Dynamic determination method and device of early warning threshold and storage medium

Abstract

The invention discloses a dynamic determination method and device for an early warning threshold and a storage medium, and relates to the technical field of data processing. The method comprises the following steps: firstly, determining a health index of equipment in a preset time window according to multi-source sensor data in the preset time window; then, fitting a linear attenuation equation according to the health index in the preset time window, and taking the slope of the linear attenuation equation as the attenuation rate of the health index; and then, determining an early warning threshold value of the multi-source sensor according to the attenuation rate. According to the method, the health index and the attenuation rate are introduced, the early warning threshold value can be dynamically adjusted according to the actual operation state and the degradation condition of the equipment, the limitation that a fixed threshold value cannot reflect the dynamic change of the equipment in real time is overcome, and the timeliness of equipment maintenance is improved.

Description

Technical Field

[0001] This application relates to the field of data processing technologies, and particularly to a method, device, and storage medium for dynamically determining an early warning threshold. Background Art

[0002] In the field of intelligent device maintenance, fixed thresholds are usually used to trigger alarms, such as setting a unified early warning threshold based on the device's factory parameters or historical experience values. However, the degradation of a device is a complex dynamic process affected by various factors, such as the device's usage environment, operating conditions, and aging degree. Fixed thresholds cannot reflect the actual degradation state of the device in different operating stages and environments in real time, resulting in the inability to maintain the device in a timely manner.

[0003] The above content is only used to assist in understanding the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0004] The main objective of this application is to provide a method, device, and storage medium for dynamically determining an early warning threshold, aiming to solve the technical problem of improving the timeliness of device maintenance.

[0005] To achieve the above objective, this application proposes a method for dynamically determining an early warning threshold, the method including: Determining a health index of a device within the preset time window according to multi-source sensor data within the preset time window; Fitting a linear decay equation according to the health index within the preset time window, and taking the slope of the linear decay equation as the decay rate of the health index; Determining an early warning threshold of the multi-source sensor according to the decay rate.

[0006] In an embodiment, the step of determining a health index of a device within the preset time window according to multi-source sensor data within the preset time window includes: Obtaining an operation instruction received by the device within the preset time window; Taking the time stamp corresponding to the operation instruction as a timing mark and mapping it to the multi-source sensor data to obtain a data sequence with a timing mark; Determining a target operation period of the device according to the timing mark corresponding to the operation instruction, and calculating a state transition probability according to the data sequence corresponding to the target operation period through a forward-backward algorithm; Segmenting the data sequence corresponding to the target operation period according to the state transition probability to obtain a target data sequence; Determining a health index of the device within the preset time window according to the target data sequence.

[0007] In one embodiment, the step of determining the health index of the device within the preset time window according to the target data sequence includes: Determine the historical sensor data sequences corresponding to each historical failure mode according to historical failure cases; Determine the weight of the target data sequence according to the similarity between the target data sequence and the historical sensor data sequences; Determine the health index according to the target data sequence and the weight.

[0008] In one embodiment, the step of determining the weight of the target data sequence according to the similarity between the target data sequence and the historical sensor data sequences includes: Obtain the initial weight of the target data sequence; Determine the target failure mode corresponding to the target data sequence according to the similarity between the target data sequence and the historical sensor data sequences; Determine the compensation weight of the data corresponding to the failure sensor in the target data sequence according to the failure sensor corresponding to the target failure mode; Determine the weight of the target data sequence according to the initial weight of the target data sequence and the compensation weight of the data corresponding to the failure sensor in the target data sequence.

[0009] In one embodiment, after the step of fitting a linear decay equation according to the health index within the preset time window and taking the slope of the linear decay equation as the decay rate of the health index, the method further includes: Create a health index sequence according to the health index within the preset time window; Perform wavelet packet decomposition on the health index sequence to obtain the low-frequency component corresponding to the health index sequence; Determine the Hurst exponent corresponding to the low-frequency component, and determine the adjustment amount of the decay rate according to the Hurst exponent, where the Hurst exponent represents the likelihood that the change trend of the health index sequence will continue in the future; Take the sum of the decay rate and the adjustment amount of the decay rate as the new decay rate.

[0010] In one embodiment, the step of determining the warning threshold of the multi-source sensor according to the decay rate includes: Determine the initial threshold of the multi-source sensor according to the health index; Determine the threshold adjustment amount of the multi-source sensor according to the decay rate; Determine the warning threshold of the multi-source sensor according to the initial threshold and the threshold adjustment amount.

[0011] In one embodiment, after the step of determining the warning threshold of the multi-source sensor according to the attenuation rate, the method further includes: Taking the health index and the target failure mode as decision variables, and taking maintenance resources as constraint variables, where the maintenance resources include at least one of the number of equipment spare parts, the maintenance time window, and the matching degree of maintenance staff; Inputting the decision variables and the constraint variables into a preset work order generation model, generating at least one reference maintenance plan through the preset work order generation model, and calculating the maintenance cost corresponding to the reference maintenance plan; Determining a target maintenance plan from the reference maintenance plans according to the maintenance cost, and generating a maintenance work order according to the target maintenance plan.

