An intelligent production management method and device applied to a multi-stage production line and a medium

Abstract

This invention discloses an intelligent production management method, device, and medium applied to multi-level production lines, relating to the technical field of intelligent production management. The method includes: deploying multimodal sensors in multiple sections of the target multi-level production line to form multiple section sensor arrays; defining the target batch of waste paint residue material and product quality constraints; drying the target batch of waste paint residue according to initial parameters; obtaining sensor data sequences based on adaptive windows and preset frequencies; extracting time-series iterative high-dimensional features of each section; performing backtracking high-dimensional linkage identification; adjusting initial drying parameters by section and transmitting the data to the drying production line for updating. This invention solves the technical problems in existing technologies, such as the inability to respond in real-time to fluctuations in waste paint residue material and changes in production equipment status, and the difficulty in comprehensively capturing the status of multiple sections across the entire production line, leading to poor product quality consistency and production efficiency. It achieves comprehensive perception and linkage control of multiple section statuses, improving product quality consistency and production efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent production management technology, specifically to an intelligent production management method, device, and medium applicable to multi-level production lines. Background Technology

[0002] Multi-stage drying is a core component of the waste paint residue resource utilization process. Through continuous processes such as crushing, multi-stage drying, fine grinding, deep drying, and ultrafine grinding, the moisture content of the paint residue is gradually reduced and its particle size distribution is controlled to obtain dried ultrafine powder that meets utilization requirements. However, the operating states of the crushing, drying, and grinding stages in the drying process are mutually influential and affected by factors such as the initial composition of the material, fluctuations in moisture content, and equipment performance degradation. Drying control is difficult to adapt to changes in operating conditions in real time, which can easily lead to problems such as insufficient or excessive drying, excessive energy consumption, and poor product uniformity.

[0003] Existing technologies such as CN112506145A (production management system, production management device and production management method for production lines), CN116468256A (management method and device for multiple production lines), CN111176223A (a safety management system and method for a production line), and CN107203848A (a field management system and management method for a production line) have poor versatility and lack comprehensive perception and in-depth analysis of the collaborative operation status of multiple sections of the entire production line. As a result, they have insufficient adaptive adjustment capabilities when facing fluctuations in incoming materials or changes in equipment status, making it difficult to achieve optimal energy efficiency and productivity while ensuring the quality constraints of the final product.

[0004] Therefore, current technologies suffer from the inability to respond in real time to fluctuations in waste paint residue and changes in the status of production equipment, and the difficulty in comprehensively capturing the status of multiple stages across the entire production line, resulting in poor product quality consistency and low production efficiency. Summary of the Invention

[0005] This invention provides an intelligent production management method, device, and medium applicable to multi-level production lines. It solves the technical problems in the prior art that it cannot respond in real time to fluctuations in waste paint residue and changes in the status of production equipment, and that it is difficult to fully capture the status of multiple sections of the entire production line, resulting in poor product quality consistency and production efficiency. The invention achieves the technical effect of realizing comprehensive perception and linkage control of the status of multiple sections, thereby improving product quality consistency and production efficiency.

[0006] This invention provides an intelligent production management method applied to a multi-level production line. The method includes: deploying multi-modal sensors in the crushing, drying, fine grinding, deep drying, ultrafine grinding, and packaging sections of a target multi-level production line to obtain multiple section sensor arrays; acquiring the waste paint residue material and product quality constraints of the target batch; starting the target multi-level production line to dry the waste paint residue material according to the initial drying parameters; using the multiple section sensor arrays to sample according to the production line adaptive window and a preset sampling frequency to obtain multiple section sensor sampling data array sequences; performing section-specific time-series iterative feature extraction on the multiple section sensor sampling data array sequences to obtain multiple section high-dimensional features; combining the product quality constraints and the multiple section high-dimensional features to perform backtracking high-dimensional linkage identification; and adjusting the initial drying parameters according to the identification results to obtain adjusted drying parameters; and transmitting the adjusted drying parameters to the target multi-level production line for drying parameter update management.

[0007] In one possible implementation, the intelligent production management method applied to multi-level production lines further performs the following processing: the production line adaptive window is the time period between the start time node of the crushing section and the time node when the number of products completed at the packaging section meets the preset product quantity threshold, and extracts the consistency principal component; and uses the fluctuation amplitude of the consistency principal component to construct an abnormal trend index.

[0008] In one possible implementation, the intelligent production management method applied to multi-level production lines further performs the following processing: extracting a first section sensor sampling data array sequence from the multiple section sensor sampling data array sequences; performing multi-scale time-series trend feature identification on the first section sensor sampling data array sequence in descending order of scale to obtain a first multi-scale data trend feature array sequence; and performing feature iteration on the first multi-scale data trend feature array sequence to obtain the high-dimensional features of the first section.

