Equipment health degree assessment method based on multi-parameter adaptive weight

通过多参数自适应权重的设备健康度评估方法,利用K-Means和Softmax算法动态计算参数权重,解决了现有技术中依赖人工设定导致的误差问题,实现了更准确和灵活的设备健康度 Evaluate.

CN119004148APending Publication Date: 2024-11-22JIANGSU RUIDIAN ZHIXIN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411016985.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-26
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The prior art relies on manual parameter weights when evaluating the health of the equipment, which has large errors and cannot dynamically adapt to parameter changes during the operation of the equipment, resulting in inaccurate health assessment.

Method used

The equipment health evaluation method is adopted with multi-parameter adaptive weights. The weight of the health evaluation parameters is dynamically calculated through the K-Means algorithm and the Softmax algorithm. It is based on data self-learning, and does not rely on expert experience, and dynamic adjustment is supported.

Benefits of technology

It improves the accuracy and adaptability of equipment health evaluation, can continuously optimize during the operation of the equipment, dynamically adapt to parameter weight changes, and reduce human errors.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004148A_ABST
    Figure CN119004148A_ABST
Patent Text Reader

Abstract

The invention discloses an equipment health degree evaluation method based on a multi-parameter adaptive weight, and the method comprises the steps: controlling the operation state of tested equipment, at least obtaining a plurality of test data of two groups of tested equipment in different operation states, obtaining two groups of operation state data, and respectively taking a measurement time sequence as a horizontal axis, taking the health degree evaluation parameter values as vertical coordinates, and respectively drawing a curve graph of each health degree evaluation parameter; calculating a distance sequence of each health degree evaluation parameter to obtain a distance array; classifying all data in the distance array by adopting a K-Means algorithm to obtain a classification result; performing data processing on the classification result to obtain an updated data sequence; calculating the updated data sequence by adopting a Softmax algorithm to obtain an adaptive weight of each health degree evaluation parameter; and the health degree of the tested equipment is calculated through the adaptive weight of each health degree evaluation parameter, so that a health degree evaluation result is obtained.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of low voltage power equipment application, and in particular to an equipment health evaluation method based on multi-parameter adaptive weights. Background Art

[0002] In low-voltage power systems, as the use time of electrical equipment increases and the external operating environment changes, the equipment continues to wear out and the health of the equipment slowly decreases, accompanied by the deterioration of various external parameters.

[0003] Currently, when evaluating the health of equipment, some parameters of the equipment are collected, and most of the evaluation coefficient weights need to be set for different parameters based on experience, and the current health of the equipment is judged based on the weights of each parameter. This method uses manual setting, which requires high personal experience, and there may be a large deviation between the artificial setting value and the actual value, resulting in a large error in the evaluation of equipment health.

[0004] Various parameters change dynamically during the operation of the equipment. Different parameters at different stages reflect the health of the equipment to different degrees. Currently, fixed values ​​are used through manual settings, which can reflect the health of the equipment within a certain period of time. If this time range is exceeded, the deviation between the set value and the actual situation becomes increasingly larger. Summary of the invention

[0005] Technical purpose: In view of the defects in the prior art, the present invention discloses a device health assessment method based on multi-parameter adaptive weights.

[0006] Technical solution: In order to achieve the above technical objectives, the present invention adopts the following technical solution.

[0007] A device health assessment method based on multi-parameter adaptive weights includes the following steps:

[0008] S1. Control the operating state of the device under test, where the operating state includes operating conditions and operating inputs. Taking an AC motor as an example, the operating conditions include the external working environment and the load size, and the operating input includes the input voltage. At least two groups of test data of the device under test are obtained for several times under different operating states to obtain two groups of operating state data, each group of operating state data includes several health evaluation parameter sequences of the device under test;

[0009] S2. For the two groups of operating status data, a curve graph of each health evaluation parameter is drawn with the measurement time sequence as the horizontal axis and the health evaluation parameter value as the vertical axis;

[0010] S3, calculating the distance sequence of each health evaluation parameter according to the two obtained curve graphs to obtain a distance array;

[0011] S4. Use K-Means algorithm to classify all data in the distance array to obtain classification results, and the initialized K value is twice the number of health evaluation parameters; perform data processing on the classification results to obtain an updated data sequence;

[0012] S5, using the Softmax algorithm to calculate the updated data sequence to obtain the adaptive weight of each health evaluation parameter;

[0013] S6. The health of the device under test is obtained by adaptively calculating the weight of each health evaluation parameter, thereby obtaining a health evaluation result.