[0012] In one embodiment, after the step of determining a target maintenance plan from the reference maintenance plans according to the maintenance cost and generating a maintenance work order according to the target maintenance plan, it further includes: Obtaining the health index and the maintenance time consumption after the equipment maintenance; Calculating a reward function value corresponding to the target maintenance plan according to the health index after the maintenance and the maintenance time consumption; If the reward function value is less than a preset reward value, re-determining the attenuation rate of the equipment according to the health index after the maintenance; Re-determining the warning threshold of the multi-source sensor according to the re-determined attenuation rate.

[0013] In addition, to achieve the above object, the present application also proposes a device for dynamically determining a warning threshold, the device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program is configured to implement the steps of the method for dynamically determining a warning threshold as described above.

[0014] In addition, to achieve the above object, the present application also proposes a storage medium, the storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium, and when the computer program is executed by a processor, it implements the steps of the method for dynamically determining a warning threshold as described above.

[0015] The present application provides a method for dynamically determining an early warning threshold. First, based on multi-source sensor data within a preset time window, the health index of the device within the preset time window is determined; then, a linear decay equation is fitted according to the health index within the preset time window, and the slope of the linear decay equation is used as the decay rate of the health index; afterwards, the early warning threshold of the multi-source sensor is determined according to the decay rate. By introducing the health index and the decay rate, this method can dynamically adjust the early warning threshold according to the actual operating state and degradation of the device, overcoming the limitation that a fixed threshold cannot reflect the dynamic changes of the device in real time, and improving the timeliness of device maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present application and, together with the specification, are used to explain the principles of the present application.

[0017] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0018] Figure 1 It is a schematic flowchart provided for the first embodiment of the method for dynamically determining the early warning threshold of the present application; Figure 2 It is a schematic flowchart provided for the second embodiment of the method for dynamically determining the early warning threshold of the present application; Figure 3 It is a schematic flowchart of the method provided for the second real-time example of the method for dynamically determining the early warning threshold of the present application; Figure 4 It is a schematic flowchart provided for the third embodiment of the method for dynamically determining the early warning threshold of the present application; Figure 5 It is a schematic flowchart of the method for dynamically determining the early warning threshold in the embodiments of the present application; Figure 6 It is a schematic diagram of the device structure of the hardware operating environment involved in the method for dynamically determining the early warning threshold in the embodiments of the present application.

[0019] The implementation, functional features and advantages of the present application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0021] To better understand the technical solutions of the present application, the following will be described in detail in conjunction with the drawings of the specification and the specific embodiments.

[0022] In the field of intelligent device maintenance, fixed thresholds are usually used to trigger alarms, such as setting a unified warning threshold based on the factory parameters of the device or historical experience values. However, the degradation of the device is a complex dynamic process affected by various factors, such as the usage environment of the device, operating conditions, aging degree, etc. Fixed thresholds cannot reflect the actual degradation state of the device in different operating stages and environments in real time, resulting in the inability to maintain the device in a timely manner.

[0023] In view of the above problems, this application proposes a method for dynamically determining the warning threshold. First, according to the multi-source sensor data within a preset time window, the health index of the device within the preset time window is determined; then, a linear decay equation is fitted based on the health index within the preset time window, and the slope of the linear decay equation is used as the decay rate of the health index; after that, the warning threshold of the multi-source sensor is determined according to the decay rate. By introducing the health index and the decay rate, this method can dynamically adjust the warning threshold according to the actual operating state and degradation of the device, overcoming the limitation that fixed thresholds cannot reflect the dynamic changes of the device in real time and improving the timeliness of device maintenance.

[0024] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication, and program running functions, such as a tablet computer, a personal computer, etc., or an electronic device capable of implementing the above functions.

[0025] Based on this, the first embodiment proposed in this application provides a method for dynamically determining the warning threshold. Refer to Figure 1 , in this embodiment, the method for dynamically determining the warning threshold includes steps S10 to S40: Step S10, according to the multi-source sensor data within a preset time window, determine the health index of the device within the preset time window.

[0026] It can be understood that traditional fixed thresholds are set based on static parameters or experience values and cannot adapt to the degradation process of the device caused by factors such as environment, working conditions, and aging. By calculating the health index based on the multi-source sensor data within the preset time window and dynamically determining the warning threshold according to the decay rate of the health index, the interference of single-sensor data noise or false alarms can be avoided, the comprehensive health state of the device can be quantified in real time, and its degradation trajectory can be dynamically reflected.

[0027] Optionally, step S10 includes steps S11 to S15: Step S11, obtain the operation instructions received by the device within the preset time window.

[0028] Step S12: Use the timestamp corresponding to the operation instruction as a timing marker and map it to the multi-source sensor data to obtain a data sequence with timing markers.

[0029] It should be noted that operation instructions refer to the control commands or operation signals received by the device during operation. These instructions can come from manual operations, automatic control systems, or upper-level management software, including start, stop, parameter adjustment, mode switching, etc. By retrieving the operation log of the device, information such as the timestamp and operation type of the operation instruction can be obtained.

[0030] During the operation of the device, multi-source sensors such as vibration, temperature, and current sensors and operation instructions such as start / stop, operation, and load change are usually collected independently, and there may be problems with inconsistent timing. Using the timestamp of the operation instruction as a timing marker and mapping it to the multi-source sensor data can synchronize the operation events with the sensor readings. Optionally, before synchronization, preprocess the multi-source sensor data by noise reduction filtering and missing value filling.

