Predictive maintenance method and system for smart manufacturing industry production line

Abstract

The application discloses a predictive maintenance method and system for a production line in an intelligent manufacturing industry, and belongs to the technical field of data processing. The method comprises the following steps: after the production line in the manufacturing industry continuously operates in a working state and reaches a set operation cycle or abnormal fluctuation occurs, basic working condition characteristic parameters of the production line in the manufacturing industry in a preset time period are called, the operation state of a cooling device of the production line in the manufacturing industry is analyzed based on the basic working condition characteristic parameters through a machine learning algorithm model, a health score is generated, and a corresponding operation result label is given; when the operation result label is accurate, the current operation state is maintained; otherwise, based on the operation state of the cooling device, an operation result verification label is determined, and a maintenance adjustment scheme is obtained synchronously; consistency between the operation result verification label and the operation result label is judged, and the technical problem that the predictive maintenance is not accurate due to the prediction deviation of the operation state of the equipment in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a predictive maintenance method and system for production lines in the smart manufacturing industry. Background Technology

[0002] Intelligent manufacturing refers to efficient, flexible, and personalized production processes utilizing information technology, automation technology, and artificial intelligence. Intelligent manufacturing production lines are multi-process collaborative production systems composed of various equipment such as CNC machine tools, automatic feeding mechanisms, machining execution components, vision inspection units, automatic handling devices, and cooling systems. Their core objective is to achieve continuous, stable, and flexible processing operations using automation, informatization, and intelligent technologies. In such production lines, cooling systems typically undertake the tasks of heat dissipation, maintaining stable temperatures in the processing area, and preventing thermal fatigue and thermal drift of critical components, making them a key link in ensuring processing accuracy and equipment lifespan. Because production lines usually operate continuously for long periods, cooling systems are prone to performance degradation, decreased flow rate, pressure fluctuations, and reduced heat exchange efficiency under high load and long-term operating conditions. Failure to identify abnormal trends in cooling systems in a timely manner can lead to deviations in production accuracy, thermal damage to equipment, or even production line shutdowns. Therefore, accurate monitoring and predictive maintenance of the cooling system's operating status are necessary. Predictive maintenance is based on real-time operating status and historical / degradation trend data of equipment / production lines. By analyzing and judging potential future failures or remaining useful life (RUL), maintenance can be scheduled in advance to reduce unplanned downtime and extend equipment life. The cooling device in this context does not refer to a specific model, but rather to a cooling system unit on a smart manufacturing production line that provides circulating heat dissipation for processing equipment, power actuators, or process units. It is installed near specific equipment or forms a "station-level / equipment-level cooling loop" integrated with the equipment. Examples include the spindle cooling loop of a CNC machine tool, the oil cooling loop of a hydraulic station, the cutting fluid circulating cooling system, and the mold cooling system (such as for die casting and injection molding).

[0003] Predictive maintenance of existing intelligent manufacturing production lines typically relies on multi-source data acquisition and machine learning models. By collecting equipment operating status parameters, preliminary prediction results are generated using random forest models and extreme gradient boosting models, respectively. These results are then corrected using linear regression models to obtain more stable fault prediction outputs. Alternatively, spatial features of operating data can be obtained by constructing an input feature extractor, which is then fused with real-time data to form hidden states. A dynamic feature analysis module infers the trend of equipment status changes, and finally, a feature integration and prediction output module generates predicted results for equipment faults.

[0004] For example, Chinese invention patent CN118070246B discloses a predictive maintenance method for IoT devices in smart factories, which includes: using a random forest model and an extreme gradient boosting model for initial prediction, and then using a linear regression model to further process the initial prediction results to obtain the final prediction results.

[0005] For example, Chinese invention patent CN118761033B discloses a fault prediction model, method, device, and medium for a CNC machine tool cooling device, including: an input feature extractor for extracting spatial features of the operating data of the CNC machine tool cooling device; a data fusion unit for fusing operating data and spatial features to obtain the current hidden state; a dynamic feature analyzer for estimating the evolution trend of the operating data based on the hidden state; a comprehensive feature integrator adaptively fusing the results of different features output by the dynamic feature analyzer; and a final prediction output generator converting the fused hidden state into a specific prediction output.

[0006] In summary, existing technologies generally rely on fixed thresholds or empirical rules for predictive maintenance of production lines. However, because cooling devices are susceptible to the combined effects of processing intensity, ambient temperature, and equipment degradation, traditional prediction methods struggle to accurately reflect subtle fluctuations in their operating status, leading to unstable prediction results and inaccurate maintenance triggering times. Consequently, existing technologies are prone to prediction bias when facing complex production conditions and high-frequency dynamic fluctuations, failing to provide timely and effective predictive maintenance support for production lines. Summary of the Invention

[0007] To address the technical problem of inaccurate predictive maintenance due to deviations in equipment operating status prediction in existing technologies, this invention provides a predictive maintenance method and system for production lines in the intelligent manufacturing industry. The technical solution is as follows:

[0008] On the one hand, a predictive maintenance method for intelligent manufacturing production lines is provided. This method includes: after the production line has been continuously operating and has reached a set operating cycle or experienced abnormal fluctuations, retrieving the basic operating condition characteristic parameters of the production line within a preset time period, and analyzing the operating status of the cooling device of the production line using a machine learning algorithm model based on the basic operating condition characteristic parameters, thereby generating a health score and providing a corresponding operating result label; when the operating result label is an accurate prediction result, maintaining the current operating status; otherwise, determining the operating result verification label based on the operating status performance of the cooling device, and simultaneously obtaining a maintenance adjustment plan; and judging the consistency between the operating result verification label and the operating result label. If they are consistent, implementing corresponding predictive maintenance based on the maintenance adjustment plan; otherwise, updating the machine learning algorithm model to obtain the operating result label.

[0009] On the other hand, a predictive maintenance system for intelligent manufacturing production lines is provided. This system includes: a production status analysis module, a defensive adjustment confirmation module, and an operation correction module. The production status analysis module retrieves basic operating condition characteristic parameters of the production line within a preset time period after the production line has been continuously operating for a set period or after abnormal fluctuations occur. Based on these parameters, a machine learning algorithm model analyzes the operating status of the cooling device, generating a health score and providing a corresponding operation result label. The defensive adjustment confirmation module maintains the current operating status when the operation result label matches the predicted result; otherwise, it determines the operation result verification label based on the cooling device's operating status and simultaneously obtains a maintenance adjustment plan. The operation correction module checks the consistency between the operation result verification label and the operation result label. If they match, corresponding predictive maintenance is implemented based on the maintenance adjustment plan; otherwise, the machine learning algorithm model is updated to obtain the operation result label.