[0009] In one possible implementation, the intelligent production management method applied to multi-level production lines further performs the following processing: extracting the first and second multi-scale data trend feature arrays from the first multi-scale data trend feature array sequence and performing feature iteration to obtain high-dimensional features of the first stage segment; using the high-dimensional features of the first stage segment to perform feature iteration on the third multi-scale data trend feature array to obtain high-dimensional features of the second stage segment; and so on, using the high-dimensional features of the stage segment obtained from the previous feature iteration to perform feature iteration on the next multi-scale data trend feature array until the last element is reached to obtain the high-dimensional features of the first segment.

[0010] In one possible implementation, the intelligent production management method applied to multi-level production lines further performs the following processing: similarity identification is performed on trend features corresponding to the same data type in the first-level multi-scale data trend feature array and the second-level multi-scale data trend feature array to obtain a first adjacency similarity set; the first adjacency similarity set is matrix-processed to obtain a first adjacency matrix; the second-level multi-scale data trend feature array is iteratively enhanced using the first adjacency matrix to obtain a first enhanced multi-scale data trend feature array, and after weighted fusion, high-dimensional features of the stage section are obtained.

[0011] In one possible implementation, the intelligent production management method applied to a multi-level production line further performs the following processing: based on the product quality constraints, anomaly identification is performed on the high-dimensional features of the multiple work sections to obtain multiple abnormal high-dimensional features; according to the work section flow of the target multi-level production line, multi-level correlation backtracking is performed on the multiple abnormal high-dimensional features to obtain multiple sets of correlated backtracking high-dimensional features, wherein each correlated backtracking high-dimensional feature has a correlation weight; the multiple abnormal high-dimensional features and the multiple sets of correlated backtracking high-dimensional features are mapped and combined, and the obtained multiple mapping combinations are used as the identification result.

[0012] In one possible implementation, the intelligent production management method applied to a multi-level production line further performs the following processing: traversing the multiple mapping combinations to perform segmented adaptive adjustments to the initial drying parameters to obtain multiple adjusted drying parameters; and performing overall linkage optimization on the multiple adjusted drying parameters with the goal of minimizing energy consumption and maximizing quality to determine the adjusted drying parameters.

[0013] In one possible implementation, the intelligent production management method applied to a multi-level production line further performs the following processing: adjusting the preset sampling frequency in segments based on the high-dimensional features of the multiple work sections to obtain multiple work section sampling frequencies; sampling the production line regular window according to the multiple work section sampling frequencies, wherein the production line regular window starts at the time point when the production line adaptive window ends and ends at the time point when the production line ends production.

[0014] This invention also provides an intelligent production management device for multi-level production lines. The device includes: a sensor array acquisition module, used to deploy multi-modal sensors in the crushing, drying, fine grinding, deep drying, ultrafine grinding, and packaging sections of the target multi-level production line to acquire multiple section sensor arrays; and a sampling data acquisition module, used to acquire the waste paint residue material and product quality constraints of the target batch, start the target multi-level production line to dry the waste paint residue material according to initial drying parameters, and utilize the multiple section sensor arrays according to the production line adaptive window. The system samples data at a preset sampling frequency to obtain multiple work section sensor sampling data array sequences. A high-dimensional feature acquisition module is used to perform work section-specific time-series iterative feature extraction on the multiple work section sensor sampling data array sequences to obtain multiple work section high-dimensional features. An adjustment drying parameter acquisition module is used to perform backtracking high-dimensional linkage identification by combining the product quality constraints and the multiple work section high-dimensional features, and to perform work section-specific adaptive adjustment of the initial drying parameters according to the identification results to obtain adjusted drying parameters. The adjusted drying parameters are then transmitted to the target multi-level production line for drying parameter update management.

[0015] The present invention also provides a computer-readable storage medium, comprising: storing thereon a computer program that, when executed by a processor, implements an intelligent production management method applicable to multi-level production lines.

[0016] This invention proposes an intelligent production management method, device, and medium for multi-level production lines. It involves deploying multimodal sensors in multiple sections of the target multi-level production line, forming multiple section sensor arrays. The method clarifies the target batch of waste paint residue material and product quality constraints, dries the target batch of waste paint residue according to initial parameters, and obtains sensor data sequences based on adaptive windows and preset frequency sampling. It extracts high-dimensional features of time-series iteration for each section, performs backtracking high-dimensional linkage identification, adjusts the initial drying parameters for each section, and transmits the data to the drying production line for updates. This solves the technical problems of existing technologies, such as the inability to respond in real-time to fluctuations in waste paint residue material and changes in production equipment status, and the difficulty in comprehensively capturing the status of multiple sections across the entire production line, leading to poor product quality consistency and production efficiency. It achieves comprehensive perception and linkage control of multiple section statuses, improving product quality consistency and production efficiency. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this invention to illustrate the operations performed by the apparatus according to the embodiments of the present invention. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0018] Figure 1 This is a schematic diagram of a smart production management method applied to multi-level production lines, provided as an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of an intelligent production management device applied to a multi-level production line, provided as an embodiment of the present invention.