[0031] It can be understood that within a preset fixed time window, the device may have multiple operating cycles. For example, an injection molding machine has intermittent downtime during production batches, and the preset fixed time window may contain invalid downtime data, that is, some sensor data shows 0. The operation instruction can provide real-time change information during the operation of the device. To ensure the accuracy of the health index calculation, the timestamp of the operation instruction is used as a timing marker and mapped to the multi-source sensor data, so that the data sequence with timing markers can reflect the changes in the multi-source sensor data of the device under different operating states.

[0032] Step S13: Determine the target operating period of the device according to the timing marker corresponding to the operation instruction, and calculate the state transition probability according to the data sequence corresponding to the target operating period by the forward-backward algorithm.

[0033] Step S14: Segment the data sequence corresponding to the target operating period according to the state transition probability to obtain the target data sequence.

[0034] It can be understood that when the device starts, stops, or the load changes, the sensor data will show non-linear transition characteristics such as increasing rotational speed and temperature fluctuations. Directly using the original data to calculate the health index will include such transient processes in the evaluation, resulting in fluctuations in the index. For example, the vibration amplitude of a bearing may increase briefly when starting because the lubricating oil film has not formed, but this does not mean that the bearing has deteriorated. Therefore, in order to more accurately obtain the health state of the device during stable operation, the state transition probability can be calculated for the data sequence corresponding to the target operating period of the device, and the sensor data representing the change process can be removed, leaving the target data sequence that can represent the device in a stable operating state.

[0035] Exemplarily, a Hidden Markov Model is used to divide the data sequence according to the timing marks of the operation instructions, so as to obtain multi-source sensor data that can represent the device during the target operation period. For example, the data sequence is divided according to the "start" instruction and the "stop" instruction to obtain multi-source sensor data during the target operation period. Then, the forward-backward algorithm is used to calculate the state transition probability for the data sequence corresponding to the target operation period obtained after the division, and the data sequence corresponding to the target operation period is divided again according to the state transition probability to obtain the target data sequence reflecting the stable operation of the device.

[0036] To better understand the solution given in this example, this example will be further described in combination with a specific application scenario.

[0037] Suppose that after dividing the data sequence according to the timing marks corresponding to the operation instructions "start" and "stop" to obtain the data sequence corresponding to the target operation period, the steps to obtain the target data sequence in the "stable operation" state from the data sequence corresponding to the target operation period are as follows: First, according to the actual operation characteristics of the device, the operation state of the device during the target operation period is modeled as a Markov chain including multiple hidden states to reflect the performance of the device in different operation stages. For example, the operation state of the device is modeled as a Markov chain including three hidden states: "start-up preheating", "stable operation", and "deceleration shutdown". Among them, start-up preheating represents the process of the device gradually accelerating from a stationary state to a stable operation speed. Stable operation represents the state where the device operates normally under rated load and speed. Deceleration shutdown represents the process of the device gradually decelerating from the stable operation state to a stop.

[0038] Next, feature data is extracted from the data sequence corresponding to the target operation period to generate an observation sequence. These features should be able to accurately reflect the operation state of the device. For example, the vibration spectrum centroid is extracted from the vibration sensor data to reflect the distribution of the vibration energy of the device, and the temperature change rate is extracted from the temperature sensor data to reflect the temperature change trend of the device during operation. The feature data can also be other relevant features such as the effective current value, pressure fluctuation, etc., which can be determined specifically according to the sensor type or user requirements, and this application does not make specific restrictions.

[0039] Subsequently, in the data sequence corresponding to the target operation period, a current moment is determined. Starting from the initial moment of the target operation period, the forward probabilities that the device is in each hidden state and observes the current observation sequence are gradually calculated to the current moment by using the transition probability and observation probability of the hidden state. And, starting from the current moment, similarly using the transition probability and observation probability of the hidden state, the backward probabilities that the device is in each hidden state and observes the subsequent observation sequence are gradually calculated to the final moment of the target operation period. Among them, the transition probability of the hidden state represents the probability that the device transfers from one hidden state to another; the observation probability of the hidden state represents the probability of observing a certain observation state under a specific hidden state, which can be set according to experience or through model training based on historical data. Then, according to the forward probabilities and backward probabilities calculated above, the transition probability and observation probability of the hidden state are iteratively optimized by the Baum-Welch algorithm.

[0040] Finally, starting from the initial moment, the maximum probabilities that the device is in each hidden state are calculated to the final moment by using the above forward probabilities and the optimized state transition probability. According to the maximum probabilities, a maximum probability path of the hidden state conversion of the sensor data is obtained, and the most likely state transition sequence during the device operation is obtained. According to this state transition sequence, the start and end time points of each complete operation cycle of the device are identified. For example, starting from the "start preheating" state and ending at the "deceleration shutdown" state, a complete operation cycle is formed. According to the identified start and end time points of the operation cycle, the target data sequence in the "stable operation" state is obtained from the data sequence corresponding to the target operation period.

[0041] Step S15: Determine the health index of the device within the preset time window according to the target data sequence.

[0042] It can be understood that device degradation usually occurs in the stable operation stage rather than the transient process. By dividing the target data sequence, the health index calculation can more focusedly learn the degradation characteristics, avoid the interference of features that may be irrelevant to degradation in the transient process, and improve the accuracy of the health index calculation.

