Methods, systems, equipment, and media for planning production equipment inspection tasks considering production load.

Abstract

This invention relates to the field of power production equipment maintenance technology, specifically providing a method, system, equipment, and medium for planning production equipment inspection tasks considering production load. The method includes: acquiring product inspection reports and equipment operating status data, wherein the status data includes equipment status parameters and real-time load rate; inputting the equipment status data into a pre-trained multi-output random forest model to obtain an expected product quality deviation vector; extracting the actual quality deviation vector of non-conforming products and calculating its similarity to the expected vector as the equipment contribution; correcting the contribution based on state mutation factors and load mutation factors to lock onto the target equipment; calculating the risk urgency based on the contribution, equipment anomaly degree, and historical fault frequency, and determining the inspection time in conjunction with load forecasting. This invention achieves precise source locking and adaptive inspection planning.

Description

Technical Field

[0001] This invention belongs to the field of power production equipment maintenance technology, specifically relating to a method, system, equipment, and medium for planning production equipment inspection tasks considering production load. Background Technology

[0002] The manufacturing process of power transmission and transformation equipment involves a large number of critical production equipment, such as drying tanks, winding machines, testing equipment, and gas current / voltage detection equipment. The operating status of these devices directly affects product quality and production efficiency. Existing inspection methods acquire historical inspection data, real-time production operation parameters, equipment operation parameters, and quality parameters, and then analyze the data using an inspection analysis model. This model dynamically generates inspection tasks based on failure probability, estimated total loss, and urgency priority. This method can trace related equipment when product quality anomalies are detected and adjust the inspection cycle based on failure risk, thus solving to some extent the problems of resource waste and delayed problem detection caused by fixed-cycle inspections.

[0003] However, in terms of target equipment location, existing solutions mainly rely on process tracing (such as finding the machine number through batch number) or simple fault probability correlation, resulting in inaccurate location. More importantly, existing solutions do not consider the nonlinear coupling effect of production load rate on the relationship between equipment operating status and product quality. Specifically, under the same vibration, temperature, and other state parameters, when the production line is running at high load, the actual stress and wear on the equipment are significantly amplified, and its negative impact on product quality is magnified; conversely, when running at low load, the same state parameters may not pose a substantial risk. Existing technology treats the relationship between equipment status and quality as static or linear, ignoring the modulation effect of load, leading to two typical problems: under high load conditions, normal fluctuations in equipment status are misjudged as the root cause of quality problems due to the load amplification effect; under low load conditions, the actual abnormal state of the equipment is not fully reflected in product quality deviations due to the lower load.

[0004] Therefore, how to accurately identify the target equipment most likely to cause quality problems is a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the prior art, the present invention provides a method, system, equipment and medium for planning production equipment inspection tasks that take into account production load, so as to solve the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides a method for planning production equipment inspection tasks considering production load, comprising: Obtain product testing reports and production equipment operating status data; the operating status data includes equipment status parameters and the real-time load rate of the workstation where the equipment is located; The operating status data of each device is input into a pre-trained random forest model to obtain the expected product quality deviation vector of each device in the current state. Extract the quality deviation vector of the non-conforming product from the product inspection report, calculate the similarity between the expected product quality deviation vector and the quality deviation vector, and use it as the quality deviation contribution of the corresponding equipment. Based on the operating status data of the equipment within a preset time window, the state mutation factor and the load mutation factor are calculated respectively, and the corresponding quality deviation contribution is corrected by the state mutation factor and the load mutation factor. At least one equipment with the highest quality deviation contribution after correction is selected as the target equipment. The risk urgency is calculated based on the quality deviation contribution of the target equipment and the equipment status parameters. The inspection time is determined according to the risk urgency and the real-time load rate, and an inspection task containing the target equipment and the inspection time is generated.

[0007] In an optional implementation, the random forest regression model adopts a multi-output structure, with its output dimension consistent with the dimension of the product quality deviation vector. The leaf nodes of each decision tree store the predicted value of the corresponding dimension, and the final output of the model is the mean of the prediction results of all decision trees.

[0008] In an optional implementation, extracting the quality deviation vector of non-conforming products from the product inspection report includes: Obtain the measured values ​​of the non-conforming products for each test item, as well as the corresponding standard values ​​for each test item; For each test item, the difference between the measured value and the standard value is calculated, and the difference is divided by the standard value. The result is used as the quality deviation component of the corresponding test item. According to the order of the test items in the product test report, the quality deviation components of each test item are combined sequentially to form the quality deviation vector.

[0009] In an optional implementation, based on the device's operating status data within a preset time window, the state mutation factor and load mutation factor are calculated respectively, including: Obtain the device status parameters and load rate at the current time and the start time of the preset time window; Calculate the norm of the difference between the device state parameters at the current time and the starting time, divide the norm of the difference by the norm of the device state parameters at the starting time, and use the resulting ratio as the state change factor. Calculate the absolute value of the difference between the load rate at the current time and the load rate at the start time, divide the absolute value of the difference by the load rate at the start time, and use the resulting ratio as the load mutation factor.

[0010] In an optional implementation, the risk urgency is calculated based on the quality deviation contribution of the target equipment and the equipment status parameters, including: Obtain the corrected quality deviation contribution of the target device; The real-time anomaly level of the device is calculated based on the extent to which each key operating parameter of the target device exceeds its preset normal range at the current moment. The ratio of the number of faults to the runtime of the target device within a historical time period is obtained as the historical fault frequency. The corrected quality deviation contribution, the real-time abnormality of the equipment, and the historical failure frequency are each assigned a corresponding weight, and the weighted sum of the three is calculated to obtain the risk urgency.