[0014] Preferably, in S4, data processing is performed on the classification result to obtain an updated distance array, including:

[0015] In the classification results, the valid results of each health evaluation parameter are obtained according to the validity of the results. The valid results corresponding to each health evaluation parameter include the data of its own health evaluation parameter and the data of other health evaluation parameters;

[0016] The data of other health evaluation parameters are removed from the valid results corresponding to each health evaluation parameter, the data of its own health evaluation parameter is retained, the mean is recalculated and the absolute value is taken to obtain an updated distance array.

[0017] Preferably, the calculation formula of the adaptive weight of each health evaluation parameter in S5 is:

[0018]

[0019] Where d′=[d′ 1 , d′ 2 ,...,d′ m ], is the updated data sequence, d′ 1 , d′ 2 , ..., d′ m are the means of m health evaluation parameters respectively; 1≤k≤m, p(k) is the weight corresponding to the kth health evaluation parameter; m is the number of health evaluation parameters.

[0020] Preferably, the calculation formula for the health of the device under test in S6 is:

[0021]

[0022] Among them, p(k) is the weight corresponding to the kth health evaluation parameter, h(k) is the health value of the kth health evaluation parameter itself, 1≤k≤m, and m is the number of health evaluation parameters.

[0023] Beneficial effects: The present invention collects data for self-learning, does not rely on expert experience, and supports dynamic adjustment capabilities, thereby improving the accuracy and adaptability of equipment health evaluation; it continuously optimizes during equipment operation and dynamically adapts to changes in the weights of different parameters during equipment operation; BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart of the method of the present invention;

[0025] Figure 2 Schematic diagrams of two curves of the present invention;

[0026] Figure 3 This is a schematic diagram of the K-Means result of the present invention with twice the number of center points;

[0027] Figure 4 This is a schematic diagram of a cluster to which samples of other indicators belong, obtained by calculating a certain indicator through K-Means in the present invention;

[0028] Figure 5 for Figure 4 Schematic diagram of the clustering results of a health evaluation parameter after removing other health evaluation parameter sample data. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0030] As attached Figure 1 As shown, a device health assessment method based on multi-parameter adaptive weights of the present invention comprises the following steps:

[0031] Step 1: The running input and running state of the device under test (hereinafter referred to as the device) remain unchanged, and n tests are performed to obtain the value of each health evaluation parameter of the device, and multiple sets of data for each parameter are obtained respectively; the sequence of health evaluation parameters is as follows:

[0032]

[0033] Among them, θ 1 ,θ 2 ...、θ m is the m health evaluation parameter sequence of the equipment, θ 1-1 ,θ 1-2 , ..., θ 1-nis the parameter value of the first health evaluation parameter in the overall health evaluation parameters of the equipment's operating status, measured by n tests;

[0034] Step 2, with the measurement time sequence as the horizontal axis and the health evaluation parameter value as the vertical axis, draw a curve graph of each health evaluation parameter;

[0035] Step 3: Change the operating conditions of the equipment under test, such as external environment temperature, load size, etc., repeat the above steps, and re-obtain the sequence corresponding to each health evaluation parameter;

[0036]

[0037] Among them, θ′ 1-1 ,θ′ 1-2 , ..., θ′ 1-n is the parameter value of the first health evaluation parameter for n tests, θ′ 1 ,θ′ 2 ...、θ′ m is a sequence of m health evaluation parameters of the equipment;

[0038] Step 4: Draw a health evaluation parameter curve chart based on the parameter values ​​obtained by re-measurement, with the measurement time sequence as the horizontal axis and the health evaluation parameter value as the vertical axis;

[0039] Step 5: For each health evaluation parameter, calculate the distance sequence between the two curves obtained in step 2 and step 4, such as Figure 2 As shown;

[0040]

[0041] Among them, d 1-1 ,d 1-2 , ..., d 1-n is the corresponding distance between the n points of the two curves of the first health evaluation parameter, d 1 ,d 2 ...、d m is the distance sequence corresponding to the m health evaluation parameters of the equipment;

[0042] Step 6: Merge the curve distances obtained by measuring the curves of multiple health evaluation parameters to form a one-dimensional array d:

[0043] d=[d 1 , d 2 , ..., d m ]