[0043] Optionally, step S15 includes steps S151 to S153: Step S151: Determine the historical sensor data sequences corresponding to each historical failure mode according to historical failure cases.

[0044] Step S152: Determine the weight of the target data sequence according to the similarity between the target data sequence and the historical sensor data sequences.

[0045] It is understandable that different fault modes of the equipment involve different physical degradation processes. For example, bearing wear is mainly caused by surface fatigue and lubrication failure, and early characteristics may be reflected in the increase of specific frequency components in the vibration spectrum. Motor overheating may be caused by winding insulation aging or cooling system failure, and the effective value of current and the rate of temperature change are key indicators. If the same weight is assigned to all sensor data, the health index may misjudge the equipment status due to ignoring the specificity of fault modes. For example, in the early stage of bearing wear, the vibration data is more sensitive than the temperature data. If the weights of the two are the same, the health index may not be able to reflect the wear degree in time. Therefore, the possible fault mode that the equipment may face can be judged through the current target data sequence, and then the weight of the corresponding sensor data can be further increased according to the possible fault mode to improve the fault specificity of the health index and calculate the health index of the equipment more accurately.

[0046] Exemplarily, collect the fault cases that have occurred to the equipment in the past from the equipment maintenance records, fault databases or expert knowledge bases. Each fault case includes the fault type such as bearing wear, gear fracture, motor overheating, etc., the occurrence time, the description of the fault phenomenon, and the sensor data before and after the fault occurrence. Classify and sort out the collected fault cases to establish a fault mode library. Each fault mode corresponds to one or more fault types and is associated with the corresponding sensor data sequence. Extract the historical sensor data sequences corresponding to each fault mode from the fault mode library. Use algorithms such as dynamic time warping, Euclidean distance, and cosine similarity to calculate the similarity between the target data sequence and the historical sensor data sequences corresponding to each historical fault mode. Determine the weight of the target data sequence according to the similarity calculation results.

[0047] Among them, the higher the similarity, the higher the matching degree between the target data sequence and a certain historical fault mode. Therefore, the sensor data corresponding to this fault mode should have a higher weight when calculating the health index. Then, combine the weight of the target data sequence and the sensor data to calculate the health index of the equipment. The calculation method can be weighted average, such as summing the data of each sensor after multiplying it by its weight, or a more complex machine learning model, such as training a health index prediction model using the weight as an input feature.

[0048] To better understand the solution given in this example, the following further explains this example in combination with a specific application scenario.

[0049] Taking the health monitoring of a wind turbine gearbox as an example, it is assumed that a fault mode library containing fault modes such as bearing wear and gear fracture has been established, and the corresponding historical sensor data sequences have been associated. The dynamic time warping algorithm is used to calculate the similarity between the target data sequence, that is, the current operating data of the gearbox, and the historical sensor data sequences corresponding to each historical fault mode. According to the similarity calculation results, the weight of the target data sequence is determined. For example, if the similarity between the target data sequence and the historical sensor data sequence of the bearing wear fault mode is high, a higher weight is assigned to the sensor data related to bearing wear. Combining the weight of the target data sequence and the sensor data, the weighted average method is used to calculate the health index of the gearbox. When the health index is low, it indicates that there may be a fault risk in the gearbox, and a maintenance work order needs to be triggered for generation.

[0050] Optionally, step S152 includes steps S1521 to S1524: Step S1521, obtain the initial weight of the target data sequence.

[0051] Exemplarily, analyze the distribution of sensor data in the historical health state of the device, calculate the variance or entropy value of each sensor data, and use it as the initial weight of each sensor data in the target data sequence. For example, the larger the variance of the vibration sensor data, the more sensitive it is to state changes, and its initial weight is higher. The initial weights of each sensor data can also be manually set by the user.

[0052] Step S1522, determine the target fault mode corresponding to the target data sequence according to the similarity between the target data sequence and the historical sensor data sequence.

[0053] Step S1523, determine the compensation weight of the data corresponding to the fault sensor in the target data sequence according to the fault sensor corresponding to the target fault mode.

[0054] Step S1524, determine the weight of the target data sequence according to the initial weight of the target data sequence and the compensation weight of the data corresponding to the fault sensor in the target data sequence.

[0055] Exemplarily, calculate the dynamic time warping distance between the target data sequence and the historical sensor data sequences corresponding to each historical failure mode to quantify the similarity between the two in terms of time series. Then, after normalizing the target data sequence and the historical sensor data sequences corresponding to each historical failure mode, calculate the cosine similarity between the target data sequence and the historical sensor data sequences corresponding to each historical failure mode to evaluate the similarity between the two in terms of direction. Weightedly sum up the similarities between the two in terms of time series and direction to obtain a similarity score. If the similarity between the target data sequence and any historical failure mode exceeds a preset threshold, then regard this historical failure mode as the target failure mode. Identify the faulty sensors corresponding to the target failure mode in the target data sequence, and set the compensation weights for the sensor data corresponding to the faulty sensors.

[0056] Finally, weightedly sum up the initial weight and the compensation weight to obtain the weight of the target data sequence.

[0057] Step S153, determine the health index according to the target data sequence and the weight.