[0011] In an optional implementation, determining the inspection time based on the risk urgency and the real-time load rate includes: Set a first risk threshold and a second risk threshold, wherein the first risk threshold is greater than the second risk threshold; Set an upper limit for the allowable load rate of inspections; Obtain the real-time load rate at the current moment and the predicted load curve for a period of time in the future; If the risk urgency is greater than or equal to the first risk threshold, then the current moment is determined as the inspection moment; If the risk urgency is between the second risk threshold and the first risk threshold, then find the first moment in the predicted load curve that satisfies the load rate being less than or equal to the upper limit of the load rate, and determine that moment as the inspection moment; If the risk urgency is less than or equal to the second risk threshold, the original scheduled inspection time for the equipment remains unchanged.

[0012] In an optional implementation, it further includes: After the inspection task is completed, the inspection result data is obtained and the inspection result data is converted into a structured anomaly feature vector. The structured anomaly feature vector includes the abnormal location, abnormal pattern, severity and development trend. The structured anomaly feature vector is input into a pre-trained root cause classification model to obtain at least one candidate root cause and its corresponding confidence level, and the root cause with the highest confidence level is determined as the main root cause. According to the main root cause query, the preset periodic correction mapping table is used to obtain the corresponding periodic correction coefficient, and the periodic correction coefficient is multiplied by the current base period to obtain the interval of the next round of inspection. Based on the main root cause, query the preset inspection content rule base to generate the inspection item list for the next round of inspection; The interval duration and the list of inspection items are used as the updated inspection task parameters.

[0013] Secondly, the present invention provides a production equipment inspection task planning system that considers production load, comprising: The data acquisition module is used to acquire product testing reports and operating status data of production equipment; the operating status data includes equipment status parameters and the real-time load rate of the workstation where the equipment is located. The deviation prediction module is used to input the operating status data of each device into a pre-trained random forest model to obtain the expected product quality deviation vector of each device in the current state. The contribution calculation module is used to extract the quality deviation vector of the non-conforming product from the product inspection report, calculate the similarity between the expected product quality deviation vector and the quality deviation vector, and use it as the quality deviation contribution of the corresponding equipment. The contribution correction module is used to calculate the state mutation factor and the load mutation factor based on the operating status data of the equipment within a preset time window, and to correct the corresponding quality deviation contribution degree using the state mutation factor and the load mutation factor, and to select at least one equipment with the highest quality deviation contribution degree after correction as the target equipment. The task generation module is used to calculate the risk urgency based on the quality deviation contribution of the target equipment and the equipment status parameters, and determine the inspection time according to the risk urgency and the real-time load rate, and generate an inspection task that includes the target equipment and the inspection time.

[0014] Thirdly, a device is provided, comprising: Memory, used to store production equipment inspection task planning programs that take production load into account; A processor is configured to implement the steps of the production equipment inspection task planning method considering production load as provided in the first aspect when executing the production equipment inspection task planning program that considers production load.

[0015] Fourthly, a computer-readable medium is provided, on which a production equipment inspection task planning program considering production load is stored, wherein when the production equipment inspection task planning program considering production load is executed by a processor, the steps of the production equipment inspection task planning method considering production load provided in the first aspect are implemented.

[0016] The beneficial effects of this invention are that the production equipment inspection task planning method, system, equipment, and medium provided by this invention, which considers production load, reveal the nonlinear coupling relationship between equipment operating status, production load rate, and product quality deviation. By introducing a load correction mechanism to dynamically adjust the equipment contribution, it eliminates the bias of misjudging normal state fluctuations as the root cause of problems under high load conditions and also compensates for the deficiency of underestimating the contribution of true abnormal states under low load conditions, significantly improving the accuracy and robustness of target equipment positioning. Determining the inspection timing based on risk urgency and real-time load rate ensures that high-risk equipment can be inspected in a timely manner while avoiding forced shutdowns during high-load periods that could impact production. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.

[0019] Figure 2 This is a schematic flowchart illustrating the inspection task optimization process of a method according to an embodiment of the present invention.

[0020] Figure 3 This is a schematic block diagram of a system according to an embodiment of the present invention.

[0021] Figure 4 This is a schematic diagram of the structure of a device provided in an embodiment of the present invention. Detailed Implementation

[0022] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0024] The production equipment inspection task planning method considering production load provided in this embodiment of the invention is executed by computer equipment, and correspondingly, the production equipment inspection task planning system considering production load runs in computer equipment.

[0025] Figure 1 This is a schematic flowchart illustrating a method according to an embodiment of the present invention. Wherein, Figure 1 The executing entity can be a production equipment inspection task planning system that takes into account production load. Depending on different requirements, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0026] like Figure 1 As shown, the method includes: S1. Obtain product testing reports and operating status data of production equipment; the operating status data includes equipment status parameters and the real-time load rate of the workstation where the equipment is located; S2. Input the operating status data of each device into the pre-trained random forest model to obtain the expected product quality deviation vector of each device in the current state; S3. Extract the quality deviation vector of the non-conforming product from the product inspection report, calculate the similarity between the expected product quality deviation vector and the quality deviation vector, and use it as the quality deviation contribution of the corresponding equipment. S4. Based on the operating status data of the equipment within a preset time window, calculate the state mutation factor and the load mutation factor respectively, and use the state mutation factor and the load mutation factor to correct the corresponding quality deviation contribution. Select at least one equipment with the highest corrected quality deviation contribution as the target equipment. S5. Calculate the risk urgency based on the quality deviation contribution of the target equipment and the equipment status parameters, and determine the inspection time according to the risk urgency and the real-time load rate, and generate an inspection task that includes the target equipment and the inspection time.