[0010] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0011] 1. The predictive maintenance method for intelligent manufacturing production lines provided by this invention retrieves the basic operating condition characteristic parameters of the production line within a preset time period and combines them with historical data to input into a machine learning algorithm model to predict health scores, thereby obtaining the predicted health scores for the current and future times. This achieves an accurate representation of the overall operating status of the production line and effectively solves the technical problem of inaccurate predictive maintenance caused by the prediction deviation of equipment operating status in the prior art.

[0012] 2. This invention performs difference analysis between the actual health score and the predicted value at the target time and generates operation result labels, thereby verifying and correcting the machine learning prediction results, ensuring the reliability and accuracy of the predictive maintenance scheme, and effectively avoiding the maintenance misjudgment problem caused by the inconsistency between the actual equipment status and the prediction results.

[0013] 3. This invention obtains high-frequency data parameters of the cooling device within an abnormal time window, analyzes its dynamic behavior status accordingly, generates operation result verification tags, and simultaneously obtains maintenance and adjustment plans. This enables real-time management and defensive maintenance of the cooling device under multi-variable coupled disturbances, effectively solving the potential equipment risk problem caused by the inability to adjust the cooling system in a timely manner due to abnormalities in the prior art.

[0014] 4. This invention obtains abnormal execution values ​​by processing abnormal execution parameters and output execution sample sets, thereby quantifying the degree of abnormal cumulative deviation of the execution parameters of the cooling device, and thus realizing targeted adjustment of abnormal parameters and optimization of the cooling processing execution cycle. This effectively improves the response speed and adjustment accuracy of the cooling system and solves the problem of low maintenance efficiency caused by the inability to accurately quantify multivariate interference in the prior art. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A flowchart illustrating the steps of a predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment;

[0017] Figure 2 A global flowchart of the predictive maintenance method for intelligent manufacturing production lines provided in the embodiments of this application;

[0018] Figure 3 A flowchart illustrating the dynamic behavior state analysis of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment;

[0019] Figure 4 A flowchart illustrating the determination of the cooling process execution cycle in the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment;

[0020] Figure 5 A flowchart illustrating the execution characteristic state analysis of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment;

[0021] Figure 6 A schematic diagram of the modules of a predictive maintenance system for a smart manufacturing production line provided in an embodiment of this application;

[0022] Figure 7 A maintenance and inspection process diagram of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment;

[0023] Figure 8 This is a schematic diagram of the health score prediction process of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment. Detailed Implementation

[0024] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0025] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0026] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0027] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0028] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0029] Embodiment 1 of this invention provides a predictive maintenance method for production lines in the intelligent manufacturing industry, such as... Figure 1 The diagram shows a flowchart of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment. The method includes the following steps: After the manufacturing production line has been continuously operating and has reached a set operating cycle or experienced abnormal fluctuations, the basic operating condition characteristic parameters of the manufacturing production line within a preset time period are retrieved. Based on the basic operating condition characteristic parameters, the operating status of the cooling device of the manufacturing production line is analyzed using a machine learning algorithm model, thereby generating a health score and providing a corresponding operating result label. When the operating result label indicates that the prediction result is accurate, the current operating status is maintained; otherwise, based on the operating status performance of the cooling device, an operating result verification label is determined, and a maintenance adjustment plan is obtained simultaneously. The consistency between the operating result verification label and the operating result label is judged. If they are consistent, corresponding predictive maintenance is implemented based on the maintenance adjustment plan; otherwise, the machine learning algorithm model is updated to obtain the operating result label.

[0030] In this embodiment, as Figure 2 As shown, Figure 2The global flowchart of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment firstly involves the system retrieving basic operating condition characteristic parameters within a preset time period when the production line reaches a set operating cycle or experiences operational fluctuations, and preprocessing the relevant data. Then, historical characteristic parameters are input into the machine learning algorithm model, outputting predicted health scores for the current and future time series. After selecting a target time, the system performs difference analysis between the predicted health score and the actual health score, thereby generating an operational result label. If the label indicates accurate prediction, the existing operating state is maintained; if the label indicates deviation from the prediction, the operating status of the cooling device is further analyzed, an operational result verification label is generated, and a maintenance adjustment plan is provided. Finally, the system executes the corresponding predictive maintenance based on the adjustment plan and updates the machine learning algorithm model after execution, forming a closed-loop, optimizable maintenance logic for the entire method.

[0031] Furthermore, the health score is obtained through the following steps: retrieving the basic operating condition characteristic parameters of the manufacturing industry production line within a preset time period. These basic operating condition characteristic parameters are used to characterize the basic operating conditions of the overall operation status of the intelligent manufacturing industry production line. After preprocessing the basic operating condition characteristic parameters, they are combined with historical basic operating condition characteristic parameters and input into a preset machine learning algorithm model to output the current health score and the predicted health score values ​​for each time period within the future preset time.

[0032] In this embodiment, the basic operating condition characteristic parameters are the overall operating condition characteristic parameters of the manufacturing production line, such as load fluctuation amplitude, average temperature rise rate of equipment group, cycle time offset, and global flow rate change rate of coolant. The executing equipment corresponding to these basic operating condition characteristic parameters is the production and processing equipment. It should be noted that this embodiment does not limit the specific type and selection of production lines; adjustments can be made according to the actual situation in practical applications. Only a few common production lines are introduced below. In specific projects, the production line that actually needs to be analyzed can be selected for evaluation. In addition to the two production lines introduced in this embodiment, other production lines with cooling systems can also be selected for evaluation.

[0033] Example A: This case is an intelligent manufacturing production line for machining aluminum alloy shells using a group of high-speed vertical machining centers. In this production line, the basic operating condition characteristic parameters include: load fluctuation amplitude, which is the spindle servo load fluctuation amplitude of the high-speed vertical machining center spindle system; average temperature rise rate of the equipment group, which is the average of the spindle motor temperature rise rate and the servo motor temperature rise rate; cycle time offset, which is the workpiece transfer waiting time (if no waiting is required, it is 0 at this time, and the waiting time is the cycle time offset, because the set target waiting time is zero); and coolant global flow rate change rate.

[0034] Example B: This case study describes an intelligent production line for precision machining of automotive engine cylinder blocks (cast iron / cast aluminum), including roughing and finishing of cylinder bores, oil passage milling, and mounting surface milling. To ensure tool life, reduce thermal deformation, and improve machining stability, the production line is equipped with a high-pressure coolant supply station and provides high-pressure, high-flow-rate coolant to multiple horizontal machining centers through a centralized cooling system. The basic operating parameters on this production line include: load fluctuation amplitude (i.e., the spindle torque fluctuation of the horizontal machining centers), average temperature rise rate of the equipment group (the rate of temperature rise of the spindle motors of the machining workstation composed of multiple horizontal machining centers, the rate of temperature rise of the three-axis servo motors, and the rate of temperature rise within the machine tool electrical control cabinet), cycle time offset (i.e., the pallet exchange time of the workpiece conveyor chain), and the global coolant flow rate change rate (i.e., the coolant return speed of the coolant pump group).