[0058] After determining the weights corresponding to each sensor data in the target data sequence, weightedly sum up the multi-source sensor data and its weights to calculate the health index of the device.

[0059] Optionally, input the multi-source sensor data into a long short-term memory network model, and through the long short-term memory network model, learn the dependency relationship between the sensor data and the health index according to the historical training data, and obtain the health index within a preset time window output by the long short-term memory network model.

[0060] Step S20, fit a linear decay equation according to the health index within the preset time window, and use the slope of the linear decay equation as the decay rate of the health index.

[0061] It should be noted that the decay rate represents the speed at which the device health index decreases over time, and can be defined as the amount of decrease in the health index per unit time. For example, if the device health index drops from 1.0 to 0.8 in 100 hours, then the decay rate is 0.002 / hour.

[0062] Optionally, determine the decay rate through a linear decay model. Assume that the health index decreases linearly over time, that is , where λ is the decay rate, is the health index at time t, is the health index at the initial moment within the preset time window.

[0063] Optionally, determine the decay rate through an exponential decay model. Assume that the health index decreases exponentially over time, that is .

[0064] Optionally, for a complex attenuation process, models such as polynomial regression, support vector regression, or neural networks can be used to fit the changing trend of the health index.

[0065] Optionally, the attenuation rate is determined by a preset formula:

[0066] where P is a preset time window, is the mean of the health index within the i-th day, is the mean of all health indices, represents the i-th day, represents the mean of t, which is obtained by calculating the arithmetic mean of all t, that is .

[0067] Step S30: Determine the warning threshold of the multi-source sensor according to the attenuation rate.

[0068] Traditional static thresholds cannot reflect the dynamic changes in the health status of the device. For example, when there is early wear of the bearing, the vibration amplitude may be lower than the fixed threshold, but the attenuation rate has increased significantly. When the attenuation rate of the device is high, the static threshold may lag behind the actual fault development, resulting in failure to give an early warning in time, inability to maintain the device in time, and causing a delay in the device maintenance response. By quantifying the degradation speed of the device through the attenuation rate and associating the warning threshold with the degradation speed of the device, the dynamic degradation process of the device can be used to achieve a more accurate early warning.

[0069] Optionally, step S30 includes steps S31 to S33: Step S31: Determine the initial threshold of the multi-source sensor according to the health index.

[0070] Step S32: Determine the threshold adjustment amount of the multi-source sensor according to the attenuation rate.

[0071] Step S33: Determine the warning threshold of the multi-source sensor according to the initial threshold and the threshold adjustment amount.

[0072] Exemplarily, first, according to the preset mapping relationship, the initial threshold of each multi-source sensor is determined according to the health index of the device. Then, the initial threshold is corrected according to the attenuation rate to determine the threshold adjustment amount of each multi-source sensor: . Where is the initial threshold, is the threshold adjustment amount. Then, the sum of the initial threshold and the threshold adjustment amount is used as the warning threshold of the multi-source sensor.

[0073] In this embodiment, by introducing a health index and a decay rate, the warning threshold can be dynamically adjusted according to the actual operating state and degradation of the device, overcoming the limitation that a fixed threshold cannot reflect the dynamic changes of the device in real time.

[0074] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar content as in the above-mentioned first embodiment can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 2 , after step S30, the method for dynamically determining the warning threshold further includes steps S40 to S60: Step S40, using the health index and the target failure mode as decision variables, and using maintenance resources as constraint variables, where the maintenance resources include at least one of the number of device spare parts, the maintenance time window, and the matching degree of maintenance staff.

[0075] Exemplarily, the target failure mode is converted into a one-hot encoding or an embedding vector, and the health index sequence and the failure mode encoding are spliced into a high-dimensional vector as the decision variable.

[0076] Step S50, inputting the decision variables and the constraint variables into a preset work order generation model, generating at least one reference maintenance plan through the preset work order generation model, and calculating the maintenance cost corresponding to the reference maintenance plan.

[0077] Exemplarily, as Figure 3 shown, a multi-objective genetic algorithm such as NSGA-II can be used as the preset work order generation model. Using the multi-objective genetic model to optimize the variables in the maintenance plan according to the decision variables and the constraint variables, and determining the target maintenance plan with a smaller maintenance cost.

[0078] Specifically, the chromosome structure of the multi-objective genetic algorithm can be encoded first as a maintenance plan. For example, chromosome = [maintenance execution time, maintenance operation, spare parts list, maintenance staff]. Randomly generate multiple groups of chromosomes within the range of constraint variables as alternative maintenance plans, and define a maintenance cost function F. Among them, the variables in the chromosome can be determined according to multiple historical maintenance plans corresponding to the target failure mode. Then, perform genetic operations through the multi-objective genetic algorithm to output a non-dominated solution set as the reference maintenance plan. Among them, the genetic operations simulate the processes of natural selection, crossover, and mutation in biological evolution, and are used to generate a new generation of chromosomes, that is, a new maintenance plan, according to the decision variables and the initially randomly generated multiple groups of chromosomes. The non-dominated solution set (Non-dominated Set) refers to a set of solutions in a multi-objective optimization problem for which there is no better solution. In this embodiment, the solution refers to the maintenance plan. Then, according to the defined maintenance cost function F, calculate the maintenance cost corresponding to the non-dominated solution set.