[0027] In one embodiment of the present invention, based on step S1, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0028] S101. Obtain product testing report Product inspection reports originate from the end of the production line or quality inspection stations at each stage of the process. For power transmission and transformation equipment manufacturing, the inspection report should include at least the following information: Unique Product Identifier: Batch ID or Serial Number can be used to trace the equipment and processing time of each product or batch of products.

[0029] Test items and measured values: Depending on the product type, test items include, but are not limited to, insulation resistance (unit: MΩ), turns ratio error (percentage), partial discharge (unit: pC), no-load current, and load loss. Actual measured values ​​are recorded for each item.

[0030] Judgment result: Based on the preset qualification standards (such as insulation resistance ≥1000MΩ, transformation ratio error ≤±0.5%, partial discharge ≤50pC), the product will be marked as "qualified" or "unqualified".

[0031] Specifically, through data interfaces with Enterprise Resource Planning (ERP) or Manufacturing Execution System (MES), inspection reports are acquired in batches at preset time intervals (e.g., every hour, every shift, or every day). For non-conforming products, the system automatically extracts the measured values ​​and corresponding standard values ​​of all inspection items, and calculates the relative deviations of each item to form a quality deviation vector.

[0032] S102. Obtain real-time operating status data of production equipment. Real-time operating status data of production equipment is acquired through various sensors and data acquisition systems deployed on the equipment. The data acquisition frequency can be set according to the importance of the equipment and the rate of parameter change, such as once per second or once per minute. The acquired data includes two types: Equipment status parameters: These refer to key physical quantities that reflect the health status of the equipment itself, including but not limited to: vibration amplitude (unit: mm / s or g), obtained by an accelerometer mounted on the bearing housing or casing; temperature (unit: °C), measured by a thermocouple or infrared sensor of the winding, bearing, or cooling medium temperature; current (unit: A), measured by a current transformer of the motor or main circuit current; pressure (unit: MPa), measured by a pressure transmitter of the hydraulic or pneumatic system pressure; other parameters such as speed, flow rate, and oil level can be selectively collected according to the equipment type.

[0033] Real-time load rate of the workstation where the equipment is located: The load rate is defined as the ratio of the current actual throughput of the workstation to its rated throughput, expressed as a percentage. For example, if a winding machine has a rated capacity of 10 coils per hour and is currently winding 8 coils per hour, then the load rate is 80%. The load rate can be calculated in real time from the production progress data in the MES, or it can be obtained by directly reading the current load command or equivalent conversion of the current duty cycle from the equipment controller.

[0034] S103. Data Alignment and Storage Since the timestamps in the product testing report may not be consistent with the time granularity of the equipment operation status data, this embodiment adopts a time window alignment method: for each product, the start and end time periods of its passage through each key device are determined, and the average value, maximum value or rate of change of each device status parameter within the time period, as well as the average load rate of the time period, are extracted to form the equipment status-load dataset corresponding to the product.

[0035] All acquired raw data and processed aligned data are stored in a time-series database (such as InfluxDB) or a relational database (such as MySQL), and indexed by key values ​​such as device ID, timestamp, and product batch number to facilitate efficient retrieval and model training.

[0036] In one embodiment of the present invention, based on step S2, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0037] S201. Construction and Training of Random Forest Model This embodiment uses a multi-output random forest regression model (Multi... The model (outputRandomForestRegressor) can predict multiple target variables simultaneously, which naturally matches the product quality deviation vector (containing deviation components of multiple detection items) in this invention.

[0038] 1. Construction of training samples Extract a training sample set from historical data. Specifically, for each manufactured product (including qualified and unqualified products), perform the following operations: Identify every critical piece of equipment the product passes through on the production line. And its corresponding processing time period.

[0039] Extract devices within this time period The operational status data is used to form the input feature vector. This vector should include at least: statistical characteristics of the vibration amplitude (such as mean and peak value), temperature change rate, current fluctuation coefficient, average pressure, and average load rate over the specified time period. .

[0040] Extract the product inspection report and calculate its actual quality deviation vector. The relative deviation between the measured value and the standard value of each test item is used as the output label for the device.

[0041] The above and A training sample is formed. All historical samples from all devices are aggregated to form the training dataset.

[0042] 2. Model Structure and Parameter Settings The random forest regression model is composed of multiple decision trees. In this embodiment, the number of decision trees is set to 100. The maximum depth of each tree is set to 10 to 20 layers according to the number of features (usually 5 to 10 dimensions). The minimum number of samples in the leaf nodes is set to 5 to balance the fitting ability and generalization performance.

[0043] The key multi-output implementation methods are as follows: Model input: Input feature vector of a single sample (dimension) ,For example ).

[0044] Model output: Product quality deviation vector Its dimensions Equal to the number of items tested (e.g.) Corresponding to insulation resistance deviation, turns ratio error deviation, and partial discharge quantity deviation.

[0045] During the training of each decision tree, the splitting criterion for a node uses the mean squared error (MSE) as a comprehensive evaluation of multiple output vectors. Specifically, for the sample set at a node, the sum of the mean squared errors of all its output dimensions is calculated as the splitting basis to find the optimal splitting feature and threshold.

[0046] 3. Leaf node storage structure Each leaf node of each decision tree stores a string of length... The vector is equal to the mean of the output label vectors of all training samples falling into the leaf node. For example, if a leaf node contains 10 training samples, each corresponding to a 3-dimensional quality bias vector, then the 10 vectors are stored in the leaf node as the 1 / 2 average of each dimension (i.e., the arithmetic mean of each dimension).

[0047] S202. Model Prediction Process When it is necessary to predict the current state of a device, perform the following steps: Step 1: Construct the input vector The current operating parameters of the equipment (vibration, temperature, current, pressure, etc.) and the real-time load rate are combined into an input vector in the same order and normalization method as the training samples. .

