Data-driven workshop production equipment flow intelligent regulation and control method

Abstract

The invention relates to the technical field of intelligent regulation and control, and discloses a data-driven workshop production equipment flow intelligent regulation and control method. Comprising the steps of S1, equipment state data acquisition in a data acquisition stage, S2, data cleaning abnormal value processing in a data preprocessing stage, S3, equipment health degree analysis model in a data analysis and modeling stage, S4, equipment scheduling decision in an intelligent regulation and control decision stage, and S5, regulation and control effect feedback and effect evaluation index production efficiency in an iteration stage. According to the data-driven workshop production equipment flow intelligent regulation and control method, the data acquisition and processing efficiency is improved, the integrity of equipment state data is improved to 98% through multi-dimensional sensor deployment (vibration 100Hz sampling and temperature 30s / time), the time consumed for automatic correction of an abnormal value is shortened from 30min to 1min, and the efficiency of automatic correction of the abnormal value is improved. And after data standardization, the multi-dimensional fusion analysis efficiency is improved by 4 times, and a foundation is laid for accurate regulation and control.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and specifically to a data-driven intelligent control method for workshop production equipment processes. Background Art

[0002] Driven by Industry 4.0 and the wave of smart manufacturing, workshop production is transforming from automation to digitalization and intelligence. Currently, data, as a new production factor, has been deeply integrated into core processes such as equipment management and process control. However, issues such as data fragmentation and lags in control under the traditional manufacturing model have become key bottlenecks hindering improvements in workshop production efficiency and quality. According to statistics, global manufacturing capacity losses due to equipment downtime exceed $200 billion annually. Inefficient process parameter optimization has led to a long-term stagnation in product qualification rates at 3%-5%. There is an urgent need to upgrade production processes through data-driven technologies to achieve intelligent control.

[0003] Insufficient data collection and processing capabilities Traditional methods rely heavily on manual inspections or single-point sensors to collect equipment status data. These methods suffer from low sampling frequencies (e.g., only 1-2 times per hour) and limited data dimensions (covering only basic parameters like temperature and current). This makes it impossible to capture micro-fault signals (e.g., abnormal vibrations that indicate early bearing wear) in real time. Furthermore, production process data (e.g., material flow and process parameters) is stored in separate systems (MES, PLC) without unified standardization. This makes data fusion analysis difficult and leads to delayed correction of outliers (e.g., false temperature sensor alarms require manual troubleshooting).

[0004] Low accuracy of equipment health assessment and production capacity prediction Traditional health assessment methods based on rule engines (such as setting fixed thresholds) cannot adapt to dynamic scenarios such as equipment aging and changing operating conditions. Fault warning accuracy is less than 60%, and false positives and omissions are common. Capacity forecasts often rely on empirical formulas and fail to integrate multi-dimensional data such as equipment health and material supply. Responses to unexpected orders or equipment downtime can be delayed by more than four hours, leading to delayed identification of capacity gaps and an on-time delivery rate of less than 85%. Summary of the Invention

[0005] (1) Technical problems solved In response to the shortcomings of the existing technology, the present invention provides a data-driven intelligent control method for workshop production equipment processes, which has the advantages of improving data collection and processing efficiency, and solves the problem of insufficient data collection and processing capabilities.

[0006] (2) Technical solution In order to achieve the above-mentioned purpose of improving data collection and processing efficiency, the present invention provides the following technical solutions: a data-driven intelligent control method for workshop production equipment processes, including equipment status data collection in the S1 data collection stage, data cleaning and outlier processing in the S2 data preprocessing stage, equipment health analysis model in the S3 data analysis and modeling stage, equipment scheduling decision-making in the S4 intelligent control decision-making stage, and effect evaluation index production efficiency in the S5 control effect feedback and iteration stage. The equipment status data collection in the S1 data collection stage includes S101 production process data collection and production task data; Among them, the data cleaning and outlier processing in the S2 data preprocessing stage includes S201 data standardization; Among them, the equipment health analysis model in the S3 data analysis and modeling stage includes S301 training set containing historical equipment failure data, S302 production process optimization model capacity prediction and S303 process parameter optimization; Among them, the equipment scheduling decision in the S4 intelligent control decision stage includes S401 production parameter control, S402 abnormal warning and intervention equipment abnormal warning; Among them, the production efficiency of S5 regulation effect feedback and iteration stage effect evaluation indicators includes S501 model iterative optimization.