[0035] The load fluctuation amplitude refers to the overall load fluctuation amplitude of the production line. This can be achieved by retrieving real-time load data (such as spindle torque, servo motor percentage load, conveyor drive load, etc.) from the production line's main control system. Within a preset statistical period, the load time series of all equipment is normalized, the maximum deviation of each equipment's load curve from its mean is calculated, and the average value of all equipment is then calculated to obtain the overall load fluctuation amplitude of the production line. The average temperature rise rate of the equipment group reflects the degree of heat accumulation of the production line's equipment cluster under continuous operation. Real-time temperature data of key equipment (such as spindles, servo drives, hydraulic stations, cooling pumps, etc.) is collected from temperature sensors. The temperature curves of each equipment are processed using first-order difference to obtain the temperature change per unit time. The average temperature rise rate of the equipment group is then calculated. Cycle time offset is used to reflect the volatility of production line process cycles. The actual cycle time of each process (such as clamping, processing, inspection, and transfer steps) is retrieved from the production line's back-end management system. Using the standard cycle time of the production line as the benchmark cycle time, the offset between the actual cycle time and the benchmark cycle time is calculated for each process, and then the average value is taken to obtain the cycle time offset. The global flow rate change rate of coolant is used to characterize the stability of the global fluid circulation of the cooling device. Real-time flow rate data of the supply and return branches of the cooling device are collected by flow meters. Multiple flow rate sampling points are synchronously normalized to construct a global flow rate feature vector. This vector is then subjected to trend fitting (e.g., linear fitting) over a preset time period to obtain the change per unit time. The average value of the changes at all flow rate monitoring points is then taken to obtain the global flow rate change rate of coolant.

[0036] When continuous production load fluctuations intensify, uneven energy output occurs during processing, transmission, and drive operations, leading to increased heat accumulation rates in internal and surrounding components. This causes the average temperature rise rate of the equipment group to trend upwards. When the temperature rise rate remains high, the equipment will lengthen or compress local operation cycles to maintain performance stability, causing a drift in the process execution rhythm and ultimately further amplifying the cycle time offset. This continuous offset in the cycle time will cause varying degrees of misalignment in the start-stop logic and coolant supply rhythm of the cooling system, altering the flow load of the coolant in different branches. This causes irregular fluctuations in the global flow rate of the coolant. Conversely, when the global flow rate of the coolant continues to increase or tends to decrease, the temperature of key heat dissipation components is difficult to maintain within a reasonable range, leading to a rapid increase in the average temperature rise rate of the equipment group. This, in turn, further exacerbates the load fluctuation, forming a cyclical coupling relationship between thermal load, mechanical load, cycle time deviation, and cooling efficiency. In this coupling mechanism, any deviation of any parameter in an unfavorable direction will cause other parameters to change in a coordinated manner, making the overall operating state of the production line in the intelligent manufacturing industry exhibit quantifiable dynamic coordinated instability characteristics.

[0037] The machine learning algorithm model is a long short-term memory neural network model based on feature sequence input. It is implemented by constructing an input layer, a time-step recurrent layer, and an output layer. First, the basic operating condition feature parameters are normalized and then arranged in chronological order to form a feature sequence. Then, the feature sequence is input into the long short-term memory recurrent layer to capture the cross-time correlation formed by equipment load, temperature rise changes, cycle time offset, and cooling flow rate fluctuations. The output of the recurrent layer is transformed by a fully connected layer to obtain the corresponding health score and future prediction value. The input data of this machine learning algorithm model includes historical basic operating condition feature parameters and corresponding historical health scores.

[0038] By retrieving basic operating condition characteristic parameters within a preset time period and using them to characterize the overall operating status of production lines in the manufacturing industry, the generation of health scores has higher operating condition sensitivity and time resolution. At the same time, current features and historical features are input into the machine learning algorithm model, so that the scoring results can take into account both the immediate state and long-term trends, thereby achieving a forward-looking prediction of future operational health and improving the reliability of predictive maintenance decisions.

[0039] Furthermore, the operation result labeling method is as follows: Select the predicted health score at a certain moment within a preset time period and mark it as the first predicted health score, and mark this moment as the target moment; obtain the actual health score at the target moment (this actual health score is the health score obtained by analysis when running to the target moment), and perform absolute difference processing with the first predicted health score to obtain the absolute difference of the health score; obtain a preset health score difference threshold and compare it with the absolute difference of the health score. If the absolute difference of the health score is below the health score difference threshold, the operation result label is "Prediction result is accurate"; otherwise, the operation result label is "Prediction result is deviated".

[0040] In this embodiment, by analyzing the difference between the predicted and actual health scores at the target time and generating an operation result label, the predictive maintenance process can verify the reliability of the machine learning algorithm model's predictions in real time. Since load disturbances, cooling efficiency fluctuations, and cycle time changes in industrial scenarios are random, relying solely on model output may lead to deviations. Introducing the operation result label allows for timely identification of model misjudgments when the predicted results deviate from the actual state, thereby avoiding equipment anomalies or production line fluctuations caused by implementing maintenance strategies based on erroneous scores. When the absolute difference in health scores is within the difference threshold, the system can confirm that the machine learning algorithm model still possesses stable predictive capabilities under the current operating conditions and can maintain the existing operating strategy. If the difference exceeds the limit, a correction mechanism can be immediately triggered, making subsequent maintenance execution logic closer to the actual operating conditions. This achieves dynamic matching between the predicted results and the actual state, improving the reliability, adaptability, and stability of the predictive maintenance process, ensuring that the production line can still conduct accurate health assessments and make scientific maintenance decisions in complex environments. It should be noted that an operation result label indicating an accurate prediction means the machine learning algorithm model's prediction is normal, while an operation result label indicating a deviation means the machine learning algorithm model's prediction is abnormal.

[0041] Furthermore, the operating status of the cooling device includes: dynamic behavior status or execution characteristic status.