[0020] Figure labeling: Sensor array acquisition module 10, sampling data acquisition module 20, high-dimensional feature acquisition module 30, and drying parameter adjustment acquisition module 40. Detailed Implementation

[0021] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0022] This invention provides an intelligent production management method applicable to multi-level production lines, such as... Figure 1 As shown, the method includes:

[0023] In step S100, multi-modal sensors are deployed in the crushing section, drying section, fine grinding section, deep drying section, ultrafine grinding section and packaging section of the target multi-stage production line to obtain multiple section sensor arrays.

[0024] In this embodiment, the target multi-stage production line refers to a multi-stage drying production line for waste paint sludge. In the production of waste paint sludge recycling, the multi-stage drying process includes multiple stages such as crushing, drying, fine grinding, deep drying, ultrafine grinding, and packaging. There are complex coupling relationships between each stage, such as temperature and humidity, wind speed, moisture content, and energy consumption. Therefore, multi-modal sensors are deployed in the crushing, drying, fine grinding, deep drying, ultrafine grinding, and packaging stages of the target multi-stage production line to measure different types of physicochemical parameters. This includes deploying multiple types of sensors at different locations in each stage to obtain multiple stage sensor arrays, so as to comprehensively perceive the overall operating status of each stage and the state of the waste paint sludge material. The multiple stage sensor arrays include a crushing stage sensor array, a drying stage sensor array, a fine grinding stage sensor array, a deep drying stage sensor array, an ultrafine grinding stage sensor array, and a packaging stage sensor array.

[0025] In this embodiment, for the crushing section sensor array, vibration sensors are installed on the crusher body to monitor mechanical conditions such as bearing wear, rotor imbalance, and component loosening; current / power sensors monitor the load current or real-time power of the crusher motor to reflect material input, hardness changes, or equipment stall; acoustic sensors collect crusher operating sounds to analyze abnormal impact and friction sounds; and industrial cameras capture images of the crushed material to estimate the particle size distribution of waste paint residue. For the drying section sensor array / deep drying section sensor array, thermocouples or resistance thermometers are deployed inside the drying equipment to plot the temperature field distribution to monitor temperature uniformity; humidity sensors are deployed at the exhaust vent and key internal points of the drying equipment to measure exhaust gas humidity or cavity air humidity to reflect the material drying rate; wind speed / pressure sensors are installed in the air duct to monitor hot air velocity and pressure drop; and an online material moisture detector is deployed at the dryer outlet to measure the instantaneous moisture content of the material in real time and non-contactly.

[0026] In this embodiment, for the fine grinding section sensor array / ultrafine grinding section sensor array, vibration sensors are installed in the grinder to monitor the vibration state of the main shaft and classifier bearings, differential pressure sensors monitor the pressure difference before and after the air jet mill nozzle or on both sides of the classifier to reflect the internal flow field state of the equipment, and a laser particle size analyzer detects and outputs the particle size distribution data of the material in real time; for the packaging section sensor array, a scale is installed in the packaging machine to accurately control the weight of each packaging bag, a metal detector is installed to detect whether there are accidental metal impurities mixed in the material before packaging, and a temperature sensor is installed to measure the material temperature before packaging to ensure that the product is packaged at a safe temperature.

[0027] Step S200: Obtain the waste paint residue material and product quality constraints of the target batch, start the target multi-stage production line to dry the waste paint residue material according to the initial drying parameters, and use the multiple section sensor arrays to sample according to the production line adaptive window and preset sampling frequency to obtain multiple section sensor sampling data array sequences.

[0028] Step S200 further includes the production line adaptive window being the time period between the start time of the crushing section and the time when the number of products completed at the packaging section meets a preset product quantity threshold.

[0029] In this embodiment of the application, when the target batch of waste paint residue is started to be processed, the initial attribute information of the waste paint residue material of the target batch is obtained, including at least the initial moisture content, the approximate proportion of the main resin type, pigments, and additives, as well as the physical form such as viscosity and block size, which is used to match the initial drying parameters such as initial temperature, wind speed, and feeding speed; at the same time, the product quality constraints that the final product needs to meet are obtained from the production order or process standard, including at least the final moisture content, particle size distribution, and impurity content, for example, requiring the moisture content of the dried powder to be less than 0.5% and requiring 90% of the particles to have a particle size of less than 10 micrometers.

[0030] In this embodiment, the control system starts the target multi-stage production line. Based on the attribute information of the waste paint residue, it selects empirical initial drying parameters from a preset process parameter library and distributes them to the crushers, dryers, pulverizers, and other execution equipment in each section to dry the target batch of waste paint residue. Then, using multiple section sensor arrays according to the production line adaptive window and preset sampling frequency, it collects the status data of the target multi-stage production line during operation. The production line adaptive window refers to the complete drying process time from the time node when the crushing section starts processing the target batch of waste paint residue to the time node when the number of qualified products produced in the packaging section meets the preset product quantity threshold. The preset sampling frequency is the data sampling time interval set for each sensor. For example, parameters that change rapidly, such as temperature, pressure, and current, are set to a high frequency sampling of 1 second / time, while parameters that change slowly or have a long analysis period, such as moisture and particle size, may be set to a low frequency sampling of 1 minute / time or 5 minutes / time, to ensure that the sampling data can accurately reflect the dynamic changes of the process in the time dimension. The final output consists of multiple sequence of sensor sampling data arrays from different work sections, including multi-dimensional data streams collected by sensor arrays in each work section arranged in chronological order, and is marked with data sampling timestamps.

