probability value of key product indicators exceeding the qualified range, and a third probability value of environmental parameters exceeding process requirements. The abnormal probability values ​​are compared with the corresponding preset abnormal thresholds, and a quality and safety prediction result is generated based on the comparison results.

5. The quality and safety management method for industrial production according to claim 4, characterized in that, The step of comparing the anomaly probability value with the corresponding preset anomaly threshold, and generating a quality and safety prediction result based on the comparison result, specifically includes: The abnormal probability value is compared with the corresponding preset abnormal threshold. When the abnormal probability value is higher than the corresponding preset abnormal threshold, the risk type corresponding to the abnormal probability value is determined to be an abnormal state. Based on the risk type, equipment control suggestions are generated. Based on the risk type, the corresponding anomaly probability value, the time interval information of the future period, and the equipment control suggestions, quality and safety prediction results are generated.

6. The quality and safety management method for industrial production according to claim 1, characterized in that, The step of analyzing the quality and safety prediction results, extracting equipment adjustment suggestions, and generating equipment control commands specifically includes: The quality and safety prediction results are analyzed to extract risk types; When the risk type is equipment failure risk, the corresponding equipment control instruction is obtained based on the preset equipment risk control instruction mapping table; When the risk type is product qualification risk, a quality inspection report is generated based on the quality and safety prediction results and sent to the quality inspection personnel management terminal.

7. The quality and safety management method for industrial production according to claim 1, characterized in that, The preprocessing and fusion of the production data to generate a quality monitoring dataset specifically includes: Based on the source sensor type of the production data, the corresponding preprocessing engine is invoked for preprocessing. When the production data comes from vibration sensors, the production data is denoised using a wavelet transform algorithm. When the production data comes from a near-infrared spectral sensor, the spectral parameters in the production data are corrected for scattering using a normal variable transformation algorithm; Based on the collection timestamp, the preprocessed production data is time-series aligned, and the time-series aligned production data is merged to generate a quality monitoring dataset.

8. The quality and safety management method for industrial production according to claim 1, characterized in that, After the control commands are sent to the corresponding production execution equipment via the industrial bus to adjust the equipment operating parameters, the method further includes: Record the control commands and the adjusted equipment parameters; Collect offline product testing results based on a preset production cycle, and calculate the deviation between the offline product testing results and the expected standard; Based on the control command, the adjusted equipment parameters, and the deviation, training samples are constructed and stored in the historical training database. Based on a preset time interval, the deep learning model is incrementally trained periodically using samples from the historical database.

9. A quality and safety management device for industrial production, characterized in that, include: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a quality and safety management method for industrial production as described in claims 1 to 8.

10. A non-volatile computer storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are configured to execute a quality and safety management method for industrial production as described in claims 1 to 8.

Images