[0042] Furthermore, based on the dynamic behavior status, a verification tag for the operation result is determined, and a maintenance adjustment plan is obtained simultaneously. Based on the maintenance adjustment plan, a defensive maintenance signal is generated after the cooling process execution cycle, thereby triggering the execution of the corresponding maintenance adjustment plan. Specifically, the high-frequency data parameters of the cooling device at each moment within the abnormal time window are obtained. These high-frequency data parameters include the transient flow rate of the coolant, the frequency of pressure pulsation, the reciprocating motion deviation of the actuator, the bearing vibration amplitude, the rate of temperature rise, and the intensity of current fluctuation. Within the abnormal time window, if the high-frequency data parameters at each moment are all within the preset set range of high-frequency data parameters, the operation result verification tag indicates normal operation, and the maintenance adjustment plan is not adjusted. If the transient flow rate of the coolant increases, and the pressure pulsation frequency... If periodic fluctuations occur (oscillating around a certain value), the operation result verification label indicates abnormal operation. In this case, the maintenance and adjustment plan is to reduce the output pressure of the regulating coolant pump. If the reciprocating motion deviation of the actuator increases and the bearing vibration amplitude deviates, the operation result verification label indicates abnormal operation. In this case, the maintenance and adjustment plan is to reduce the speed of the regulating pump. If the rate of temperature rise and the intensity of current fluctuation increase simultaneously, the operation result verification label indicates abnormal operation. In this case, the maintenance and adjustment plan is to increase the flow rate of the regulating pump. Otherwise, the operation result verification label indicates severe abnormal operation. In this case, the maintenance and adjustment plan is to issue a warning message and stop the production line operation in the manufacturing industry. The above abnormal situations may occur simultaneously. If the above abnormal situations occur simultaneously, the corresponding maintenance and adjustment plans will be executed simultaneously.

[0043] In this embodiment, as Figure 3 As shown, Figure 3 This application provides a flowchart of the dynamic behavior state analysis process for a predictive maintenance method for intelligent manufacturing production lines. The system first collects operating data of the cooling device within an abnormal time window, acquiring high-frequency operating parameters including transient coolant flow rate, pressure pulsation frequency, actuator reciprocating motion deviation, bearing vibration amplitude, temperature surge rate, and current fluctuation intensity. It then determines whether the operating result verification tag indicates an abnormal operation. If it is determined to be normal, the maintenance adjustment plan is to not adjust; if it is determined to be abnormal, the cause of the abnormality is further identified and a corresponding maintenance adjustment plan is matched: if the transient coolant flow rate increases and is accompanied by periodic fluctuations in pressure pulsation frequency, the coolant pump output pressure is reduced; if the actuator reciprocating motion deviation increases and the bearing vibration amplitude shifts, the pump speed is reduced; if the temperature surge rate and current fluctuation intensity increase synchronously, the pump flow rate is increased. The process terminates after each branch completes its maintenance adjustment. It should be noted that if the high-frequency data parameters of the cooling device are not within the preset range of high-frequency data parameters, it indicates that the cooling device is abnormal. If any one or more of the above three abnormal operating conditions are met, the maintenance and adjustment plan for the cooling device is the maintenance and adjustment plan mentioned above.

[0044] An abnormal time window refers to the time interval between the time corresponding to the first predicted health score and the target time when making a health score prediction. The preset time period is the time period that starts from the first time of the abnormal time window and spans a preset duration. In other words, the preset time period is a duration within the abnormal time window starting from the first time.

[0045] The specific methods for reducing the output pressure of the regulating pump, reducing the speed of the regulating pump, and increasing the flow rate of the regulating pump are as follows: Matching the comprehensive interference complexity value with a preset sample set, where the sample set includes various comprehensive interference complexity sample values ​​(historical comprehensive interference complexity values ​​obtained in real-time during historical operation; specifically, these historical comprehensive interference complexity values ​​refer to the comprehensive interference complexity sample values ​​when the operating state of the equipment was directly improved from abnormal to normal through primary equipment adjustment during the fault occurrence). The comprehensive interference complexity value is compared with each comprehensive interference complexity sample value, and the comprehensive interference complexity sample value closest to the current comprehensive interference complexity value is taken as the reference comprehensive interference complexity value. The adjustment parameter sample value corresponding to this reference comprehensive interference complexity value is obtained as the actual parameter adjustment value, where the actual parameter adjustment value includes the cooling pump output pressure adjustment value, the cooling pump speed adjustment value, and the cooling pump flow rate adjustment value (coolant flow rate output by the cooling pump). Based on this, the cooling pump output pressure is reduced and / or the regulating pump speed is reduced and / or the regulating pump flow rate is increased in response.

[0046] Transient coolant velocity refers to the instantaneous change in coolant flow velocity per unit time in the pipeline. It characterizes the dynamic flow state of the coolant delivery chain and the transient supply capacity of the pump. It is obtained through a high-speed turbine flow meter installed on the main supply pipeline. Pressure pulsation frequency refers to the periodic change frequency of pressure fluctuations in the cooling pipeline. It characterizes the pump impeller operation stability, flow channel blockage tendency, and fluid impact intensity in the pipeline. It can be obtained by acquiring the fluid pressure change curve over time using a high-frequency pressure sensor and calculating the number of pressure pulsations per unit time. Actuator reciprocating motion deviation refers to the displacement offset of the actuator (valve body drive rod) during reciprocating motion in the cooling device. It characterizes the consistency of mechanical motion and positioning accuracy of the actuator. It can be obtained by monitoring the real-time displacement trajectory of the actuator using a magnetic encoder and calculating the difference between the set motion trajectory and the actual motion trajectory. Bearing vibration amplitude refers to the vibration intensity of the cooling pump or actuator bearing. It characterizes the balance, lubrication status, and wear degree of rotating components. It can be obtained by analyzing the vibration amplitude at a high sampling frequency using an acceleration vibration sensor installed on the pump body. The rate of temperature rise refers to the rapid increase in temperature of a key component of a cooling system per unit time (rapid rise means that the rate of temperature change per unit time exceeds a preset temperature change rate threshold). It is used to characterize cooling efficiency degradation, heat exchange imbalance, or abnormal local heating. The transient temperature rise rate can be obtained by acquiring temperature curves using thermocouples and calculating the transient rate of temperature rise using differential methods. The current fluctuation intensity refers to the degree of rapid fluctuation in the current driving a pump, motor, or actuator within a short period of time. It characterizes changes in motor load, increased internal friction, or abnormal pump end resistance. It can be obtained by acquiring current curves using current transformers and calculating the maximum fluctuation amplitude; this fluctuation amplitude is the current fluctuation intensity.