[0031] Step S300: Perform segment-based time-series iterative feature extraction on the multiple segment sensor sampling data array sequence to obtain multiple segment high-dimensional features.

[0032] Step S300 further includes step S310, extracting the first section sensor sampling data array sequence from the plurality of section sensor sampling data array sequences; step S320, performing multi-scale time-series trend feature identification on the first section sensor sampling data array sequence in descending order of scale to obtain the first multi-scale data trend feature array sequence; step S330, performing feature iteration on the first multi-scale data trend feature array sequence to obtain the high-dimensional features of the first section.

[0033] In this embodiment, segment-by-segment time-series iterative feature extraction is performed on multiple segment-by-segment sensor sampling data array sequences. Segment-by-segment time-series iterative feature extraction refers to parallel multi-scale time-series trend feature identification and feature iteration of the sampling data of each segment, and fusion into a high-dimensional feature vector of the segment in chronological order. Specifically, the first segment sensor sampling data array sequence is extracted from the multiple segment sensor sampling data array sequences. The first segment sensor sampling data array sequence is any one of the multiple segment sensor sampling data array sequences and contains temperature, humidity, wind speed, and other data from all sensors in that segment.

[0034] In this embodiment, the sensor sampling data array sequence of the first process section is subjected to multi-scale temporal trend feature identification in descending order of scale. Simultaneously, different dynamic patterns in the long-term, medium-term, and short-term stages of the production process are captured. Specifically, a longer time window is used for large-scale scanning to identify and determine the overall trend, such as whether the temperature rises slowly, falls slowly, or remains stable throughout the entire process section. A medium time window is used for medium-scale analysis to identify periodic fluctuations caused by equipment start-up, shutdown, or control cycles, or to determine whether the process has entered a stable state. A very short time window is used for small-scale analysis to capture instantaneous events, such as sudden temperature spikes or brief pressure drops, which may characterize equipment anomalies or sudden changes in material properties. The sensor sampling data array sequence of the first process section is then transformed into a set of multi-scale trend features to obtain the first multi-scale data trend feature array sequence. Then, feature iteration is performed on the first multi-scale data trend feature array sequence to encode the contextual dependencies of the time series into the final features to reflect the current state and the evolution process and inertia of the state. Finally, the high-dimensional features of the first section are output. Multi-scale time series trend feature recognition is performed on the sensor sampling data array sequences of all sections to determine the high-dimensional features of multiple sections.

[0035] Furthermore, step S330 also includes step S331, extracting the first and second multi-scale data trend feature arrays from the first multi-scale data trend feature array sequence and performing feature iteration to obtain the high-dimensional features of the first stage section; step S332, using the high-dimensional features of the first stage section to perform feature iteration on the third multi-scale data trend feature array to obtain the high-dimensional features of the second stage section; step S333, and so on, using the high-dimensional features of the stage section obtained from the previous feature iteration to perform feature iteration on the next multi-scale data trend feature array until the last element is reached to obtain the high-dimensional features of the first stage section.

[0036] In this embodiment, the first and second multi-scale data trend feature arrays are extracted from the first multi-scale data trend feature array sequence to represent the operating trend of the work section in two adjacent time periods. Then, feature iteration is performed on them. Specifically, the similarity of corresponding parts of the same data type in the first and second multi-scale data trend feature arrays is calculated to obtain an adjacency similarity set, which is then matrix-processed to express the pattern relationship from state change. Enhanced fusion is then performed to generate the high-dimensional features of the first stage work section. Then, the high-dimensional features of the first stage work section are used to iterate the features of the third multi-scale data trend feature array to obtain the high-dimensional features of the second stage work section. This process continues for all time points, using the high-dimensional features of the stage work section obtained from the previous feature iteration to iterate the features of the next multi-scale data trend feature array until the last multi-scale trend feature in the work section's sensing sequence is processed. Finally, the high-dimensional features of the first work section are output, fusing all key trend information of the entire production batch in the work section, as well as the complete path and history of the work section's state evolution, thereby greatly improving the accuracy and reliability of intelligent analysis and decision-making.

[0037] Furthermore, step S331 also includes: performing similarity identification on trend features corresponding to the same data type in the first-order multi-scale data trend feature array and the second-order multi-scale data trend feature array to obtain a first adjacency similarity set; performing matrix processing on the first adjacency similarity set to obtain a first adjacency matrix; using the first adjacency matrix to iteratively enhance the second-order multi-scale data trend feature array to obtain a first enhanced multi-scale data trend feature array; and performing weighted fusion to obtain the high-dimensional features of the stage section.