[0007] Preferably, the device status data is collected during the S1 data collection phase: Deploy sensors at key locations on workshop production equipment (such as processing machine tools, conveyor lines, and storage facilities). These sensors include vibration sensors that collect vibration data at a 100Hz sampling frequency to monitor equipment operational stability. The threshold is set at 0.5g (g is the acceleration due to gravity). If this threshold is exceeded, an early warning of potential equipment failure will be issued. Temperature sensors collect temperature data on the device's motors, transmission components, and other components every 30 seconds. The temperature threshold varies depending on the device type. For example, a machine tool spindle temperature of ≤60°C triggers cooling or load reduction instructions if the temperature exceeds 60°C. The current sensor collects the equipment's operating current in real time with an accuracy of 0.1A. This is used to determine the equipment's load condition. The difference between no-load and full-load current is set to ≥30% of the rated current. If the difference is abnormal, the rationality of production task allocation is analyzed. Through the industrial Internet of Things gateway, the collected data is uploaded to the edge server every 1 minute using the MQTT protocol for temporary storage.

[0008] Preferably, the S101 production process data collection production task data: obtained from the production management system (MES), including order number, product model, production quantity, process requirements (such as processing accuracy ±0.02mm, assembly sequence, etc.), delivery cycle, etc., and synchronized to the data center every 5 minutes; Material flow data: RFID readers are installed at the exit of the material warehouse and the entrance of the production station. Combined with the station barcode scanner, they record material batches, entry time, and usage quantity, and update the material location and consumption status at a frequency of 1 second. Process parameter data: collected from the equipment control system (such as PLC), covering processing speed (such as machine tool cutting speed adjustable from 50 to 200m / min), pressure (hydraulic equipment pressure 0 to 10MPa, etc.), time (such as welding time 5 to 30s), etc., with a sampling period of 100 milliseconds to ensure process execution accuracy monitoring.

[0009] Preferably, the S2 data preprocessing stage includes data cleaning and outlier processing: Using the statistically based 3σ principle, we screen equipment status data such as vibration, temperature, and current, as well as production data such as processing speed and material consumption, and eliminate data points that deviate from the mean by three standard deviations. For example, if a machine tool temperature sensor falsely reports 100°C (normally ≤60°C), it is identified as an outlier and corrected. The correction strategy is to take the mean of the five adjacent normal data points as the replacement. Missing value filling: For missing data such as material flow and process parameters, if it is continuous data (such as temperature and current), linear interpolation is used to fill the missing values. Missing values ​​with a time interval of ≤5 minutes are filled by fitting the curve of the data before and after. If it is discrete data (such as material batches and order numbers), it is supplemented by retrospective analysis from related data sources (such as MES systems and warehouse management systems) based on the production process logic (such as material sequence and order association relationships).

[0010] Preferably, the S201 data is standardized: Normalize the equipment status and production process data using the min-max normalization method to map the data to the [0,1] interval. The formula is: \(x_{norm} = \frac{x - x_{min}}{x_{max} - x_{min}}\); Where x represents the raw data, and x_{min} and x_{max} represent the minimum and maximum values ​​of that data dimension. For example, equipment vibration acceleration data originally ranges from 0 to 1g, but after normalization, it corresponds to 0 to 1g, facilitating subsequent multi-dimensional data fusion analysis.

[0011] Preferably, the equipment health analysis model in the S3 data analysis and modeling phase: Build a machine learning-based equipment health assessment model, input pre-processed vibration, temperature, and current data, and use the random forest algorithm to classify equipment health status into five levels (healthy, sub-healthy, warning, faulty, and severe faulty). The S301 training set contains historical equipment failure data (labeled with fault type, occurrence time, and pre-fault equipment status parameters) and normal operating data. During training, the number of decision trees was set to 50, with a maximum depth of 8 layers. Vibration frequency, temperature change rate, and current fluctuation values ​​were selected as key features through feature importance analysis. The model outputs an equipment health index (0-100, with higher values ​​indicating healthier equipment). A fault warning process is triggered when the index is ≤60.

[0012] Preferably, the production process optimization model capacity prediction in S302 is based on an LSTM neural network, which inputs historical production task data (order quantity, product type), equipment status data (health, operating time), and material supply data (inventory, delivery cycle), and predicts the workshop capacity for the next 24 hours, 7 days, and 30 days. The time step is set to 1 hour, the number of hidden layer neurons is 64, and the capacity prediction curve is output. The curve is compared with the order demand to identify capacity gaps or excesses. S303 process parameter optimization: Using genetic algorithms, with product qualification rate (target value ≥ 99%), production energy consumption (target value ≤ 10% of industry standard), and production cycle (target value ≤ standard cycle) as optimization targets, process parameters such as processing speed, pressure, and time are optimized; Among them, the equipment health analysis model in the S3 data analysis and modeling stage sets the population size to 50, the genetic generation to 100 generations, the crossover probability to 0.8, and the mutation probability to 0.1, and outputs the optimal process parameter combination adapted to different product models and equipment status. For example, for the processing of a certain model of parts, the optimized cutting speed is 120m / min, the pressure is 5MPa, and the processing time is 15s, which increases the pass rate by 2% and reduces energy consumption by 5% compared with the original parameters.