[0048] Step 2: Each decision tree makes independent predictions For each decision tree in the random forest ( , Starting from the root node, based on the input vector The values ​​of each feature are compared with the node splitting conditions, navigating downwards layer by layer until finally landing on a leaf node. The pre-stored output vector in that leaf node is then read. .

[0049] Step 3: Integrate the prediction results of all trees The arithmetic mean of the predictions from all decision trees is taken to obtain the final output expected product quality deviation vector:

[0050] Should It is a length of vectors, for example .

[0051] Step 4: Output Results Use this vector as a device The expected product quality deviation vector under the current state and load is used for subsequent similarity calculations with the actual deviation vector of non-conforming products.

[0052] S103. Model Update and Maintenance To adapt to long-term changes such as equipment aging and process modifications, this embodiment adopts a periodic incremental training strategy. Newly collected data (equipment status, load rate, actual product quality deviation) are added to the training set monthly or quarterly, and the random forest model is retrained or incrementally updated using an online random forest algorithm.

[0053] In one embodiment of the present invention, based on step S3, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0054] S301. Extract the quality deviation vector of non-conforming products from the product inspection report. When a product deemed non-conforming appears on the production line, the following sub-steps are triggered: Step 1: Obtain basic testing data for non-conforming products Retrieve the complete inspection report for the non-conforming product from the Manufacturing Execution System (MES) or quality database. The inspection report must include at least the following fields: unique product identifier (batch number or serial number), name and measured value of each inspection item, and the corresponding pass / fail criterion value for each inspection item (usually stored as upper / lower limits or standard values). The inspection items are predetermined based on the specific type of power transmission and transformation equipment. For example, for transformers, typical inspection items include insulation resistance, turns ratio error, partial discharge, and no-load current; for reactors, they may include inductance and quality factor.

[0055] Simultaneously, the standard value corresponding to each test item is obtained. The standard value comes from the company's internal control standards or national standards. Taking the transformer turns ratio error as an example, the standard value is usually expressed as "rated turns ratio", with an allowable deviation of ±0.5%; the standard value for insulation resistance is "not less than 1000MΩ". For the sake of uniform calculation, this implementation method interprets the standard value as the ideal target value (for upper limit type indicators, the standard value is the upper limit; for lower limit type indicators, the standard value is the lower limit; for range type indicators, the standard value is the midpoint or nominal value of the range, and the deviation calculation is based on the range exceeding the range).

[0056] Step 2: Calculate the quality deviation component for each test item. For each inspection item, the quality deviation component is calculated using the relative deviation method. The specific calculation formula is as follows: Quality deviation component = (Measured value - Standard value) / Standard value The measured value represents the actual measurement result of the product in this project, while the standard value represents the target value set in advance for this project. This formula converts test items of different dimensions and orders of magnitude into dimensionless relative deviations, facilitating subsequent vector calculations.

[0057] This explanation will be based on three key testing items of a transformer: Insulation resistance: The standard value is 1000MΩ, and the measured value is 820MΩ. Therefore, the deviation component = (820-1000) / 1000 = -0.18. A negative value indicates that the measured value is lower than the standard, which is an unfavorable deviation.

[0058] Transformer ratio error: The standard value is 0% (i.e., ideal transformation ratio), and the measured value is +0.7%. The deviation component = (0.7% - 0%) / 0% cannot be calculated because a standard value of 0 would lead to a division-by-zero error. Therefore, for the special case where the standard value is zero or close to zero, the absolute deviation is used instead: Deviation component = (Measured value - Upper limit of allowable deviation) / Upper limit of allowable deviation. Transformer ratio error usually has an upper limit of allowable deviation (e.g., 0.5%). This implementation uses the upper limit of allowable deviation as the normalization benchmark. Assuming the upper limit of allowable deviation is 0.5% and the measured value is 0.7%, then the excess is 0.2%, and the deviation component = 0.2% / 0.5% = 0.4%.

[0059] Partial discharge quantity: The standard value is 50pC, the measured value is 78pC, and the deviation component is (78-50) / 50=0.56.

[0060] More generally, for range-type indicators (such as temperature allowed to be between 10℃ and 40℃), if the measured value is within the range, the deviation component is 0; if it exceeds the upper limit, the deviation component = (measured value - upper limit) / upper limit; if it is below the lower limit, the deviation component = (lower limit - measured value) / lower limit.

[0061] Step 3: Form the quality deviation vector Following the fixed order of the test items in the product test report (e.g., insulation resistance, turns ratio error, partial discharge), the calculated quality deviation components are arranged sequentially to form a one-dimensional vector, denoted as... The dimension of this vector. It equals the number of items tested.

[0062] Continuing with the previous example, if the three components are [ [0.18, 0.4, 0.56], then the quality deviation vector This vector visually reflects the direction and extent of deviation of non-conforming products across various quality dimensions.

[0063] S302. Calculate the similarity between the expected product quality deviation vector and the actual quality deviation vector. For every piece of equipment on the production line that may have participated in the processing of the defective product. The system has obtained the expected product quality deviation vector of the equipment in its current state through the aforementioned steps. (and (Same dimensions). This implementation uses cosine similarity to measure the similarity between two vectors. Cosine similarity reflects the consistency of vectors in direction, is not sensitive to amplitude, and is suitable for comparing the similarity of deviation patterns.

[0064] Similarity calculation formula: set up , ,but:

[0065] If the product of the two moduli in the denominator is zero (when both vectors are zero vectors), then the similarity is directly defined as 0.

[0066] Calculation steps: Calculate the numerator: Multiply the components of the corresponding dimensions of the two vectors, and then sum them.