[0047] Increased flow velocity indicates accelerated coolant flow, while periodic fluctuations in pressure pulsation frequency usually signify flow instability in the pump impeller or piping, such as localized cavitation, backflow, or fluid shock. Maintaining the original pressure may exacerbate pressure fluctuations, leading to increased mechanical shock at the pump. Reducing the coolant pump output pressure can alleviate fluid shock, continuously reducing the amplitude of pressure fluctuations within the piping, and allowing the system to return to a steady-state flow range. Increased actuator motion deviation and bearing vibration amplitude shift often occur simultaneously under conditions of increased mechanical load or abnormally high friction coefficient, indicating a dynamic imbalance in the system's mechanical transmission link. Reducing the pump speed can effectively alleviate the motion load and vibration coupling intensity of the actuator, preventing further wear accumulation due to vibration, and allowing the actuator to re-enter a normal reciprocating operating trajectory. A sudden temperature rise indicates a significant decrease in cooling efficiency, while increased current fluctuations mean the pump is operating under higher resistance, potentially indicating overheating due to substandard circulation efficiency.

[0048] Furthermore, the cooling process execution cycle involves the following steps: First, obtain the high-frequency data parameters at the first moment within the abnormal time window and use them as reference data. Second, compare the high-frequency data parameters at each moment within the abnormal time window with the reference data to obtain the trend of change in the high-frequency data parameters. (For example, if a parameter within the high-frequency data parameters, such as the transient flow rate of the coolant, is greater at the first moment than at the last moment, and the transient flow rate gradually decreases in the intermediate moments (the overall trend is decreasing; if there is a trend at a certain moment that differs from the overall trend, it is ignored), then the trend of change in the transient flow rate of the coolant is decreasing.) This trend includes increases, decreases, or periodic fluctuations. If there is a trend in the change of a certain high-frequency data parameter... If the trend of change is the same as the preset trend in the maintenance and adjustment plan (e.g., the transient flow rate of coolant increases, and the frequency of pressure pulsation fluctuates periodically, the reciprocating motion deviation of the actuator increases, and the bearing vibration amplitude shifts, and the temperature rise rate and current fluctuation intensity increase simultaneously), then the parameter is marked as an abnormal management parameter. (It should be noted that this means that the parameter at each moment within the abnormal time window is an abnormal management parameter. Therefore, the value of the abnormal management parameter at this time is specifically the abnormal management parameter at the last moment within the abnormal time window, i.e., the abnormal management parameter at the target moment. The last moment is the same as the target moment.) Thus, each abnormal management parameter is obtained. Based on the high-frequency data parameters at the last moment, they are compared with the control data. The process involves several steps. First, if the high-frequency data parameter at the last moment (target moment) within the abnormal time window belongs to an anomaly management parameter, then an anomaly handling value is obtained. This is achieved by dividing the anomaly management parameter at the last moment by the corresponding reference data. For example, if the coolant transient flow rate is an anomaly management parameter, then the coolant transient flow rate is divided by the corresponding reference data value. Otherwise, a preset handling value is obtained. This preset handling value is obtained by acquiring a preset high-frequency data parameter data analysis mapping table. This table stores various high-frequency data parameters, and each high-frequency data parameter corresponds to a fixed preset handling value. When a high-frequency data parameter does not belong to an anomaly management parameter, the preset handling value can be determined based on the high-frequency data parameter. The corresponding preset processing value is obtained by querying the data parameter data analysis mapping table. Then, the abnormal processing value and the preset processing value are introduced into the corresponding weights (these weights are pre-set weights) and multiplied and superimposed to obtain the comprehensive interference complexity value. The comprehensive interference complexity value is used to characterize the degree of composite influence of multivariate coupled disturbances of the cooling device. Based on the comparison between the comprehensive interference complexity value and the preset comprehensive interference complexity threshold, when the comprehensive interference complexity value is less than the comprehensive interference complexity threshold, the preset cooling cycle is obtained as the cooling processing execution cycle. Otherwise, the first delay shortening time is obtained based on the comprehensive interference complexity value, and the cooling execution time of the cooling processing execution cycle is obtained by subtracting the first delay shortening time from the cooling time corresponding to the cooling cycle.

[0049] In this embodiment, as Figure 4 As shown, Figure 4 The flowchart for determining the cooling process execution cycle of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment first obtains high-frequency data at the first moment of the abnormal time window as reference data, and compares it item by item with the high-frequency data at each moment within the abnormal time window. When the change trend of a certain high-frequency parameter is consistent with the preset trend in the maintenance adjustment plan, the parameter is marked as an abnormal management parameter and an abnormal processing value is calculated; otherwise, the preset processing value is calculated. Subsequently, the abnormal processing value and the preset processing value are superimposed according to their corresponding weights to obtain a comprehensive disturbance complexity value, which is used to reflect the comprehensive degree of multivariate coupled disturbances experienced by the cooling device. If the value is lower than the comprehensive disturbance complexity threshold, the preset cooling cycle is adopted; if it exceeds the threshold, the corresponding delay shortening time is matched to shorten the cooling execution time, so that the cooling process responds more promptly to the level of complex disturbances.

[0050] It should be noted that preset processing values ​​are used for high-frequency data parameters that are not part of the anomaly management parameters. This is to establish a stable reference benchmark and avoid interference from normal fluctuations in anomaly detection. The preset processing values ​​are derived from the high-frequency data parameter data analysis mapping table. This ensures that non-anomaly parameters participate in the calculation of the comprehensive disturbance complexity value with uniform, fixed values, maintaining a balanced parameter weight distribution. This allows the comprehensive disturbance complexity value to accurately characterize the degree of multivariate coupled disturbances in the cooling device. By distinguishing between anomaly processing values ​​and preset processing values, the system can clearly identify the contribution of anomaly factors, preventing the cooling processing execution cycle from being affected by non-anomaly factors, and ensuring the controllability and reliability of cycle adjustment decisions.

[0051] The first delay reduction duration is obtained based on the matching of comprehensive interference complexity values. The specific method is as follows: a preset sample set is obtained, which includes several sample comprehensive interference complexity values ​​and the sample first delay reduction duration corresponding to the sample comprehensive interference complexity values. The comprehensive interference complexity value is compared with the sample comprehensive interference complexity value respectively. The sample comprehensive interference complexity value that is closest to the comprehensive interference complexity value is taken as the reference sample comprehensive interference complexity value. The sample first delay reduction duration corresponding to the reference sample comprehensive interference complexity value is obtained as the first delay reduction duration.