[0079] Step S60: Determine a target maintenance plan in the reference maintenance plan according to the maintenance cost, and generate a maintenance work order according to the target maintenance plan.

[0080] Exemplarily, a maintenance cost threshold can be defined, and the reference plan with a maintenance cost lower than the maintenance cost threshold is determined as the target maintenance plan. Then, a maintenance work order is generated according to the target maintenance plan. The maintenance work order at least includes various variables in the target maintenance plan. The urgency of the maintenance work order can also be confirmed according to the health index of the device. The lower the health index, the higher the urgency of the work order.

[0081] Optionally, when the data collected by the multi-source sensor is greater than the warning threshold and the health index of the device is less than the first threshold, an alarm is triggered and a repair work order is generated. When the data collected by the multi-source sensor is greater than the warning threshold and the health index of the device is greater than the second threshold, a maintenance work order is generated. Wherein, the first threshold is less than the second threshold.

[0082] Based on the above embodiments of the present application, in the third embodiment of the present application, the same or similar content as the above embodiments can be referred to the above introduction and will not be repeated hereinafter. On this basis, please refer to Figure 4 , after step S30, the method for dynamically determining the warning threshold further includes steps S70 to S100: Step S70: Create a health index sequence according to the health index within the preset time window.

[0083] Step S80: Perform packet decomposition on the health index sequence to obtain the low-frequency component corresponding to the health index sequence.

[0084] Exemplarily, the health indices within the preset time window are arranged in ascending order of time stamps to generate an equally spaced time series HI = . Select a wavelet basis according to the signal characteristics, determine the decomposition level according to the length of the health index sequence, and use discrete wavelet transform to decompose the HI sequence to obtain the low-frequency component and high-frequency component corresponding to the HI sequence. Among them, the low-frequency component can represent the long-term trend of device degradation, such as the progressive process of bearing wear; the high-frequency component can represent the sudden noise or fault impact of the sensor.

[0085] Specifically, first perform convolution on the health index sequence using the low-pass filter and high-pass filter corresponding to the wavelet basis. After each layer of decomposition, downsample the low-pass filter coefficients by a factor of 2. Repeat the above process for the remaining low-frequency components after each layer of decomposition until the determined decomposition level is reached.

[0086] Step S90, determine the Hurst exponent corresponding to the low-frequency component, and determine the adjustment amount of the attenuation rate according to the Hurst exponent, where the Hurst exponent represents the likelihood of the change trend of the health index sequence continuing in the future.

[0087] Step S100, use the sum of the attenuation rate and the adjustment amount of the attenuation rate as the new attenuation rate.

[0088] Among them, a low-frequency component will be obtained in each layer of decomposition. The obtained low-frequency component is divided into N sub-intervals with a length of τ. For each sub-interval, calculate the range R of the cumulative deviation and the standard deviation S, and then calculate the rescaled range R / S value of each sub-interval. Then calculate the power-law relationship between the rescaled range R / S value and the time window length to obtain the Hurst exponent H: . Where C is a constant.

[0089] The Hurst exponent is an index used to measure the long-term memory of a time series and can reflect the possibility of the current change trend continuing or reversing in the future. When the Hurst exponent is large, it indicates that the health index sequence has long-term memory, and the current change trend may continue in the future for a period of time, that is, the health index will continue to decay according to the current trend.

[0090] Exemplarily, if the Hurst exponent H > 0.5, it means that the equipment degradation trend may continue, and the attenuation rate is not adjusted, that is = , where represents the adjustment amount of the attenuation rate. If the Hurst exponent H 0.5, it means that the possibility of the equipment degradation trend continuing is small, and a reversal may occur, that is, the current attenuation rate may overestimate the future risk. At this time, the attenuation rate is reduced to avoid premature triggering of maintenance, such as , where is the adjustment coefficient. Finally, use the sum of the attenuation rate and the adjustment amount of the attenuation rate as the new attenuation rate: .

[0091] Through the above steps, combining wavelet decomposition and Hurst exponent analysis, the change trend and characteristics of the equipment health index can be considered more comprehensively, and the calculation accuracy of the attenuation rate can be improved.

[0092] Based on the above embodiments of the present application, in the fourth embodiment of the present application, the same or similar content as in the above Embodiment 1 can be referred to the above introduction and will not be repeated hereinafter. On this basis, after step S60, the method for dynamically determining the warning threshold further includes steps S110~S140: Step S110, obtain the health index and maintenance time-consuming after the equipment is maintained.

[0093] Step S120: Calculate the reward function value corresponding to the target maintenance plan according to the maintained health index and the maintenance time consumed.

[0094] Step S130: If the reward function value is less than the preset reward value, re-determine the attenuation rate of the device according to the maintained health index.

[0095] Step S140: Re-determine the warning threshold of the multi-source sensor according to the re-determined attenuation rate.

[0096] Collect multi-source sensor data within a preset time period after maintenance, and calculate the health index after maintenance according to the multi-source sensor data within the preset time period after maintenance. At the same time, record the maintenance time consumed during the actual maintenance process. Substitute the health index after maintenance and the maintenance time consumed into the defined reward function to calculate the reward function value corresponding to the target maintenance plan.