[0067] Calculate the denominator: Calculate the magnitude of each vector (the square root of the sum of the squares of each component) and then multiply them.

[0068] Divide the numerator by the denominator to obtain the similarity value, which ranges from -1 to 1.

[0069] A similarity close to 1 indicates that the expected deviation vector and the actual deviation vector are almost in the same direction. This means that under the current state and load conditions of the equipment, the expected quality deviation pattern highly matches the actual deviation pattern exhibited by the non-conforming product. Therefore, the equipment contributes significantly to this non-conformity. A similarity close to 0 indicates that the two vectors are orthogonal and the deviation patterns are unrelated. A negative similarity value indicates that the directions are opposite, and the equipment may actually mitigate the deviation in the opposite direction.

[0070] S303. Output quality deviation contribution The calculated similarity is directly used as the device. Contribution of quality deviation to current non-conforming products ,Right now:

[0071] For each piece of equipment on the production line involved in processing this product, perform the above calculations one by one to obtain a set of contribution values. The higher the contribution, the more likely the equipment is to be the root cause of product quality problems.

[0072] In one embodiment of the present invention, based on step S4, the following will provide a possible embodiment and describe its specific implementation in a non-limiting manner.

[0073] S401. Setting the preset time window A sliding time window is maintained for each piece of equipment. The window length can be pre-configured based on the equipment type and process characteristics, such as setting it to the most recent 30 minutes, 1 hour, or a shift. The start point of the window is the current time minus the window length, and the end point is the current time. For scenarios with high real-time requirements, the window length can be set to 10 minutes; for equipment with slow changes (such as drying tanks), it can be set to 2 hours. This window length can also be dynamically adjusted by the user through the human-machine interface.

[0074] S402. Calculate the state change factor For each device The preset time window length is (For example, 30 minutes), the window starts at... The current time is Define the state mutation factor as:

[0075] in: For equipment The state parameter vector (including normalized values ​​of vibration, temperature, current, pressure, etc.); It represents the Euclidean norm (i.e., the square root of the sum of the squares of the vector's components). For a very small positive number (e.g.) (), used to prevent the denominator from being zero.

[0076] This formula reflects the relative change in equipment status within a time window.

[0077] S403. Calculate the load mutation factor Define the mutation factor as:

[0078] in: For equipment The real-time load rate of the workstation; It is a very small positive number, used to prevent division by zero errors when the initial load rate is zero.

[0079] (iv) Correcting the contribution of quality bias using mutation factors For each device Its original quality deviation contribution (Based on similarity calculations) It needs to be amplified or reduced according to the state mutation factor and load mutation factor to reflect the additional contribution of sudden state changes and drastic load fluctuations to the quality impact.

[0080] The corrected formula is in product form:

[0081] in: This is the original contribution (similarity) calculated in step four.

[0082] It is a state mutation factor.

[0083] It is a load mutation factor.

[0084] and The preset weighting coefficient typically ranges from 0.3 to 1.0 and can be obtained by the user based on experience or through optimization using historical data. In this embodiment, the default value is [missing value]. , This indicates that state mutations and load mutations contribute equally to the contribution correction.

[0085] The physical meaning of this correction formula is that when there is a significant change in equipment status or load, even if the original contribution is not very high, its suspicion level should be increased, because sudden changes often indicate the occurrence of abnormal events. Conversely, if the equipment status is stable and the load is steady, the original contribution level should be kept unchanged or slightly reduced.

[0086] S405. Select target device Calculate the corrected contribution for all devices. Then, sort them in descending order. Based on the preset selection strategy, perform the following operations: Select the top N devices with the highest contribution after correction (N is set according to requirements) as the target device set.

[0087] In one embodiment of the present invention, based on step S5, a possible embodiment will be given below, and its specific implementation will be described in a non-limiting manner.

[0088] S501. Calculate the urgency of the risk For each locked target device, the system calculates its risk urgency using the following sub-steps. .

[0089] Step 1: Obtain the corrected contribution of quality deviation The corrected contribution of the target device can be directly read from the aforementioned steps. This value incorporates the impact of equipment condition and load on quality deviations, reflecting the importance of equipment in quality attribution.

[0090] Step 2: Calculate the real-time anomaly level of the equipment. Real-time anomaly level of equipment This system is used to quantify the degree to which the current operating state of equipment deviates from the normal range. For each key operating parameter of the equipment (such as vibration, temperature, current, pressure, etc.), the system pre-sets its own normal operating range, including a lower limit. and upper limit value For the measured values ​​collected at the current moment. The over-limit range of this parameter is calculated according to the following rules: like exist If the parameter exceeds the limit within the specified range, the over-limit range is 0. like The over-limit range is ; like The over-limit range is .

[0091] The real-time anomaly level of the equipment is obtained by taking the arithmetic mean of the out-of-limit ranges of all key parameters. The larger this value, the more abnormal the current state of the device.

[0092] Step 3: Obtain historical fault frequencies Retrieve fault records for the target equipment from the equipment maintenance management system for a historical period (e.g., the last 30 or 90 days). Count the number of faults that occurred within that period. And calculate the total runtime of the device during that time period. (Unit: hour or day). Historical fault frequency. Calculate using the following formula:

[0093] This ratio reflects the long-term frequency of equipment failures. The higher the frequency, the lower the reliability of the equipment itself and the greater the risk.

[0094] Step 4: Calculate the weighted sum to obtain the risk urgency. Regarding the above three factors—the adjusted contribution Real-time anomaly level Historical failure frequency Assign weight coefficients respectively , , And satisfy The weighting coefficients can be adjusted by the user according to the actual scenario, for example, increasing them in scenarios that emphasize quality attribution results. Increase in scenarios where the current status of the equipment is important In this embodiment, the default value is... , , .