[0038] In this embodiment, the Pearson correlation coefficient is used to identify the pairwise similarity of trend features of the same data type in the first-dimensional and second-dimensional multi-scale data trend feature arrays until all dimensions are calculated, resulting in multiple similarities. These similarities are used to quantify the change pattern of each feature dimension, thus forming a first adjacency similarity set. The first adjacency similarity set is then matrix-processed, constructing a diagonal matrix by placing each similarity score on the main diagonal of the matrix, resulting in a first adjacency matrix. Next, the first adjacency matrix is ​​used to iteratively enhance the second-dimensional multi-scale data trend feature array, i.e., performing matrix multiplication transformation operations to process the second-dimensional multi-scale data trend feature array according to its state change pattern. If the dimension change is stable, it is retained or slightly amplified; if the dimension change is drastic, it is suppressed or corrected, resulting in a first enhanced multi-scale data trend feature array. Finally, the first enhanced multi-scale data trend feature array and the first-dimensional multi-scale data trend feature array are weighted and fused to obtain the high-dimensional features of the stage segment.

[0039] Step S400: Combine the product quality constraints and high-dimensional features of multiple work sections to perform backtracking high-dimensional linkage identification, and adjust the initial drying parameters according to the identification results to obtain adjusted drying parameters. Transmit the adjusted drying parameters to the target multi-level production line for drying parameter update management.

[0040] Step S400 further includes step S410, which involves identifying anomalies in the high-dimensional features of the multiple work sections based on the product quality constraints, thereby obtaining multiple anomalous high-dimensional features of work sections; step S420, which involves performing multi-level correlation backtracking on the high-dimensional features of the multiple anomalous work sections according to the work section flow of the target multi-level production line, thereby obtaining multiple sets of correlation backtracking high-dimensional features of work sections, wherein each correlation backtracking high-dimensional feature of work sections has a correlation weight; and step S430, which involves mapping and combining the multiple anomalous high-dimensional features of work sections and the multiple sets of correlation backtracking high-dimensional features of work sections, and using the obtained multiple mapping combinations as the identification result.

[0041] In this embodiment, a backtracking high-dimensional linkage identification is performed by combining product quality constraints and high-dimensional features of multiple work sections. Specifically, the final product quality requirements are used as the standard, and the high-dimensional features of each work section are used as evidence to infer the abnormal operating sections that lead to the final product's non-compliance. This achieves accurate fault tracing and impact analysis. Specifically, anomaly identification is performed on the high-dimensional features of multiple work sections. This involves monitoring and comparing multiple final quality indicators with the product quality constraints. If the quality is acceptable, it indicates that the entire target multi-level production line is operating normally. If the quality is unacceptable, a diagnostic process is initiated. A fault diagnosis model is used for backtracking to identify the high-dimensional feature patterns of the work sections that cause the current product quality to exceed the standard. The fault diagnosis model learns the complex mapping relationship between the high-dimensional features of each work section and the final product quality through training with historical data. For example, the feature patterns of the current drying section and the deep drying section are most correlated with faults involving excessive moisture content. Finally, multiple abnormal high-dimensional features of the work sections are output.

[0042] In this embodiment, based on the process flow of the target multi-level production line, multiple abnormal high-dimensional features are back-traced back through multi-level correlation. This involves tracing back upstream along the reverse direction of the production process. For example, if the product particle size is substandard, and abnormal vibration characteristics are identified in the ultrafine grinding section, the process traces back to the deep drying section upstream of the ultrafine grinding section to verify whether the uneven moisture content of the output material causes agglomeration and adhesion. Then, it verifies whether the abnormality in the deep drying section is caused by unstable particle size in the output of the fine grinding section upstream of it. This process is repeated level by level upstream, and a correlation weight is calculated for each traced upstream section feature based on the causal relationship strength in historical data. Finally, multiple sets of high-dimensional features of the back-traced sections are output, each feature accompanied by a correlation weight. Finally, the multiple abnormal high-dimensional features and the multiple sets of high-dimensional features of the back-traced sections are mapped and combined, that is, fault phenomena and root causes are paired and combined to generate multiple complete fault diagnosis chains. These multiple mapping combinations are used as the identification results, with each combination containing the faulty section and the chain of root cause sections leading to the fault.

[0043] Furthermore, step S400 also includes step S440, traversing the multiple mapping combinations to perform segmented adaptive adjustments on the initial drying parameters to obtain multiple adjusted drying parameters; step S450, with the goal of minimizing energy consumption and maximizing quality, performing overall linkage optimization on the multiple adjusted drying parameters to determine the adjusted drying parameters.

[0044] In this embodiment, multiple mapping combinations are traversed and each combination is processed sequentially. Specifically, for each mapping combination, a preset parameter adjustment rule library is invoked to perform segment-based adaptive adjustments to the initial drying parameters. That is, the initial drying parameters are adjusted in a targeted manner according to the segments and their associated weights contained in the mapping combination results. For example, for mapping combinations where the abnormality in the ultrafine grinding segment is caused by the deep drying segment and the fine grinding segment with associated weights of 0.7 and 0.3 respectively, the feeding speed of the deep drying segment is significantly reduced, while the classifier speed of the fine grinding segment is slightly increased. For mapping combinations where the moisture content in the packaging segment exceeds the standard due to the drying segment with an associated weight of 0.8, the hot air temperature and target moisture content setting value of the drying segment are significantly increased. Thus, a parameter adjustment scheme for the drying production line is generated for each mapping combination, and multiple adjusted drying parameters are output, which can effectively solve quality problems and save energy to the maximum extent.