[0052] By dynamically analyzing the cooling process execution cycle, real-time quantitative assessment of multivariate disturbances in the cooling device within an abnormal time window is achieved. High-frequency data parameters at the first moment of the abnormal time window are used as reference data, and trends are compared with high-frequency data parameters at subsequent moments. This allows for accurate identification and labeling of abnormal features as anomaly management parameters, avoiding misjudgments caused by single parameter errors. Automated and differentiated adjustment of the cooling process execution cycle is achieved by comparing the comprehensive disturbance complexity value with a comprehensive disturbance complexity threshold. When the comprehensive disturbance complexity value is below the threshold, it indicates a low level of disturbance, and the cooling device remains within a controllable range; therefore, a preset cooling cycle can be directly used to maintain operational stability. When the comprehensive disturbance complexity value exceeds the threshold, it indicates a significantly increased level of disturbance to the cooling device, posing risks such as a sharp drop in cooling efficiency, fluid transport instability, or increased dynamic load on the actuator. In this case, appropriately shortening the cooling process execution cycle allows for earlier intervention of maintenance measures, thereby suppressing further amplification of the anomaly and preventing the system from entering an irreversible fault propagation stage. Therefore, this invention achieves highly sensitive anomaly response capabilities by adaptively adjusting the cooling process execution cycle based on the comprehensive disturbance complexity value. The higher the overall interference complexity value, the more severe the multivariate coupling anomaly. Therefore, the shorter the cooling process execution cycle, the more timely the cooling device can be intervened at the most critical time, thereby improving the safety margin of the cooling system and the reliability of predictive maintenance, and achieving precise protection of the operational stability of the cooling device.

[0053] Furthermore, based on the execution characteristic status, a verification label for the running result is determined, and a maintenance and adjustment plan is obtained simultaneously. Based on the maintenance and adjustment plan, defensive maintenance is performed after the cooling process execution cycle. The specific method is as follows: the actual execution parameters of the cooling device are obtained, which are used to characterize the comprehensive execution performance of the cooling device under heat dissipation conditions; the corresponding output execution sample set in the database is obtained and compared with the actual execution parameters one by one. If a certain actual execution parameter exceeds the preset offset range of the corresponding output execution sample value, the parameter is marked as an abnormal execution parameter, thus obtaining each abnormal execution parameter; based on the analysis of each abnormal execution parameter, the abnormal execution value is obtained; the preset abnormal execution threshold in the database is obtained and compared with the abnormal execution value. If the abnormal execution value is less than the abnormal execution threshold, the preset cooling time is obtained as the cooling process execution cycle; otherwise, the second delay shortening time is obtained based on the abnormal execution value, and the cooling process execution cycle is obtained by subtracting the second delay shortening time from the preset cooling time; thus, the frequency of the variable frequency drive of the cooling pump motor is increased and adjusted within the cooling process execution cycle.

[0054] In the second embodiment of this solution, as Figure 5 As shown, Figure 5This document presents a flowchart illustrating the execution characteristic state analysis process of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment. The flowchart explains the process of calculating abnormal execution values ​​and adjusting the cooling process execution cycle based on execution characteristic states. The system first acquires the actual execution parameters of the cooling system and compares them one by one with the output execution sample set stored in the database, marking parameters exceeding the offset range as abnormal execution parameters. Then, it calculates the abnormal execution value based on the abnormal execution parameters and compares it with an abnormal execution threshold. If the abnormal execution value is greater than the threshold, it indicates a high degree of execution offset. In this case, the duration is shortened by matching the delay, thus shortening the cooling process execution cycle. If the abnormal execution value does not exceed the threshold, a preset cooling cycle is used as the final execution cycle. Finally, the frequency of the cooling pump is adjusted based on the final execution cycle to make the cooling process more suitable for the current execution state.

[0055] The output execution sample set is a set of parameters collected under normal steady-state operating conditions on the production line. The actual execution parameters include the rate of temperature rise, the rate of flow rate decay, the instantaneous flow rate, and the pipeline pressure difference.

[0056] The rate of temperature rise is used to characterize the rate of temperature change of the cooled object (production equipment in manufacturing production lines) within an abnormal time window. It can be obtained by subtracting the temperature value at the beginning of the abnormal time window from the temperature value at the end of the abnormal time window, and then dividing the temperature difference by the length of the abnormal time window. This reflects whether the heat absorption capacity of the cooling device is sluggish or the heat transfer efficiency is reduced. The flow rate decay rate is used to characterize the delivery decay characteristics of the coolant in the circulation path, reflecting the decline in pump output, increase in pipeline resistance, or local blockage. It is measured by the rate of change in the output flow rate of the cooling device at the beginning and end of the time window. The instantaneous flow rate value is used to characterize the actual delivery volume of the coolant at the current moment and is a core indicator for measuring whether the circulation supply meets the expected load requirements. The pipeline pressure difference is used to characterize the change in the flow resistance of the coolant between the inlet and outlet channels, reflecting changes in fluid dynamic characteristics such as pipeline pressure loss, flow channel blockage, internal wear, or air resistance formation.

[0057] Furthermore, the abnormal execution value is obtained by the following method: The difference between each abnormal execution parameter and the output execution sample set is processed to obtain each abnormal execution difference; each abnormal execution difference is divided by the time length corresponding to the abnormal time window to obtain each unit abnormal execution difference; each unit abnormal execution difference is normalized with the output execution sample set and then multiplied by the corresponding weight (which is a pre-stored sample weight) to obtain each product value; the product values ​​are then superimposed to obtain the abnormal execution value. The abnormal execution value is used to characterize the degree of abnormal cumulative offset of the cooling device's execution parameters.

[0058] In this embodiment, by performing absolute difference analysis on each abnormal execution parameter of the cooling device, a quantitative characterization mechanism for abnormal execution values ​​is established, enabling accurate assessment of the cumulative deviation of abnormal execution parameters of the cooling device. By performing difference processing on the abnormal execution parameters and the output execution sample set, and combining this with the time length of the abnormal time window to calculate the unit abnormal execution difference, followed by normalization and weighted summation, a comprehensive abnormal execution value is obtained. This value comprehensively reflects the dynamic abnormal performance of the cooling device and its cumulative impact within a specific time period.

[0059] The decision to analyze only abnormal execution parameters, rather than processing all parameters, is based on the consideration that normal parameter fluctuations in the cooling device are within a controllable range and contribute little to health scoring and maintenance decisions. Analyzing all parameters would introduce interference signals and reduce prediction accuracy. Processing only abnormal parameters can significantly reduce the amount of data computation, ensuring real-time judgment and rapid response in high-frequency data acquisition environments. Therefore, this invention achieves highly sensitive monitoring of the cooling device's status and quantitative assessment of the cumulative effects of anomalies through refined analysis of abnormal execution parameters, thereby improving the accuracy of predictive maintenance and the stability of system operation, and ensuring that the cooling device can receive timely and effective maintenance intervention under multivariate coupled disturbances.