[0013] Preferably, the equipment scheduling decision in the S4 intelligent control decision stage is: Based on the output of the equipment health analysis model and combined with production capacity forecast results, an equipment scheduling strategy is formulated: Equipment with a health score of 80 or higher is prioritized for high-load, high-precision production tasks, such as key parts processing. Equipment with a health score of 60-80 is scheduled for routine production tasks, and preventive maintenance plans are initiated simultaneously (such as increasing inspection frequency and replacing wearing parts). Equipment with a health score of less than 60 has production tasks suspended, triggering the fault repair process. During the repair period, backup equipment is deployed through the equipment ledger or the production schedule is adjusted to ensure order delivery. S401 production parameter control: The optimal process parameters output by the production process optimization model are dynamically adjusted through the equipment control system (PLC or DCS). For equipment that supports real-time parameter adjustment (such as CNC machine tools), the optimized processing speed, pressure and other parameters are automatically sent to the equipment controller in a 1-minute cycle, achieving adaptive adjustment of process parameters. For equipment that requires downtime for adjustments (such as some older production lines), operations and maintenance personnel will manually adjust the parameters based on the parameter optimization results during the production task switching period (such as shift handover and order completion). After the adjustment, the parameter change information will be fed back to the system via a barcode scanner or data entry terminal to update the production process data. S402 Abnormal Warning and Intervention Equipment abnormal warning: When the equipment health index is ≤60 or the sensor monitoring data exceeds the threshold (such as vibration acceleration >0.5g, temperature >60°C), the system automatically triggers a three-level warning (warning, fault, serious fault) and pushes it to the operation and maintenance personnel through the workshop LED screen and mobile app. The warning information includes the equipment number, fault type (such as abnormal vibration, excessive temperature), and recommended maintenance measures (such as checking bearings and cleaning heat dissipation channels). Operation and maintenance personnel must respond within 10 minutes and arrive at the site within 30 minutes to handle the problem. Production abnormality intervention: If there is a material shortage (inventory ≤ 10% of safety stock), a production capacity gap > 10%, or a product qualification rate < 95%, the system initiates an abnormality intervention process, including: automatically triggering an emergency material purchase request (linking to the supplier management system to push the purchase order to the preferred supplier), adjusting production task allocation (such as transferring some tasks to idle equipment or outsourcing), and re-optimizing process parameters (shortening the genetic algorithm optimization cycle to 30 generations and quickly iterating parameters) to ensure continuous and stable production processes.

[0014] Preferably, the S5 control effect feedback and iterative stage effect evaluation index production efficiency: Calculate the output per unit time (e.g. pieces / hour) and the ratio of effective equipment operating time (effective operating time / planned operating time × 100%), with target values ​​of increasing by 10% and ≥90% respectively; Product quality: Calculate the product qualification rate (qualified product quantity / total production quantity × 100%) and the distribution of defective product types (such as dimensional deviation and surface defect ratio). The target value is a qualified rate of ≥99% and a defective product ratio of ≤1%. Energy consumption cost: Calculate unit product energy consumption (kWh / unit) and equipment energy consumption reduction rate ((original energy consumption - current energy consumption) / original energy consumption × 100%). The target value is to reduce unit energy consumption by 8% and the equipment energy consumption reduction rate by ≥5%.

[0015] Preferably, the S501 model is iteratively optimized: Every seven days, data on control effectiveness (production efficiency, product quality, and energy costs) is collected and compared with the model's predicted values. The error rate is calculated (|actual value - predicted value| / predicted value × 100%). If the error rate is greater than 15%, the equipment health analysis model and production process optimization model are iteratively updated: new failure case data and production anomaly data are added to the training set, the model is retrained, and parameters such as the number of decision trees in the random forest algorithm and the number of hidden layer neurons in the LSTM neural network are adjusted. Optimize the fitness function of the genetic algorithm and incorporate new evaluation indicators (such as equipment coordination efficiency and material delivery timeliness) to improve the model's adaptability to complex production scenarios and ensure that the control method continues to fit the actual production needs of the workshop.

[0016] (3) Beneficial effects Compared with the existing technology, the present invention provides a data-driven intelligent control method for workshop production equipment processes, which has the following beneficial effects: 1. This data-driven intelligent control method for workshop production equipment processes improves data collection and processing efficiency. Multi-dimensional sensor deployment (vibration sampling at 100Hz and temperature sampling at 30 seconds / time) increases the integrity of equipment status data to 98%. The time required for automatic correction of outliers is reduced from 30 minutes to 1 minute. After data standardization, the efficiency of multi-dimensional fusion analysis increases by 4 times, laying the foundation for precise control.

[0017] 2. This data-driven intelligent control method for workshop production equipment processes has achieved breakthroughs in the accuracy of equipment health and capacity prediction. The health model based on random forests has increased the accuracy of fault warnings to 92%, the warning response time is ≤10 minutes, and the early fault identification rate has increased by 3 times. The LSTM capacity prediction model has an error rate of ≤8% for 24-hour capacity, which is a 50% improvement compared to traditional empirical formulas. Capacity gaps can be identified in advance to 4 hours, and the on-time delivery rate of orders has increased to more than 95%.