[0045] In this embodiment, with the goals of minimizing energy consumption and maximizing quality, multiple adjustable drying parameters are optimized in a holistic manner. Using the multi-level drying history of waste paint residue as a training dataset, a quality prediction model and an energy consumption prediction model are constructed and trained respectively. By taking the same set of drying production line parameters as input, the final product quality and the total energy consumption of the entire production line can be predicted. Then, multiple adjustable drying parameters are input into the model to output the predicted quality score and the predicted energy consumption value. Through genetic optimization or multi-objective particle swarm optimization, multi-objective optimization decision-making is performed to evaluate all schemes and determine the scheme that achieves the best balance between product quality and minimizing production energy consumption. Finally, the adjustable drying parameters are output.

[0046] In this embodiment, the drying parameters are adjusted and transmitted to the target multi-stage production line via an industrial communication network for parameter updates. This includes sending the adjusted drying parameters for different stages to the corresponding controllers and converting them into specific control signals for parameter updates. For example, the programmable logic controller (PLC) of the drying stage transmits a setpoint of 185°C to the temperature control component, which adjusts the heater power or steam flow rate via a power regulator or control valve. The PLC of the fine grinding stage transmits a setpoint of 3200 rpm to the frequency converter, precisely controlling the speed of the classifier motor. Ultimately, the entire target multi-stage production line is adjusted from the initial drying parameters to the latest optimized parameters. A multi-modal sensor array continues to monitor the execution effect and form a closed-loop feedback, thereby achieving comprehensive perception and coordinated control of the multi-stage status, ensuring improved product quality consistency and production efficiency.

[0047] Furthermore, an intelligent production management method applied to multi-level production lines also includes: adjusting the preset sampling frequency in segments based on the high-dimensional features of the multiple work sections to obtain multiple work section sampling frequencies; and sampling the production line regular window according to the multiple work section sampling frequencies, wherein the production line regular window starts at the time point when the production line adaptive window ends and ends at the time point when the production line ends production.

[0048] In this embodiment, the state patterns of high-dimensional features of multiple work sections are analyzed, and the preset sampling frequency is dynamically adjusted in segments. If the features of a work section exhibit drastic fluctuations, instability, or are on the verge of anomalies, it is determined that the work section is a key monitoring point in current production, and the sampling frequency of that work section is increased to capture data more precisely. If the features of a work section exhibit a highly stable and smooth state, it is determined that the work section is operating well, and the sampling frequency of that work section is reduced to save resources. This process generates an independent sampling frequency for each work section, resulting in multiple work section sampling frequencies. Then, the sampling of the production line's regular window is performed using these multiple work section sampling frequencies. This achieves refined management by focusing on key aspects of the process while relaxing monitoring of stable aspects, thereby significantly reducing the total amount of data collection, transmission, and storage, while saving resources and energy. The production line's regular window refers to the time period that begins at the end of the production line's adaptive window and ends at the end of the production line's production period.

[0049] In the above text, refer to Figure 1 A method for intelligent production management applied to a multi-level production line according to an embodiment of the present invention is described in detail. Next, reference will be made to... Figure 2 This invention describes an intelligent production management device applied to a multi-level production line according to an embodiment of the present invention.

[0050] According to an embodiment of the present invention, an intelligent production management device applied to a multi-level production line is used to solve the technical problems existing in the prior art, such as the inability to respond in real time to fluctuations in waste paint residue materials and changes in the status of production equipment, and the difficulty in comprehensively capturing the status of multiple sections of the entire production line, resulting in poor product quality consistency and production efficiency. The device achieves the technical effect of realizing comprehensive perception and linkage control of the status of multiple sections, thereby improving product quality consistency and production efficiency. Figure 2 As shown, an intelligent production management device for multi-level production lines includes: a sensor array acquisition module 10, a sampling data acquisition module 20, a high-dimensional feature acquisition module 30, and a drying parameter adjustment acquisition module 40.

[0051] The sensor array acquisition module 10 is used to deploy multi-modal sensors in the crushing, drying, fine grinding, deep drying, ultrafine grinding, and packaging sections of the target multi-level production line to obtain multiple section sensor arrays. The sampling data acquisition module 20 is used to acquire the waste paint residue material and product quality constraints of the target batch, start the target multi-level production line to dry the waste paint residue material according to the initial drying parameters, and use the multiple section sensor arrays to sample according to the production line adaptive window and preset sampling frequency to obtain multiple section sensor sampling data array sequences. The high-dimensional feature acquisition module 30 is used to perform segment-by-segment time-series iterative feature extraction on the multiple section sensor sampling data array sequences to obtain multiple section high-dimensional features. The drying parameter adjustment acquisition module 40 is used to perform backtracking high-dimensional linkage identification by combining the product quality constraints and multiple section high-dimensional features, and to perform segment-by-segment adaptive adjustment of the initial drying parameters according to the identification results to obtain adjusted drying parameters, and transmit the adjusted drying parameters to the target multi-level production line for drying parameter update management.