[0060] Furthermore, the machine learning algorithm model is updated to obtain the result label. Specifically, the difference between the predicted first health score and the actual health score at the target time is calculated to obtain the absolute difference in health scores. Based on the absolute difference in health scores, a mapping and matching process is performed to obtain the data removal amount. If the actual health score at the target time is lower than the predicted first health score, the historical basic operating condition characteristic parameters are removed according to the data removal amount based on the time sequence. If the actual health score at the target time is greater than the predicted first health score, the fault data of the historical basic operating condition characteristic parameters within a preset time period are removed according to the data removal amount based on the time sequence.

[0061] In this embodiment, the data removal amount is obtained by mapping and matching based on the absolute difference of health scores. The specific method is as follows: obtain the health score sample set stored in the database. The health score sample set includes several health score difference sample values. By comparing the current absolute difference of health scores with each health score difference sample value, the health score difference sample value that is closest to the current absolute difference of health scores is taken as the reference health score difference sample value. The data removal amount sample value corresponding to the reference health score difference sample value is obtained as the data removal amount.

[0062] By updating the machine learning algorithm model, the consistency between the predicted health score and the actual operating status of the equipment is optimized. Because production lines in the intelligent manufacturing industry operate under complex conditions, historical data may contain too many abnormal samples or noisy data, causing bias in the machine learning algorithm during training. This results in the predictive model outputting an abnormal score when the equipment is actually in a normal state. To eliminate this bias, this invention maps and matches the difference between the actual health score and the predicted value at the target time to determine the amount of data to be removed, and adjusts the historical basic operating condition characteristic parameters accordingly. Specifically, when the actual health score at the target time is lower than the predicted value, it indicates that the equipment may have abnormal behavior that has not yet been fully learned by the model. To avoid over-adjustment, a portion of historical data, including both normal and abnormal samples, is removed in chronological order to reduce the excessive influence of abnormal samples on the model. When the actual health score at the target time is higher than the predicted value, it indicates that the normal state of the equipment is misjudged as abnormal by the model. In this case, abnormal samples related to faults in the historical data are removed to prevent erroneous learning from further affecting prediction accuracy. This data removal and model update mechanism can eliminate the negative impact of historical abnormal data on machine learning models, enabling models to gradually optimize their learning capabilities, improve the accuracy and reliability of health score predictions, and thus generate more accurate operational result labels to support predictive maintenance decisions for smart manufacturing production lines.

[0063] like Figure 6 As shown, Figure 6 This is a schematic diagram of the modules of a predictive maintenance system for a smart manufacturing production line provided in this application embodiment. The system includes: a production status analysis module, a defensive adjustment confirmation module, and an operation correction module. The production status analysis module is used to retrieve basic operating condition characteristic parameters of the manufacturing production line within a preset time period after the production line has been continuously operating for a set operating cycle or has experienced abnormal fluctuations. Based on these basic operating condition characteristic parameters, a machine learning algorithm model is used to analyze the operating status of the cooling device of the production line, thereby generating a health score and providing a corresponding operation result label. The defensive adjustment confirmation module is used to maintain the current operating status when the operation result label is an accurate prediction result; otherwise, based on the operating status performance of the cooling device, a verification label for the operation result is determined, and a maintenance adjustment plan is obtained simultaneously. The operation correction module is used to perform a consistency judgment based on the verification label and the operation result label. If they are consistent, corresponding predictive maintenance is implemented based on the maintenance adjustment plan; otherwise, the machine learning algorithm model is updated to obtain the operation result label.

[0064] like Figure 7 As shown, Figure 7The diagram illustrates the maintenance and inspection process of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment. The diagram shows the trend of the predicted and actual values ​​of the cooling device bearing temperature over 1400 seconds, the temperature error value predicted by machine learning, and the trend of the actual and predicted values ​​of the friction coefficient during this process.

[0065] like Figure 8 As shown, Figure 8 This is a health score prediction diagram of the maintenance detection process of the predictive maintenance method for intelligent manufacturing production lines provided in this application embodiment. When the time is 0, it represents the current state. Thus, the health score prediction value can be predicted for the next 1000 seconds after the current state. Here, only the health score prediction value after 1000 seconds is shown, but it is not limited to predicting the health score within 100 seconds.

[0066] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0067] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0068] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0069] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0070] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0071] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A predictive maintenance method for production lines in the intelligent manufacturing industry, characterized in that, Includes the following steps: After a manufacturing production line has been in continuous operation and has reached a set operating cycle or experienced abnormal fluctuations, the basic operating condition characteristic parameters of the manufacturing production line within a preset time period are retrieved. The basic operating condition characteristic parameters include load fluctuation amplitude, average temperature rise rate of equipment group, cycle time offset, and global flow rate change rate of coolant. Based on the basic operating condition characteristic parameters, the operating status of the cooling device of the manufacturing production line is analyzed by a machine learning algorithm model, thereby generating a health score and giving a corresponding operating result label. When the operation result label indicates that the prediction result is accurate, the current operation status is maintained; otherwise, based on the operation status performance of the cooling device, the operation result verification label is determined, and a maintenance and adjustment plan is obtained simultaneously. The operating status of the cooling device includes: dynamic behavior status; Based on the dynamic behavior status, a verification label for the running results is determined, and a maintenance and adjustment plan is obtained synchronously. Based on the maintenance and adjustment plan, a defensive maintenance signal is generated after the cooling process execution cycle, thereby triggering the execution of the corresponding maintenance and adjustment plan. The specific method is as follows: The high-frequency data parameters of the cooling device at each moment within the abnormal time window are obtained. The high-frequency data parameters include the transient flow rate of the coolant, the frequency of pressure pulsation, the reciprocating motion deviation of the actuator, the bearing vibration amplitude, the temperature rise rate, and the current fluctuation intensity. If the high-frequency data parameters at each time point are within the preset range of high-frequency data parameters within the abnormal time window, the operation result verification label is normal, and the maintenance and adjustment plan is not to be adjusted. If the transient flow rate of the coolant increases and the frequency of pressure pulsation fluctuates periodically, the operation result verification label indicates abnormal operation. In this case, the maintenance and adjustment plan is to reduce the output pressure of the cooling pump. If the reciprocating motion deviation of the actuator increases and the bearing vibration amplitude deviates, the operation result verification label will be "operational abnormality". In this case, the maintenance and adjustment plan is to reduce the speed of the regulating pump. If the rate of temperature rise and the intensity of current fluctuation increase simultaneously, the operation result verification label will be "operational abnormality". In this case, the maintenance and adjustment plan is to increase the flow rate of the adjustment pump. Otherwise, if the verification label of the running result is "serious abnormality", the maintenance and adjustment plan is to issue a warning message and stop the operation of the manufacturing production line. The health score is obtained through the following steps: Retrieve basic operating condition characteristic parameters of the production line in the manufacturing industry within a preset time period. The basic operating condition characteristic parameters are used to characterize the basic operating conditions of the overall operating status of the production line in the intelligent manufacturing industry. After preprocessing the basic working condition characteristic parameters, they are combined with historical basic working condition characteristic parameters and input into a preset machine learning algorithm model to output the current health score and the predicted health score for each time within a preset time period. The consistency between the verification label and the operation result label is determined. If they are consistent, corresponding predictive maintenance is implemented based on the maintenance and adjustment plan. Otherwise, the machine learning algorithm model is updated to obtain the operation result label.