[0018] 3. This data-driven intelligent control method for workshop production equipment processes achieves automated control decision-making and collaborative optimization. The equipment scheduling strategy ensures a 90% match rate between equipment with a health level ≥80 and high-load tasks, increasing the effective operating time of equipment from 75% to 92%. Genetic algorithm process parameter optimization shortens single optimization time to 15 minutes, increases the pass rate by 2% (to 99%), reduces energy consumption by 5%, and achieves "zero trial and error" parameter adjustment. The abnormal intervention process shortens the response time to material shortages from 8 hours to 2 hours, reducing the risk of production line shutdown by 70%.

[0019] 4. This data-driven intelligent control method for workshop production equipment processes implements model iteration and continuous optimization capabilities. The model is automatically iterated every 7 days, and an emergency update is triggered when the error rate is greater than 15%. The adaptation period of the equipment health model to new fault types is shortened from 3 months to 1 week, and effect feedback data is synchronized in real time (delay ≤ 5 minutes). Production efficiency is increased by 12% (target 10%), unit energy consumption is reduced by 9% (exceeding the target by 1%), and the proportion of defective products is controlled at 0.8% (≤ 1%), achieving continuous optimization of production indicators.

[0020] 5. This data-driven intelligent control method for workshop production equipment processes achieves system compatibility and flexible expansion. It is compatible with data access from heterogeneous systems such as PLC and MES, supports 10+ types of sensor protocols such as vibration and current, and has a model parameter adjustment cycle of ≤1 day when adding new equipment or processes. It meets the needs of rapid switching in multi-variety, small-batch production scenarios and provides a universal solution for the digital transformation of workshops. DETAILED DESCRIPTION