[0097] Exemplarily, the reward function can be . Wherein, is the reward function value; is the health index after maintenance; is the health index before maintenance; is the maintenance time window in the target maintenance plan; is the maintenance time consumed during the actual maintenance process; is the maintenance cost; 、 and are weight coefficients.

[0098] If the calculated reward function value is less than the preset reward value, it means that the current maintenance plan has not achieved the expected effect, and the possibility of equipment failure is still relatively high. At this time, re-determine the attenuation rate of the equipment according to the maintained health index, and re-determine the warning threshold of the multi-source sensor according to the re-determined attenuation rate, or re-adjust the work order generation rule.

[0099] In this embodiment, the maintenance effect is quantified through the reward function. When the maintenance effect does not meet the expectation, the system automatically triggers parameter re-calculation to adapt to the new degradation state of the equipment and give timely warnings, which can effectively reduce the response delay of equipment maintenance.

[0100] Exemplarily, to help understand the implementation process of the dynamic determination method of the warning threshold obtained by combining this embodiment with the above embodiment, please refer to Figure 5 , Figure 5 which provides a brief flow diagram of the dynamic determination method of the warning threshold. Specifically: First, collect multi-source sensor data within a preset time window. After preprocessing the multi-source sensor data, calculate the health index of the device based on the multi-source sensor data. Then, calculate the attenuation rate of the device according to the health index, and dynamically update the warning threshold of the sensor according to the attenuation rate. When the data collected in real time by the multi-source sensor data exceeds the warning threshold, trigger the health index threshold judgment. When the health index of the device is less than the first threshold, trigger an alarm and generate a maintenance work order. When the health index of the device is greater than the second threshold, generate a maintenance work order. Among them, the first threshold is less than the second threshold.

[0101] It should be noted that the above examples are only for understanding the present application and do not constitute a limitation on the method for dynamically determining the warning threshold of the present application. Based on this technical concept, more forms of simple transformation are within the protection scope of the present application.

[0102] The present application provides a device for dynamically determining a warning threshold. The device for dynamically determining a warning threshold includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method for dynamically determining the warning threshold in the first embodiment above.

[0103] Next, refer to Figure 6 , which shows a schematic structural diagram of a device for dynamically determining a warning threshold suitable for implementing the embodiments of the present application. The device for dynamically determining a warning threshold in the embodiments of the present application may include, but is not limited to, mobile terminals such as laptop computers and tablet computers (PAD, Portable Application Description) and fixed terminals such as desktop computers. Figure 6 The device for dynamically determining a warning threshold shown is only an example and should not bring any limitation to the functions and usage scope of the embodiments of the present application.

[0104] As Figure 6As shown in the figure, the device for dynamically determining the warning threshold may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 1002 or the program loaded from the storage device 1003 into the random access memory (RAM) 1004. In the random access memory 1004, various programs and data required for the operation of the device for dynamically determining the warning threshold are also stored. The processing device 1001, the read-only memory 1002, and the random access memory 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow the device for dynamically determining the warning threshold to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows a device for dynamically determining the warning threshold with various systems, it should be understood that it is not required to implement or have all the shown systems. More or fewer systems may be implemented or had alternatively.

[0105] Specifically, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from the network through the communication device, or installed from the storage device 1003, or installed from the ROM 1002. When the computer program is executed by the processing device 1001, the above functions defined in the methods of the embodiments disclosed in the present application are executed.

[0106] The device for dynamically determining the warning threshold provided by the present application adopts the method for dynamically determining the warning threshold in the above embodiments, and can solve the technical problem of how to improve the timeliness of device maintenance. Compared with the prior art, the beneficial effects of the device for dynamically determining the warning threshold provided by the present application are the same as those of the method for dynamically determining the warning threshold provided by the above embodiments, and other technical features in the device for dynamically determining the warning threshold are the same as the features disclosed in the method of the previous embodiment, and will not be elaborated here.

[0107] It should be understood that each part disclosed in this application can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0108] As described above, the above are only specific embodiments of this application, but the protection scope of this application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed in this application, and all of them should be covered by the protection scope of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

[0109] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., computer programs) stored thereon, and the computer-readable program instructions are used to execute the method for dynamically determining the warning threshold in the above embodiments.

[0110] The computer-readable storage medium provided by this application can be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or flash memory, optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or combined with an instruction execution system, device, or device. The program code contained on the computer-readable storage medium can be transmitted by any suitable medium, including but not limited to: wires, optical cables, radio frequency (RF), etc., or any suitable combination of the above.

[0111] The above computer-readable storage medium can be included in the device for dynamically determining the warning threshold; or it can exist separately without being assembled into the device for dynamically determining the warning threshold.

[0112] The above computer-readable storage medium carries one or more programs. When the one or more programs are executed by a device for dynamically determining an early warning threshold, the device for dynamically determining an early warning threshold can write computer program code for performing the operations of the present application in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, or executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, by using an Internet service provider to connect through the Internet).

[0113] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or operations, or can be implemented by a combination of dedicated hardware and computer instructions.

[0114] The modules involved in the embodiments described in the present application can be implemented in software or in hardware. Among them, the name of the module does not constitute a limitation to the unit itself in some cases.