2. The predictive maintenance method for intelligent manufacturing production lines as described in claim 1, characterized in that: The specific method for defining the output label is as follows: Select the predicted health score value at a certain moment within a preset time period, mark it as the first predicted health score value, and mark that moment as the target moment; Obtain the actual health score at the target time and perform absolute difference processing with the first predicted health score to obtain the absolute difference of the health score; Obtain the preset health score difference threshold and compare it with the absolute difference of the health score. If the absolute difference of the health score is below the health score difference threshold, the result is labeled as accurate prediction; otherwise, the result is labeled as deviated prediction.

3. The predictive maintenance method for intelligent manufacturing production lines as described in claim 1, characterized in that: The cooling process execution cycle consists of the following steps: Obtain high-frequency data parameters at the first moment within the abnormal time window and use them as reference data; Based on the comparison of high-frequency data parameters at each time point within the abnormal time window with the control data, if the changing trend of a certain high-frequency data parameter is the same as the preset changing trend in the maintenance and adjustment plan, then the parameter is marked as an abnormal management parameter, thereby obtaining each abnormal management parameter; Based on the high-frequency data parameters at the last moment, the comparison and normalization processes are performed with the control data to obtain the changing trend of the high-frequency data parameters. If the high-frequency data parameters at the last moment within the abnormal time window belong to the abnormal management parameters, the abnormal processing value is obtained; otherwise, the preset processing value is obtained. The abnormal processing value and the preset processing value are then multiplied by their respective weights and superimposed to obtain the comprehensive interference complexity value. The comprehensive interference complexity value is used to characterize the degree of composite influence of multivariable coupled disturbances of the cooling device. The comprehensive interference complexity value is compared with the preset comprehensive interference complexity threshold. When the comprehensive interference complexity value is less than the comprehensive interference complexity threshold, the preset cooling cycle is obtained as the cooling processing execution cycle. Otherwise, the first delay shortening time is obtained based on the comprehensive interference complexity value, and the cooling execution time of the cooling processing execution cycle is obtained by subtracting the first delay shortening time from the cooling time corresponding to the cooling cycle.

4. The predictive maintenance method for intelligent manufacturing production lines as described in claim 1, characterized in that: The operating status of the cooling device also includes the execution characteristic status, which is as follows: Based on the execution characteristic status, determine the execution result verification label and simultaneously obtain the maintenance and adjustment plan. Based on the maintenance and adjustment plan, perform defensive maintenance after the cooling process execution cycle. The specific method is as follows: The actual execution parameters of the cooling device within the abnormal time window are obtained, and the actual execution parameters are used to characterize the overall performance of the cooling device under heat dissipation conditions. Obtain the corresponding output execution sample set from the database and compare it with the actual execution parameters one by one. If there is an actual execution parameter that exceeds the preset offset range of the corresponding output execution sample value, mark the parameter as an abnormal execution parameter, and thus obtain each abnormal execution parameter. Based on the analysis of each abnormal execution parameter, the abnormal execution value is obtained; Obtain the preset abnormal execution threshold in the database and compare it with the abnormal execution value. If the abnormal execution value is less than the abnormal execution threshold, obtain the preset cooling time as the cooling processing execution cycle. Otherwise, obtain the second delay shortening time based on the abnormal execution value and obtain the cooling processing execution cycle by subtracting the second delay shortening time from the preset cooling time. Therefore, the frequency of the variable frequency drive of the cooling pump motor is increased and adjusted during the cooling process execution cycle.

5. The predictive maintenance method for intelligent manufacturing production lines as described in claim 4, characterized in that: The specific method for obtaining the abnormal execution value is as follows: The execution difference of each exception is obtained by performing difference processing on the execution parameters of each exception and the output execution sample set; The abnormal execution difference is obtained by dividing each abnormal execution difference by the time length corresponding to the abnormal time window. The abnormal execution difference of each unit is then normalized with the output execution sample set and multiplied by the corresponding weight to obtain the product value. The product values ​​are then superimposed to obtain the abnormal execution value. The abnormal execution value is used to characterize the degree of abnormal cumulative offset of the execution parameters of the cooling device.

6. The predictive maintenance method for intelligent manufacturing production lines as described in claim 1, characterized in that: The specific method for updating the machine learning algorithm model to obtain the running result label is as follows: The difference between the predicted first health score and the actual health score at the target time is calculated to obtain the absolute difference in the health score. The amount of data to be removed is obtained by mapping and matching based on the absolute difference in health scores; If the actual health score at the target time is below the first predicted health score, then the historical basic working condition characteristic parameters will be removed according to the amount of data removed based on the time sequence. If the actual health score at the target time is greater than the predicted value of the first health score, then the fault data of the historical basic working condition characteristic parameters within the preset time period are processed by removing data according to the data removal amount based on the time sequence, thereby obtaining the remaining data and using it as the input data after the machine learning algorithm model is updated.

7. A system applying the predictive maintenance method for intelligent manufacturing production lines as described in any one of claims 1-6, characterized in that, include: Production status analysis module, defensive adjustment confirmation module, and operation correction module; The production status analysis module is used to retrieve the basic operating condition characteristic parameters of the manufacturing industry production line within a preset time period after the production line has been in continuous operation and has reached the set operating cycle or has experienced abnormal fluctuations. Based on the basic operating condition characteristic parameters, the module analyzes the operating status of the cooling device of the manufacturing industry production line through a machine learning algorithm model, thereby generating a health score and giving the corresponding operating result label. The defensive adjustment confirmation module is used to maintain the current operating state when the operating result label indicates that the prediction result is accurate; otherwise, it determines the operating result verification label based on the operating state performance of the cooling device and obtains the maintenance and adjustment plan simultaneously. The operation correction module is used to make a consistency judgment between the operation result verification label and the operation result label. If they are consistent, the module will implement corresponding predictive maintenance based on the maintenance adjustment plan. Otherwise, the module will update the machine learning algorithm model to obtain the operation result label.

Images