[0021] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0022] This solution provides a technical solution, specifically, a data-driven intelligent control method for workshop production equipment processes, including the following control methods: Equipment status data collection in the S1 data collection phase: Deploy sensors at key locations on workshop production equipment (such as processing machine tools, conveyor lines, and storage facilities). These sensors include vibration sensors that collect vibration data at a 100Hz sampling frequency to monitor equipment operational stability. The threshold is set at 0.5g (g is the acceleration due to gravity). If this threshold is exceeded, an early warning of potential equipment failure will be issued. Temperature sensors collect temperature data on the device's motors, transmission components, and other components every 30 seconds. The temperature threshold varies depending on the device type. For example, a machine tool spindle temperature of ≤60°C triggers cooling or load reduction instructions if the temperature exceeds 60°C. The current sensor collects the equipment's operating current in real time with an accuracy of 0.1A. This is used to determine the equipment's load condition. The difference between no-load and full-load current is set to ≥30% of the rated current. If the difference is abnormal, the rationality of production task allocation is analyzed. Through the industrial IoT gateway, the collected data is uploaded to the edge server every 1 minute using the MQTT protocol for temporary storage; S101 Production Process Data Collection Production task data: obtained from the production management system (MES), including order number, product model, production quantity, process requirements (such as processing accuracy ±0.02mm, assembly sequence, etc.), delivery cycle, etc., and synchronized to the data center every 5 minutes; Material flow data: RFID readers are installed at the exit of the material warehouse and the entrance of the production station. Combined with the station barcode scanner, they record material batches, entry time, and usage quantity, and update the material location and consumption status at a frequency of 1 second. Process parameter data: collected from the equipment control system (such as PLC), covering processing speed (such as machine tool cutting speed adjustable from 50 to 200 m / min), pressure (hydraulic equipment pressure 0 to 10 MPa, etc.), time (such as welding time 5 to 30 seconds), etc., with a sampling period of 100 milliseconds to ensure process execution accuracy monitoring; S2 data preprocessing stage data cleaning outlier processing: Using the statistically based 3σ principle, we screen equipment status data such as vibration, temperature, and current, as well as production data such as processing speed and material consumption, and eliminate data points that deviate from the mean by three standard deviations. For example, if a machine tool temperature sensor falsely reports 100°C (normally ≤60°C), it is identified as an outlier and corrected. The correction strategy is to take the mean of the five adjacent normal data points as the replacement. Missing value filling: For missing data on material flow and process parameters, if the data is continuous (such as temperature and current), linear interpolation is used to fill the gaps. Missing values ​​with a time interval of ≤5 minutes are filled by fitting the curve of the data before and after. If the data is discrete (such as material batches and order numbers), the gaps are filled by retrospectively searching related data sources (such as MES systems and warehouse management systems) based on the production process logic (such as the order of materials and the relationship between orders). S201 Data Standardization: Normalize the equipment status and production process data using the min-max normalization method to map the data to the [0,1] interval. The formula is: \(x_{norm} = \frac{x - x_{min}}{x_{max} - x_{min}}\); Where x is the raw data, and x_{min} and x_{max} are the minimum and maximum values ​​of the data dimension. For example, the original range of equipment vibration acceleration data is 0-1g, which is normalized to 0-1, facilitating subsequent multi-dimensional data fusion analysis. Equipment health analysis model in the S3 data analysis and modeling phase: Build a machine learning-based equipment health assessment model, input pre-processed vibration, temperature, and current data, and use the random forest algorithm to classify equipment health status into five levels (healthy, sub-healthy, warning, faulty, and severe faulty). The S301 training set contains historical equipment failure data (labeled with fault type, occurrence time, and pre-fault equipment status parameters) and normal operating data. During training, the number of decision trees was set to 50, with a maximum depth of 8 layers. Vibration frequency, temperature change rate, and current fluctuation values ​​were selected as key features through feature importance analysis. The model outputs an equipment health index (0-100, with higher values ​​indicating healthier equipment). A failure warning process is triggered when the index is ≤60. S302 Production Process Optimization Model Capacity Forecast: Based on an LSTM neural network, the model inputs historical production task data (order quantity, product type), equipment status data (health, operating hours), and material supply data (inventory, delivery cycle). It predicts workshop capacity for the next 24 hours, 7 days, and 30 days. The time step is set to 1 hour, the number of hidden layer neurons is 64, and the output capacity forecast curve is compared with order demand to identify capacity gaps or excesses. S303 process parameter optimization: Using genetic algorithms, with product qualification rate (target value ≥ 99%), production energy consumption (target value ≤ 10% of industry standard), and production cycle (target value ≤ standard cycle) as optimization targets, process parameters such as processing speed, pressure, and time are optimized; The equipment health analysis model in the S3 data analysis and modeling phase sets a population size of 50, a genetic generation of 100, a crossover probability of 0.8, and a mutation probability of 0.1. It outputs the optimal process parameter combination for different product models and equipment states. For example, for a certain part model, the optimized cutting speed is 120 m / min, the pressure is 5 MPa, and the processing time is 15 s, which increases the pass rate by 2% and reduces energy consumption by 5% compared to the original parameters. Equipment scheduling decision-making in the S4 intelligent control decision-making stage: Based on the output of the equipment health analysis model and combined with production capacity forecast results, an equipment scheduling strategy is formulated: Equipment with a health score of 80 or higher is prioritized for high-load, high-precision production tasks, such as key parts processing. Equipment with a health score of 60-80 is scheduled for routine production tasks, and preventive maintenance plans are initiated simultaneously (such as increasing inspection frequency and replacing wearing parts). Equipment with a health score of less than 60 has production tasks suspended, triggering the fault repair process. During the repair period, backup equipment is deployed through the equipment ledger or the production schedule is adjusted to ensure order delivery. S401 production parameter control: The optimal process parameters output by the production process optimization model are dynamically adjusted through the equipment control system (PLC or DCS). For equipment that supports real-time parameter adjustment (such as CNC machine tools), the optimized processing speed, pressure and other parameters are automatically sent to the equipment controller in a 1-minute cycle, achieving adaptive adjustment of process parameters. For equipment that requires downtime for adjustments (such as some older production lines), operations and maintenance personnel will manually adjust the parameters based on the parameter optimization results during the production task switching period (such as shift handover and order completion). After the adjustment, the parameter change information will be fed back to the system via a barcode scanner or data entry terminal to update the production process data. S402 Abnormal Warning and Intervention Equipment abnormal warning: When the equipment health index is ≤60 or the sensor monitoring data exceeds the threshold (such as vibration acceleration >0.5g, temperature >60°C), the system automatically triggers a three-level warning (warning, fault, serious fault) and pushes it to the operation and maintenance personnel through the workshop LED screen and mobile app. The warning information includes the equipment number, fault type (such as abnormal vibration, excessive temperature), and recommended maintenance measures (such as checking bearings and cleaning heat dissipation channels). Operation and maintenance personnel must respond within 10 minutes and arrive at the site within 30 minutes to handle the problem. Production Abnormal Intervention: If there is a material shortage (inventory ≤ 10% of safety stock), a production capacity gap > 10%, or a product qualification rate < 95%, the system initiates the abnormal intervention process, including: automatically triggering an emergency material purchase request (linking to the supplier management system to push the purchase order to the preferred supplier), adjusting production task allocation (such as transferring some tasks to idle equipment or outsourcing), and re-optimizing process parameters (shortening the genetic algorithm optimization cycle to 30 generations for rapid parameter iteration) to ensure continuous and stable production processes; S5 Control Effect Feedback and Iteration Phase Effect Evaluation Index Production Efficiency: Calculate the output per unit time (e.g. pieces / hour) and the ratio of effective equipment operating time (effective operating time / planned operating time × 100%), with target values ​​of increasing by 10% and ≥90% respectively; Product quality: Calculate the product qualification rate (qualified product quantity / total production quantity × 100%) and the distribution of defective product types (such as dimensional deviation and surface defect ratio). The target value is a qualified rate of ≥99% and a defective product ratio of ≤1%. Energy consumption cost: Calculate unit product energy consumption (kWh / unit) and equipment energy consumption reduction rate ((original energy consumption - current energy consumption) / original energy consumption × 100%). The target value is to reduce unit energy consumption by 8% and the equipment energy consumption reduction rate by ≥5%. S501 model iterative optimization: Every seven days, data on control effectiveness (production efficiency, product quality, and energy costs) is collected and compared with the model's predicted values. The error rate is calculated (|actual value - predicted value| / predicted value × 100%). If the error rate is greater than 15%, the equipment health analysis model and production process optimization model are iteratively updated: new failure case data and production anomaly data are added to the training set, the model is retrained, and parameters such as the number of decision trees in the random forest algorithm and the number of hidden layer neurons in the LSTM neural network are adjusted. Optimize the fitness function of the genetic algorithm and incorporate new evaluation indicators (such as equipment coordination efficiency and material delivery timeliness) to improve the model's adaptability to complex production scenarios and ensure that the control method continues to fit the actual production needs of the workshop.