[0095] Risk urgency The calculation formula is:

[0096] The value ranges from 0 to 1 (each input factor is pre-normalized to this range), and the higher the value, the more urgent the equipment needs to be checked.

[0097] S502. Determine the inspection time based on risk urgency and real-time load rate. Based on the calculated risk urgency Based on the current and future production line load, the specific timing for inspection is determined according to the three-level decision-making rules.

[0098] Step 1: Set decision parameters Set the first risk threshold (e.g., 0.8) and the second risk threshold (e.g., 0.5), where .

[0099] Set the upper limit of the allowable inspection load rate. (For example, 0.6 means that it is only suitable to schedule inspections when the load rate does not exceed 60%).

[0100] Get the current time Real-time load rate And the projected load curve for a future period of time (e.g., the next 4 hours). ( The predicted load curve can be generated in real time by the production scheduling system based on order plans and scheduling information.

[0101] Step 2: Three-level decision-making logic High-risk scenario: If This indicates that the equipment poses an extremely high risk and must be inspected immediately without delay. The inspection time should be set at this point. Determined as the current time Even if the current load rate is high, inspections will be mandatory, and short-term shutdowns can be requested if necessary.

[0102] Medium-risk scenario: If This indicates a significant risk to the equipment, but it is possible to wait for a suitable low-load window. The system from the current moment... Begin, along the predicted load curve Sequential scanning, searching for the first instance that satisfies a load factor less than or equal to The moment Once found, the inspection time will be checked. Determined as If no low-load period meeting the criteria is found within the predicted timeframe (e.g., the next 4 hours), the system can issue a prompt, allowing operators to decide whether to increase the load limit or force a check.

[0103] Low-risk scenario: If This indicates that the equipment risk is low and there is no need to disrupt the original plan. The system will maintain the original scheduled inspection time for this equipment. Unchanged, that is .

[0104] Step 3: Output Inspection Task Define the target equipment and inspection time. The inspection task, along with subsequent inspection items (generated from other steps), is combined into a complete inspection task and pushed to the inspector's mobile terminal or displayed on the workshop dashboard. The inspection task includes necessary information such as equipment identification, planned execution time, and inspection item list.

[0105] Based on the above embodiments, to improve the rationality of inspection task planning, in one embodiment, the inspection task planning scheme is optimized, such as... Figure 2 As shown: S601. Acquisition and Structured Transformation of Inspection Result Data After the inspection task is completed, the inspectors enter the inspection results data through handheld terminals or computers. At the same time, the online monitoring system will also automatically upload some detection data (such as vibration spectrum, infrared thermography, etc.). These semi-structured raw data are then transformed into unified structured anomaly feature vectors.

[0106] Structured anomaly feature vectors contain information in the following four dimensions: Abnormal Location: Identifies which component or area of ​​the equipment the fault occurs in, such as "bearing," "coil," "contact," "insulating surface," or "cooling system." This information can be obtained by inspectors selecting preset locations or automatically based on image recognition and vibration localization technology.

[0107] Abnormal patterns: Describe the manifestation of the fault, such as "wear," "overheating," "loosening," "discharge," "contamination," "poor lubrication," and "deformation." Abnormal patterns can be extracted from keywords in inspection records or automatically determined through feature analysis (such as sideband components in the vibration spectrum indicating loosening).

[0108] Severity: The severity of the abnormal state is graded. This embodiment uses a four-level scoring method: Level 1 (minor, no immediate action required), Level 2 (moderate, action recommended soon), Level 3 (severe, action required as soon as possible), and Level 4 (critical, immediate shutdown required). The severity level can be obtained by mapping based on the extent of exceeding limits, expert rules, or historical fault levels.

[0109] Development trend: Reflects the rate of change of anomalies over time, and is divided into three categories: stable (no significant change compared to the last inspection), slow growth (change rate between 5% and 20%), and rapid growth (change rate exceeding 20%). The development trend is calculated by comparing the rate of change of characteristic parameters of the same location and pattern in the current inspection with those in historical inspections.

[0110] The values ​​of the above four dimensions are combined to form a structured anomaly feature vector. .

[0111] S602. Identification of principal root causes based on root cause classification model The structured anomaly feature vectors are input into a pre-trained root cause classification model, which outputs the probability distribution of each candidate root cause.

[0112] 1. Construction of Root Cause Classification Model In this embodiment, the root cause classification model employs either a Naive Bayes classifier or a random forest classifier. The samples required for model training are derived from historical inspection records. Each sample includes a structured anomaly feature vector and an actual root cause label confirmed by engineers. The root cause labels are preset based on the equipment and fault mechanism, such as "mechanical loosening," "insulation degradation," "thermal aging," "bearing wear," "process parameter drift," and "lubrication failure."

[0113] During model training, the input is an anomaly feature vector, and the output is the posterior probability of each root cause. Taking Naive Bayes as an example, for a given anomaly feature vector... Calculate each root cause Posterior probability:

[0114] Where prior probability Conditional probability is obtained from historical root cause frequency statistics. It is obtained by frequency estimation after discretizing the eigenvalues.

[0115] 2. Perform root cause prediction The structured anomaly feature vector from the current inspection is input into the model. The model outputs the probability of all candidate root causes and sorts them from highest to lowest probability. The root cause with the highest probability is determined as the primary root cause. The corresponding probability value is used as the confidence level. If the confidence level is lower than a preset threshold (e.g., 0.6), the system can push the top two root causes to the inspection engineer for manual confirmation, and use the confirmation results as new samples to update the model.