[0052] The specific configuration of the sampling data acquisition module 20 will be described in detail below. The sampling data acquisition module 20 further includes: the production line adaptive window is the time period between the start time of the crushing section and the time when the number of products completed at the packaging section meets the preset product quantity threshold.

[0053] The specific configuration of the high-dimensional feature acquisition module 30 will be described in detail below. The high-dimensional feature acquisition module 30 further includes: extracting a first section sensor sampling data array sequence from the plurality of section sensor sampling data array sequences; performing multi-scale time-series trend feature identification on the first section sensor sampling data array sequence in descending order of scale to obtain a first multi-scale data trend feature array sequence; and performing feature iteration on the first multi-scale data trend feature array sequence to obtain the high-dimensional features of the first section.

[0054] The specific configuration of the high-dimensional feature acquisition module 30 will be described in detail below. The high-dimensional feature acquisition module 30 further includes: extracting the first and second multi-scale data trend feature arrays from the first multi-scale data trend feature array sequence and performing feature iteration to obtain the high-dimensional features of the first stage segment; using the high-dimensional features of the first stage segment to perform feature iteration on the third multi-scale data trend feature array to obtain the high-dimensional features of the second stage segment; and so on, using the high-dimensional features of the stage segment obtained from the previous feature iteration to perform feature iteration on the next multi-scale data trend feature array until the last element is reached, thus obtaining the high-dimensional features of the first stage segment.

[0055] The specific configuration of the high-dimensional feature acquisition module 30 will be described in detail below. The high-dimensional feature acquisition module 30 further includes: performing similarity identification on trend features corresponding to the same data type in the first-order multi-scale data trend feature array and the second-order multi-scale data trend feature array to obtain a first adjacency similarity set; performing matrix processing on the first adjacency similarity set to obtain a first adjacency matrix; using the first adjacency matrix to iteratively enhance the second-order multi-scale data trend feature array to obtain a first enhanced multi-scale data trend feature array; and performing weighted fusion to obtain the high-dimensional features of the stage section.

[0056] The specific configuration of the drying parameter adjustment acquisition module 40 will be described in detail below. The drying parameter adjustment acquisition module 40 further includes: anomaly identification of the multiple high-dimensional features of the various work sections based on the product quality constraints, obtaining multiple abnormal high-dimensional features of the work sections; multi-level correlation backtracking of the multiple abnormal high-dimensional features of the work sections according to the work section flow of the target multi-level production line, obtaining multiple sets of correlated backtracking high-dimensional features of the work sections, wherein each correlated backtracking high-dimensional feature of the work sections has a correlation weight; mapping and combining the multiple abnormal high-dimensional features of the work sections and the multiple sets of correlated backtracking high-dimensional features of the work sections, and using the obtained multiple mapping combinations as the identification result.

[0057] The specific configuration of the drying parameter acquisition module 40 will be described in detail below. The drying parameter acquisition module 40 further includes: traversing the multiple mapping combinations to perform segmented adaptive adjustments on the initial drying parameters to obtain multiple adjusted drying parameters; and performing overall linkage optimization on the multiple adjusted drying parameters with the goal of minimizing energy consumption and maximizing quality to determine the adjusted drying parameters.

[0058] The following describes in detail the specific configuration of an intelligent production management device applied to a multi-level production line. The intelligent production management device further includes: segmenting and adjusting the preset sampling frequency based on the high-dimensional features of the multiple work sections to obtain multiple work section sampling frequencies; sampling the production line's regular window according to the multiple work section sampling frequencies, wherein the production line's regular window starts at the end of the production line's adaptive window and ends at the end of the production line's production time.

[0059] The intelligent production management device for multi-level production lines provided in this embodiment of the invention can execute the intelligent production management method for multi-level production lines provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0060] Based on the foregoing embodiments, this invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement an intelligent production management method for multi-level production lines as described in any of the preceding embodiments.