[0023] Furthermore, this solution improves data collection and processing efficiency. Multi-dimensional sensor deployment (vibration sampling at 100Hz and temperature sampling at 30 seconds / time) increases the integrity of equipment status data to 98%. Automatic correction of outliers is reduced from 30 minutes to 1 minute. After data standardization, the efficiency of multi-dimensional fusion analysis increases fourfold, laying the foundation for precise control. Furthermore, this solution achieved breakthroughs in equipment health and capacity prediction accuracy. The health model based on random forests increased fault warning accuracy to 92%, with a warning response time of ≤10 minutes and a threefold increase in early fault identification rate. The LSTM capacity prediction model achieved an error rate of ≤8% for 24-hour capacity, a 50% improvement over traditional empirical formulas. Capacity gaps can be identified up to 4 hours earlier, and the on-time delivery rate for orders has increased to over 95%. Furthermore, this solution achieves automated control and coordination optimization. The equipment scheduling strategy ensures that equipment with a health score of 80 or higher is matched with high-load tasks 90% of the time, increasing the effective operating time of equipment from 75% to 92%. Genetic algorithm process parameter optimization reduces single-time optimization time to 15 minutes, increases the pass rate by 2% (to 99%), reduces energy consumption by 5%, and enables "trial-and-error" parameter adjustment. The abnormal intervention process reduces the response time to material shortages from 8 hours to 2 hours, reducing the risk of production line downtime by 70%. Furthermore, this solution enables model iteration and continuous optimization. The model is automatically iterated every seven days, with an emergency update triggered when the error rate exceeds 15%. The equipment health model's adaptation cycle for new fault types has been shortened from three months to one week, with real-time feedback data synchronized (delay ≤ 5 minutes). This has resulted in a 12% increase in production efficiency (target 10%), a 9% decrease in unit energy consumption (exceeding the target by 1%), and a 0.8% defective product ratio (≤ 1%), enabling continuous optimization of production indicators. Furthermore, this solution achieves system compatibility and flexible expansion, is compatible with data access from heterogeneous systems such as PLC and MES, supports 10+ types of sensor protocols such as vibration and current, and has a model parameter adjustment cycle of ≤1 day when adding new equipment or processes. It meets the rapid switching needs of multi-variety and small-batch production scenarios and provides a universal solution for the digital transformation of workshops.

[0024] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A data-driven intelligent control method for workshop production equipment processes includes the following: S1: data collection phase for equipment status data collection; S2: data preprocessing phase for data cleaning and outlier processing; S3: data analysis and modeling phase for equipment health analysis model; S4: intelligent control decision phase for equipment scheduling decision; and S5: control effect feedback and iteration phase for effect evaluation of production efficiency indicators. The method is characterized by: The equipment status data collection in the S1 data collection phase includes S101 production process data collection and production task data; Among them, the data cleaning and outlier processing in the S2 data preprocessing stage includes S201 data standardization; Among them, the equipment health analysis model in the S3 data analysis and modeling stage includes S301 training set containing historical equipment failure data, S302 production process optimization model capacity prediction and S303 process parameter optimization; Among them, the equipment scheduling decision in the S4 intelligent control decision stage includes S401 production parameter control, S402 abnormal warning and intervention equipment abnormal warning; Among them, the production efficiency of S5 regulation effect feedback and iteration stage effect evaluation indicators includes S501 model iterative optimization.