[0116] S603. Root Cause-Based Periodic Correction A built-in periodic correction mapping table records the periodic correction coefficient corresponding to each root cause. This coefficient is based on historical statistics: for a given root cause, the average remaining lifespan from the discovery of the root cause to the actual occurrence of the failure is statistically analyzed and compared with the baseline lifespan of the equipment under standard operating conditions to calculate the correction coefficient. The mapping table is shown in Table 1. Table 1 Periodic Correction Mapping Table

[0117] According to the principal root cause By querying the mapping table above, the corresponding correction coefficient can be obtained. .

[0118] Current benchmark period This can be the equipment's original fixed inspection cycle (e.g., every 7 days) or a dynamic cycle adjusted for load (e.g., the cycle calculated using the aforementioned equivalent operating load). The interval for the next round of inspections will then be determined. for:

[0119] Meanwhile, to prevent the cycle from being too short or too long, upper and lower limits are set: the minimum interval is no less than 2 hours, and the maximum interval does not exceed twice the original reference cycle.

[0120] S604. Root Cause-Based Inspection Content Generation It also maintains a rule base for inspection content, which maps each root cause to a specific set of inspection items, detection methods, and sampling frequencies. By querying the rule base, a list of inspection items for the next round of inspections can be automatically generated.

[0121] The rule base is shown in Table 2: Table 2 Inspection Content Rules Table

[0122] For those with low confidence levels (such as...) In cases where the model is uncertain, additional general check items are added when generating the list, such as appearance inspection, abnormal noise listening, and cleanliness inspection, to compensate for the uncertainty of the model.

[0123] S605. Update Inspection Task Parameters The calculated next round of inspection interval duration The generated checklist is used as the updated inspection task parameters and stored in the inspection plan database. The current time will be added... This triggers the next inspection task and pushes the corresponding checklist to the inspection personnel.

[0124] Meanwhile, the structured anomaly feature vectors from this inspection, the main root causes predicted by the model and their confidence levels, and the actual confirmed root causes (if manually corrected) are used as new training samples. The root cause classification model and the periodic correction mapping table are updated incrementally on a regular basis to achieve continuous evolution of the knowledge base.

[0125] It should be noted that this scheme assumes that the contributions of each device to product quality are additive to a certain degree of approximation, i.e., higher-order interaction terms are ignored. This assumption is based on the following engineering fact: in the manufacturing of power transmission and transformation equipment, the quality deviations of the vast majority of defective products can be attributed to significant anomalies in a single device; for the rare case of concurrent anomalies in multiple devices, this scheme covers them by selecting Top-N (N≥2) target devices. Practical application shows that this assumption significantly reduces model complexity and computational cost while ensuring attribution accuracy, demonstrating good engineering feasibility.

[0126] In some embodiments, the production equipment inspection task planning system considering production load may include multiple functional modules composed of computer program segments. The computer programs for each program segment in the production equipment inspection task planning system considering production load may be stored in the memory of a computer device and executed by at least one processor to perform (see details). Figure 1 (Description) Function to plan production equipment inspection tasks considering production load.

[0127] In this embodiment, the production equipment inspection task planning system that considers production load can be divided into multiple functional modules according to the functions it performs, such as... Figure 3 As shown. The module referred to in this invention is a series of computer program segments that can be executed by at least one processor and perform a fixed function, and is stored in memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0128] The data acquisition module is used to acquire product testing reports and operating status data of production equipment; the operating status data includes equipment status parameters and the real-time load rate of the workstation where the equipment is located. The deviation prediction module is used to input the operating status data of each device into a pre-trained random forest model to obtain the expected product quality deviation vector of each device in the current state. The contribution calculation module is used to extract the quality deviation vector of the non-conforming product from the product inspection report, calculate the similarity between the expected product quality deviation vector and the quality deviation vector, and use it as the quality deviation contribution of the corresponding equipment. The contribution correction module is used to calculate the state mutation factor and the load mutation factor based on the operating status data of the equipment within a preset time window, and to correct the corresponding quality deviation contribution degree using the state mutation factor and the load mutation factor, and to select at least one equipment with the highest quality deviation contribution degree after correction as the target equipment. The task generation module is used to calculate the risk urgency based on the quality deviation contribution of the target equipment and the equipment status parameters, and determine the inspection time according to the risk urgency and the real-time load rate, and generate an inspection task that includes the target equipment and the inspection time.

[0129] Figure 4The production equipment inspection task planning method considering production load provided in the embodiments of this application can be applied to equipment. Those skilled in the art will understand that the equipment structure involved in the embodiments of this invention does not constitute a limitation on the equipment. The equipment may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements. Specifically, the equipment 400 may include: a processor 410, a memory 420, and a communication unit 430. These components communicate through one or more buses. Those skilled in the art will understand that the server structure shown in the figures does not constitute a limitation on the invention; it may be a bus topology or a star topology, and may include more or fewer components than shown in the figures, or combine certain components, or have different component arrangements.

[0130] The present invention also provides a computer medium, wherein the computer medium may store a program, which, when executed, may include some or all of the steps provided in the embodiments of the present invention.

[0131] The same or similar parts between the various embodiments in this specification can be referred to mutually. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.

[0132] In the embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between systems or modules may be electrical, mechanical, or other forms.

[0133] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0134] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0135] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the present invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the present invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should also be covered within the protection scope of the present invention.

Claims

1. A method for planning production equipment inspection tasks considering production load, characterized in that, include: Obtain product testing reports and production equipment operating status data; the operating status data includes equipment status parameters and the real-time load rate of the workstation where the equipment is located; Input the operating status data of each device into the pre-trained random forest model to obtain the expected product quality deviation vector of each device in the current state; Extract the quality deviation vector of the non-conforming product from the product inspection report, calculate the similarity between the expected product quality deviation vector and the quality deviation vector, and use it as the quality deviation contribution of the corresponding equipment. Based on the operating status data of the equipment within a preset time window, the state mutation factor and the load mutation factor are calculated respectively, and the corresponding quality deviation contribution is corrected by the state mutation factor and the load mutation factor. At least one equipment with the highest quality deviation contribution after correction is selected as the target equipment. The risk urgency is calculated based on the quality deviation contribution of the target equipment and the equipment status parameters. The inspection time is determined according to the risk urgency and the real-time load rate, and an inspection task containing the target equipment and the inspection time is generated.