[0115] The readable storage medium provided by this application is a computer-readable storage medium. The computer-readable storage medium stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned method for dynamically determining warning thresholds, and can solve the technical problem of how to improve the timeliness of equipment maintenance. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided by this application are the same as those of the method for dynamically determining warning thresholds provided by the above embodiments, and will not be elaborated here.

[0116] The above are only partial embodiments of this application, and thus do not limit the patent scope of this application. Any equivalent structural transformation made under the technical concept of this application by using the content of the specification and drawings of this application, or any direct / indirect application in other related technical fields, is included in the patent protection scope of this application.

Claims

1. A method for dynamically determining an early warning threshold, characterized in that The method for dynamically determining the warning threshold includes: Determine the health index of the device within the preset time window according to the multi-source sensor data within the preset time window; Fit a linear decay equation according to the health index within the preset time window, and use the slope of the linear decay equation as the decay rate of the health index; Determine the warning threshold of the multi-source sensor according to the decay rate.

2. The method for dynamically determining the warning threshold according to claim 1, wherein The step of determining the health index of the device within the preset time window according to the multi-source sensor data within the preset time window includes: Obtain the operation instructions received by the device within the preset time window; Use the time stamp corresponding to the operation instruction as a timing mark, and map it to the multi-source sensor data to obtain a data sequence with timing marks; Determine the target operation period of the device according to the timing mark corresponding to the operation instruction, and calculate the state transition probability according to the data sequence corresponding to the target operation period through the forward-backward algorithm; Segment the data sequence corresponding to the target operation period according to the state transition probability to obtain a target data sequence; Determine the health index of the device within the preset time window according to the target data sequence.

3. The method for dynamically determining the warning threshold according to claim 2, wherein The step of determining the health index of the device within the preset time window according to the target data sequence includes: Determine the historical sensor data sequences corresponding to each historical failure mode according to historical failure cases; Determine the weight of the target data sequence according to the similarity between the target data sequence and the historical sensor data sequences; Determine the health index according to the target data sequence and the weight.

4. The method for dynamically determining the warning threshold according to claim 3, wherein, The step of determining the weight of the target data sequence according to the similarity between the target data sequence and the historical sensor data sequences includes: Obtain the initial weight of the target data sequence; Determine the target failure mode corresponding to the target data sequence according to the similarity between the target data sequence and the historical sensor data sequences; Determine the compensation weight of the data corresponding to the failure sensor in the target data sequence according to the failure sensor corresponding to the target failure mode; Determine the weight of the target data sequence according to the initial weight of the target data sequence and the compensation weight of the data corresponding to the failure sensor in the target data sequence.

5. The method for dynamically determining the warning threshold according to claim 1, characterized in that After the step of fitting a linear decay equation according to the health index within the preset time window and using the slope of the linear decay equation as the decay rate of the health index, it further includes: Create a health index sequence according to the health index within the preset time window; Perform packet decomposition on the health index sequence to obtain the low-frequency component corresponding to the health index sequence; Determine the Hurst index corresponding to the low-frequency component, and determine the adjustment amount of the decay rate according to the Hurst index, where the Hurst index represents the likelihood that the change trend of the health index sequence will continue in the future; Use the sum of the decay rate and the adjustment amount of the decay rate as the new decay rate.

6. The dynamic determination method of the warning threshold according to any one of claims 1-5, characterized in that, The step of determining the warning threshold of the multi-source sensor according to the decay rate includes: Determine an initial threshold of the multi-source sensor according to the health index; Determine an adjustment amount of the threshold of the multi-source sensor according to the attenuation rate; Determine an early warning threshold of the multi-source sensor according to the initial threshold and the adjustment amount of the threshold.

7. The method for dynamically determining the warning threshold according to claim 4, wherein, After the step of determining the early warning threshold of the multi-source sensor according to the attenuation rate, the method further includes: Use the health index and the target fault mode as decision variables, and use maintenance resources as constraint variables, where the maintenance resources include at least one of the number of equipment spare parts, the maintenance time window, and the matching degree of maintenance staff; Input the decision variables and the constraint variables into a preset work order generation model, generate at least one reference maintenance plan through the preset work order generation model, and calculate the maintenance cost corresponding to the reference maintenance plan; Determine a target maintenance plan from the reference maintenance plans according to the maintenance cost, and generate a maintenance work order according to the target maintenance plan.

8. The method for dynamically determining the warning threshold according to claim 7, wherein, After the step of determining a target maintenance plan from the reference maintenance plans according to the maintenance cost and generating a maintenance work order according to the target maintenance plan, it further includes: Obtain the health index and the maintenance time consumption after the equipment maintenance; Calculate a reward function value corresponding to the target maintenance plan according to the health index after the maintenance and the maintenance time consumption; If the reward function value is less than a preset reward value, re-determine the attenuation rate of the equipment according to the health index after the maintenance; Re-determine the early warning threshold of the multi-source sensor according to the re-determined attenuation rate.

9. A device for dynamically determining an early warning threshold, characterized in that, The device includes: a memory, a processor, and a computer program stored on the memory and executable on the processor, and the computer program is configured to implement the steps of the method for dynamically determining the early warning threshold according to any one of claims 1 to 8.

10. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium, and when the computer program is executed by a processor, it implements the steps of the method for dynamically determining the early warning threshold according to any one of claims 1 to 8.

Images