2. The data-driven intelligent control method for workshop production equipment processes according to claim 1 is characterized by: The S1 data collection phase collects device status data: Deploy sensors at key locations on workshop production equipment (such as processing machine tools, conveyor lines, and storage facilities). These sensors include vibration sensors that collect vibration data at a 100Hz sampling frequency to monitor equipment operational stability. The threshold is set at 0.5g (g is the acceleration due to gravity). If this threshold is exceeded, an early warning of potential equipment failure will be issued. Temperature sensors collect temperature data on the device's motors, transmission components, and other components every 30 seconds. The temperature threshold varies depending on the device type. For example, a machine tool spindle temperature of ≤60°C triggers cooling or load reduction instructions if the temperature exceeds 60°C. The current sensor collects the equipment's operating current in real time with an accuracy of 0.1A. This is used to determine the equipment's load condition. The difference between no-load and full-load current is set to ≥30% of the rated current. If the difference is abnormal, the rationality of production task allocation is analyzed. Through the industrial Internet of Things gateway, the collected data is uploaded to the edge server every 1 minute using the MQTT protocol for temporary storage.

3. The data-driven intelligent control method for workshop production equipment processes according to claim 1 is characterized by: The S101 production process data collection production task data: obtained from the production management system (MES), including order number, product model, production quantity, process requirements (such as processing accuracy ±0.02mm, assembly sequence, etc.), delivery cycle, etc., and synchronized to the data center every 5 minutes; Material flow data: RFID readers are installed at the exit of the material warehouse and the entrance of the production station. Combined with the station barcode scanner, they record material batches, entry time, and usage quantity, and update the material location and consumption status at a frequency of 1 second. Process parameter data: collected from the equipment control system (such as PLC), covering processing speed (such as machine tool cutting speed adjustable from 50-200m / min), pressure (hydraulic equipment pressure 0-10MPa, etc.), time (such as welding time 5-30s), etc., with a sampling period of 100 milliseconds to ensure process execution accuracy monitoring.

4. The data-driven intelligent control method for workshop production equipment processes according to claim 1 is characterized by: In the S2 data preprocessing stage, data cleaning and outlier processing are performed: Using the statistically based 3σ principle, we screen equipment status data such as vibration, temperature, and current, as well as production data such as processing speed and material consumption, and eliminate data points that deviate from the mean by three standard deviations. For example, if a machine tool temperature sensor falsely reports 100°C (normally ≤60°C), it is identified as an outlier and corrected. The correction strategy is to take the mean of the five adjacent normal data points as the replacement. Missing value filling: For missing data such as material flow and process parameters, if it is continuous data (such as temperature and current), linear interpolation is used to fill the missing values. Missing values ​​with a time interval of ≤5 minutes are filled by fitting the curve of the data before and after. If it is discrete data (such as material batches and order numbers), it is supplemented by retrospective analysis from related data sources (such as MES systems and warehouse management systems) based on the production process logic (such as material sequence and order association relationships).

5. The data-driven intelligent control method for workshop production equipment processes according to claim 1 is characterized by: The S201 data standardization: Normalize the equipment status and production process data using the min-max normalization method to map the data to the [0,1] interval. The formula is: \(x_{norm} = \frac{x - x_{min}}{x_{max} - x_{min}}\); Where x represents the raw data, and x_{min} and x_{max} represent the minimum and maximum values ​​of that data dimension. For example, equipment vibration acceleration data originally ranges from 0 to 1g, but after normalization, it corresponds to 0 to 1g, facilitating subsequent multi-dimensional data fusion analysis.

6. The data-driven intelligent control method for workshop production equipment processes according to claim 1 is characterized by: The equipment health analysis model in the S3 data analysis and modeling phase: Build a machine learning-based equipment health assessment model, input pre-processed vibration, temperature, and current data, and use the random forest algorithm to classify equipment health status into five levels (healthy, sub-healthy, warning, faulty, and severe faulty). The S301 training set contains historical equipment failure data (labeled with fault type, occurrence time, and pre-fault equipment status parameters) and normal operating data. During training, the number of decision trees was set to 50, with a maximum depth of 8 layers. Vibration frequency, temperature change rate, and current fluctuation values ​​were selected as key features through feature importance analysis. The model outputs an equipment health index (0-100, with higher values ​​indicating healthier equipment). A fault warning process is triggered when the index is ≤60.

7. The data-driven intelligent control method for workshop production equipment processes according to claim 1 is characterized by: The S302 production process optimization model capacity prediction is based on an LSTM neural network. It inputs historical production task data (order quantity, product type), equipment status data (health, operating hours), and material supply data (inventory, delivery cycle). It predicts workshop capacity for the next 24 hours, 7 days, and 30 days. The time step is set to 1 hour, the number of hidden layer neurons is 64, and the output capacity prediction curve is compared with order demand to identify capacity gaps or excesses. S303 process parameter optimization: Using genetic algorithms, with product qualification rate (target value ≥ 99%), production energy consumption (target value ≤ 10% of industry standard), and production cycle (target value ≤ standard cycle) as optimization targets, process parameters such as processing speed, pressure, and time are optimized; Among them, the equipment health analysis model in the S3 data analysis and modeling stage sets the population size to 50, the genetic generation to 100 generations, the crossover probability to 0.8, and the mutation probability to 0.1, and outputs the optimal process parameter combination adapted to different product models and equipment status. For example, for the processing of a certain model of parts, the optimized cutting speed is 120m / min, the pressure is 5MPa, and the processing time is 15s, which increases the pass rate by 2% and reduces energy consumption by 5% compared with the original parameters.