2. The method according to claim 1, characterized in that, The random forest regression model adopts a multi-output structure, and its output dimension is consistent with the dimension of the product quality deviation vector. The leaf nodes of each decision tree store the predicted value of the corresponding dimension, and the final output of the model is the mean of the prediction results of all decision trees.

3. The method according to claim 1, characterized in that, Extract the quality deviation vector of the non-conforming products from the product inspection report, including: Obtain the measured values ​​of the non-conforming products for each test item, as well as the corresponding standard values ​​for each test item; For each test item, the difference between the measured value and the standard value is calculated, and the difference is divided by the standard value. The result is used as the quality deviation component of the corresponding test item. According to the order of the test items in the product test report, the quality deviation components of each test item are combined sequentially to form the quality deviation vector.

4. The method according to claim 1, characterized in that, Based on the equipment's operating status data within a preset time window, calculate the state mutation factor and load mutation factor, including: Obtain the device status parameters and load rate at the current time and the start time of the preset time window; Calculate the norm of the difference between the device state parameters at the current time and the starting time, divide the norm of the difference by the norm of the device state parameters at the starting time, and use the resulting ratio as the state change factor. Calculate the absolute value of the difference between the load rate at the current time and the load rate at the start time, divide the absolute value of the difference by the load rate at the start time, and use the resulting ratio as the load mutation factor.

5. The method according to claim 1, characterized in that, The risk urgency is calculated based on the contribution of the target equipment's quality deviation and the equipment's status parameters, including: Obtain the corrected quality deviation contribution of the target device; The real-time anomaly level of the device is calculated based on the extent to which each key operating parameter of the target device exceeds its preset normal range at the current moment. The ratio of the number of faults to the runtime of the target device within a historical time period is obtained as the historical fault frequency. The corrected quality deviation contribution, the real-time abnormality of the equipment, and the historical failure frequency are each assigned a corresponding weight, and the weighted sum of the three is calculated to obtain the risk urgency.

6. The method according to claim 1, characterized in that, Determining the inspection time based on the risk urgency and the real-time load rate includes: Set a first risk threshold and a second risk threshold, wherein the first risk threshold is greater than the second risk threshold; Set an upper limit for the allowable load rate of inspections; Obtain the real-time load rate at the current moment and the predicted load curve for a period of time in the future; If the risk urgency is greater than or equal to the first risk threshold, then the current moment is determined as the inspection moment; If the risk urgency is between the second risk threshold and the first risk threshold, then find the first moment in the predicted load curve that satisfies the load rate being less than or equal to the upper limit of the load rate, and determine that moment as the inspection moment; If the risk urgency is less than or equal to the second risk threshold, the original scheduled inspection time for the equipment remains unchanged.

7. The method according to claim 1, characterized in that, The method further includes: After the inspection task is completed, the inspection result data is obtained and the inspection result data is converted into a structured anomaly feature vector. The structured anomaly feature vector includes the abnormal location, abnormal pattern, severity and development trend. The structured anomaly feature vector is input into a pre-trained root cause classification model to obtain at least one candidate root cause and its corresponding confidence level, and the root cause with the highest confidence level is determined as the main root cause. According to the main root cause query, the preset periodic correction mapping table is used to obtain the corresponding periodic correction coefficient, and the periodic correction coefficient is multiplied by the current base period to obtain the interval of the next round of inspection. Based on the main root cause, query the preset inspection content rule base to generate the inspection item list for the next round of inspection; The interval duration and the list of inspection items are used as the updated inspection task parameters.

8. A production equipment inspection task planning system considering production load, characterized in that, include: The data acquisition module is used to acquire product testing reports and operating status data of production equipment; The operating status data includes equipment status parameters and the real-time load rate of the workstation where the equipment is located; The deviation prediction module is used to input the operating status data of each device into a pre-trained random forest model to obtain the expected product quality deviation vector of each device in the current state. The contribution calculation module is used to extract the quality deviation vector of the non-conforming product from the product inspection report, calculate the similarity between the expected product quality deviation vector and the quality deviation vector, and use it as the quality deviation contribution of the corresponding equipment. The contribution correction module is used to calculate the state mutation factor and the load mutation factor based on the operating status data of the equipment within a preset time window, and to correct the corresponding quality deviation contribution degree using the state mutation factor and the load mutation factor, and to select at least one equipment with the highest quality deviation contribution degree after correction as the target equipment. The task generation module is used to calculate the risk urgency based on the quality deviation contribution of the target equipment and the equipment status parameters, and determine the inspection time according to the risk urgency and the real-time load rate, and generate an inspection task that includes the target equipment and the inspection time.

9. A production equipment inspection task planning device that considers production load, characterized in that, include: Memory, used to store production equipment inspection task planning programs that take production load into account; A processor is configured to implement the steps of the production equipment inspection task planning method considering production load as described in any one of claims 1-7 when executing the production equipment inspection task planning program that considers production load.

10. A computer-readable medium storing a computer program, characterized in that, The readable medium stores a production equipment inspection task planning program that takes into account production load. When the production equipment inspection task planning program that takes into account production load is executed by a processor, it implements the steps of the production equipment inspection task planning method that takes into account production load as described in any one of claims 1-7.

Images