[0061] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A smart production management method applied to multi-level production lines, characterized in that, The method includes: Step S100: Deploy multi-modal sensors in the crushing section, drying section, fine grinding section, deep drying section, ultrafine grinding section and packaging section of the target multi-stage production line to obtain multiple section sensor arrays; Step S200: Obtain the waste paint residue material and product quality constraints of the target batch, start the target multi-stage production line to dry the waste paint residue material according to the initial drying parameters, and use the multiple section sensor arrays to sample according to the production line adaptive window and preset sampling frequency to obtain multiple section sensor sampling data array sequences. Step S300: Perform segment-by-segment time-series iterative feature extraction on the multiple segment sensor sampling data array sequences to obtain high-dimensional features of multiple segments; Step S400: Combine the product quality constraints and high-dimensional features of multiple work sections to perform backtracking high-dimensional linkage identification, and make work section adaptive adjustments to the initial drying parameters according to the identification results to obtain the adjusted drying parameters. Transmit the adjusted drying parameters to the target multi-level production line for parameter update management. Step S300: Perform segment-specific time-series iterative feature extraction on the multiple segment-based sensor sampling data array sequences to obtain multiple segment-specific high-dimensional features, including: Extract the first section sensor sampling data array sequence from the multiple section sensor sampling data array sequences; The first multi-scale data trend feature array sequence is obtained by performing multi-scale temporal trend feature identification on the sensor sampling data array sequence of the first section in descending order of scale. The first multi-scale data trend feature array sequence is subjected to feature iteration to obtain the high-dimensional features of the first work section. The first multi-scale data trend feature array sequence is iterated to obtain the high-dimensional features of the first work section, including: Extract the first multi-scale data trend feature array and the second multi-scale data trend feature array from the first multi-scale data trend feature array sequence, perform feature iteration, and obtain the high-dimensional features of the first stage section. The high-dimensional features of the first-stage process section are used to iterate the features of the trend feature array of the third-dimensional multi-scale data to obtain the high-dimensional features of the second-stage process section. By analogy, the high-dimensional features of the stage section obtained from the previous feature iteration are used to iterate the features of the subsequent multi-scale data trend feature array until the last position is reached, thus obtaining the high-dimensional features of the first stage. The first and second multi-scale data trend feature arrays in the first multi-scale data trend feature array sequence are extracted and iterated to obtain the high-dimensional features of the first stage section, including: Similarity identification is performed on the trend features of the same data type in the first-order multi-scale data trend feature array and the second-order multi-scale data trend feature array to obtain the first adjacency similarity set. The first adjacency similarity set is matrixed to obtain the first adjacency matrix; The first adjacency matrix is ​​used to iteratively enhance the second multi-scale data trend feature array to obtain the first enhanced multi-scale data trend feature array. After weighted fusion, the high-dimensional features of the stage section are obtained. Step S400: Combining the product quality constraints and high-dimensional features of multiple work sections, perform backtracking high-dimensional linkage identification, and based on the identification results, make work section-specific adaptive adjustments to the initial drying parameters to obtain adjusted drying parameters, including: Based on the product quality constraints, anomaly identification is performed on the high-dimensional features of the multiple work sections to obtain the high-dimensional features of multiple abnormal work sections. Based on the process flow of the target multi-level production line, multi-level correlation backtracking is performed on the high-dimensional features of the multiple abnormal process sections to obtain multiple sets of high-dimensional features of the correlated backtracking process sections, wherein each high-dimensional feature of the correlated backtracking process section has a correlation weight. The high-dimensional features of the multiple abnormal work sections and the high-dimensional feature sets of the multiple associated backtracking work sections are mapped and combined, and the resulting multiple mapping combinations are used as the identification results.

2. The intelligent production management method applied to multi-level production lines as described in claim 1, characterized in that, The production line adaptive window is the time period between the start time of the crushing section and the time when the number of products completed at the packaging section meets the preset product quantity threshold.

3. The intelligent production management method applied to multi-level production lines as described in claim 1, characterized in that, Also includes: The initial drying parameters are adjusted in stages by traversing the multiple mapping combinations to obtain multiple adjusted drying parameters; With the goal of minimizing energy consumption and maximizing quality, the multiple adjustable drying parameters are optimized in a holistic manner to determine the adjusted drying parameters.

4. The intelligent production management method applied to multi-level production lines as described in claim 1, characterized in that, Based on the high-dimensional features of the multiple work sections, the preset sampling frequency is adjusted in segments to obtain multiple work section sampling frequencies; The production line's regular window is sampled according to the sampling frequency of the multiple work sections. The regular window of the production line starts at the time point when the production line's adaptive window ends and ends at the time point when the production line finishes production.

5. An intelligent production management device applied to multi-level production lines, characterized in that, The apparatus is used to implement the intelligent production management method for multi-level production lines as described in any one of claims 1 to 4, the apparatus comprising: The sensor array acquisition module is used to deploy multi-modal sensors in the crushing section, drying section, fine grinding section, deep drying section, ultrafine grinding section and packaging section of the target multi-stage production line to obtain multiple section sensor arrays. The sampling data acquisition module is used to acquire the waste paint residue material and product quality constraints of the target batch, start the target multi-stage production line to dry the waste paint residue material according to the initial drying parameters, and use the multiple section sensor arrays to sample according to the production line adaptive window and preset sampling frequency to obtain multiple section sensor sampling data array sequences. The high-dimensional feature acquisition module is used to perform segment-by-segment time-series iterative feature extraction on the multiple segment sensor sampling data array sequences to obtain high-dimensional features of multiple segments. The module for obtaining adjusted drying parameters is used to perform backtracking high-dimensional linkage identification by combining the product quality constraints and high-dimensional features of multiple work sections, and to make adaptive adjustments to the initial drying parameters according to the identification results, thereby obtaining adjusted drying parameters. The adjusted drying parameters are then transmitted to the target multi-level production line for drying parameter update management.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements an intelligent production management method for multi-level production lines as described in any one of claims 1-4.

Images