8. The data-driven intelligent control method for workshop production equipment processes according to claim 1 is characterized by: Equipment scheduling decision in the S4 intelligent control decision stage: Based on the output of the equipment health analysis model and combined with production capacity forecast results, an equipment scheduling strategy is formulated: Equipment with a health score of 80 or higher is prioritized for high-load, high-precision production tasks, such as key parts processing. Equipment with a health score of 60-80 is scheduled for routine production tasks, and preventive maintenance plans are initiated simultaneously (such as increasing inspection frequency and replacing wearing parts). Equipment with a health score of less than 60 has production tasks suspended, triggering the fault repair process. During the repair period, backup equipment is called up through the equipment ledger or the production schedule is adjusted to ensure order delivery. S401 production parameter control: The optimal process parameters output by the production process optimization model are dynamically adjusted through the equipment control system (PLC or DCS). For equipment that supports real-time parameter adjustment (such as CNC machine tools), the optimized processing speed, pressure and other parameters are automatically sent to the equipment controller in a 1-minute cycle, achieving adaptive adjustment of process parameters. For equipment that requires downtime for adjustments (such as some older production lines), operations and maintenance personnel will manually adjust the parameters based on the parameter optimization results during the production task switching period (such as shift handover and order completion). After the adjustment, the parameter change information will be fed back to the system via a barcode scanner or data entry terminal to update the production process data. S402 Abnormal Warning and Intervention Equipment abnormal warning: When the equipment health index is ≤60 or the sensor monitoring data exceeds the threshold (such as vibration acceleration >0.5g, temperature >60°C), the system automatically triggers a three-level warning (warning, fault, serious fault) and pushes it to the operation and maintenance personnel through the workshop LED screen and mobile app. The warning information includes the equipment number, fault type (such as abnormal vibration, excessive temperature), and recommended maintenance measures (such as checking bearings and cleaning heat dissipation channels). Operation and maintenance personnel must respond within 10 minutes and arrive at the site within 30 minutes to handle the problem. Production abnormality intervention: If there is a material shortage (inventory ≤ 10% of safety stock), a production capacity gap > 10%, or a product qualification rate < 95%, the system initiates an abnormality intervention process, including: automatically triggering an emergency material purchase request (linking to the supplier management system to push the purchase order to the preferred supplier), adjusting production task allocation (such as transferring some tasks to idle equipment or outsourcing), and re-optimizing process parameters (shortening the genetic algorithm optimization cycle to 30 generations and quickly iterating parameters) to ensure continuous and stable production processes.

9. The data-driven intelligent control method for workshop production equipment processes according to claim 1, characterized in that: The S5 control effect feedback and iterative stage effect evaluation index production efficiency: Calculate the output per unit time (e.g. pieces / hour) and the ratio of effective equipment operating time (effective operating time / planned operating time × 100%), with target values ​​of increasing by 10% and ≥90% respectively; Product quality: Calculate the product qualification rate (qualified product quantity / total production quantity × 100%) and the distribution of defective product types (such as dimensional deviation and surface defect ratio). The target value is a qualified rate of ≥99% and a defective product ratio of ≤1%. Energy consumption cost: Calculate unit product energy consumption (kWh / unit) and equipment energy consumption reduction rate ((original energy consumption - current energy consumption) / original energy consumption × 100%). The target value is to reduce unit energy consumption by 8% and the equipment energy consumption reduction rate by ≥5%.

10. The data-driven intelligent control method for workshop production equipment process according to claim 1, characterized in that: The S501 model is iteratively optimized: Every seven days, data on control effectiveness (production efficiency, product quality, and energy costs) is collected and compared with the model's predicted values, and the error rate is calculated (|actual value - predicted value| / predicted value × 100%). If the error rate exceeds 15%, the equipment health analysis model and production process optimization model are iteratively updated: new failure case data and production anomaly data are added to the training set, the model is retrained, and parameters such as the number of decision trees in the random forest algorithm and the number of hidden layer neurons in the LSTM neural network are adjusted. The fitness function of the genetic algorithm is optimized, incorporating new evaluation metrics (such as equipment coordination efficiency and material delivery timeliness) to improve the model's adaptability to complex production scenarios and ensure that the control method continues to meet the actual production needs of the workshop.