[0044] Step 7: Perform K-Means calculation on the one-dimensional array, where K is twice the category of the health evaluation parameter, to separate the positive correlation and negative correlation, such as Figure 3 As shown;

[0045] Step 8. Calculate the results based on K-Means calculations, and calculate the category to which each health evaluation parameter belongs according to the proportion of quantity. Among the 2K results of the classification, obtain the valid results corresponding to m health evaluation parameters according to the validity of the results, such as: retain the cluster center to which each health evaluation parameter belongs, and select the cluster center containing the most corresponding health evaluation parameters among the 2K results as the cluster center of the health evaluation parameter. The corresponding results of each health evaluation parameter include both the sample data related to the health evaluation parameter and the sample data of other health evaluation parameters, such as Figure 4 As shown;

[0046] Step 9: Get the list of all samples of the health evaluation parameter in the cluster classification result of each health evaluation parameter in the K-Means result, remove the corresponding data of non-health evaluation parameters, and reduce the influence of other health evaluation parameters on the health evaluation parameter, such as Figure 5 As shown;

[0047] Step 10: Recalculate the corresponding mean using the own sample list;

[0048] Step 11: judge the recalculated mean, if the result is negative, take the absolute value of the result;

[0049] Step 12: The mean value list calculated for each health evaluation parameter forms a data sequence d of m data:

[0050] d′=[d′ 1 , d′ 2 , ..., d′ m ]

[0051] Among them, d′ 1 , d′ 2 , ..., d′ m are the means of m health evaluation parameters respectively;

[0052] Step 13: Softmax is used to calculate the weight of each health evaluation parameter using the inverse of the curve distance of each health evaluation parameter; the calculation formula is:

[0053]

[0054] Where, 1≤k≤m, p(k) is the weight corresponding to the kth health evaluation parameter; m is the number of health evaluation parameters;

[0055] Step 14: Use the Softmax result as the corresponding parameter evaluation health weight coefficient, and use the sign of the K-Means calculation result as the sign of the weight to ensure that the correlation between each health evaluation parameter and the detected device is consistent.

[0056] Step 1 5. After calculating the weight corresponding to each parameter, when calculating the overall health of the device, calculate based on the weight corresponding to each parameter:

[0057]

[0058] Among them, H is the overall health of the equipment, with a value range of [0, 100]. The larger the value, the higher the health. p(k) is the weight corresponding to the kth health evaluation parameter, and h(k) is the health value of the kth health evaluation parameter itself. The health value of each parameter itself is independently defined in advance and is a preset parameter value. Taking temperature data as an example, a temperature of 30 degrees is excellent, and its own health value is 0.95. 50 degrees is good, and its own health value is 0.9.

[0059] In the prior art, the health of equipment is mainly evaluated by experts using multiple dimensions of equipment data based on their experience. 1) This setting relies on expert experience; 2) There are data differences between different instances of the same type of equipment, and experts cannot distinguish them one by one; 3) The weights of the parameters of multiple dimensions of equipment for evaluating the health of the equipment should change dynamically as the equipment runs, and manual settings cannot dynamically adapt. The equipment health assessment method based on multi-parameter adaptive weights proposed in the present invention self-learns through collected data, does not rely on expert experience, and supports dynamic adjustment capabilities, thereby improving the accuracy and adaptability of equipment health evaluation. For the data collected from each device, the corresponding parameter weights of different devices are calculated separately, so that 1 device has 1 parameter; continuous optimization is carried out during the operation of the equipment to dynamically adapt to changes in the weights of different parameters during equipment operation;

[0060] Example:

[0061] In this embodiment, an AC motor is taken as an example. The motor speed v, working current i, working temperature t, and vibration intensity a are collected each time. The data are collected n times in total to obtain vectors corresponding to different indicators:

[0062]

[0063] Change the motor operating environment temperature and the load size of the motor, re-collect the motor speed v, working current i, working temperature t, and vibration intensity a, and get a new data list:

[0064]

[0065] Based on the data obtained from the above measurements, the distance between the two sets of data curves for each parameter is calculated:

[0066]

[0067] Where: d v ={d v-1 d v-2 …d v-n} represents the distance sequence between n groups of data of two curves of speed parameters, and the others are similar; d v-1 =v′ 1-1 -v 1-1 It represents the distance between the first set of corresponding data of the first curve and the second curve of the speed curve, and the others are similar;

[0068] The speed, current, temperature, vibration and other parameters are calculated to obtain the curve distance and merged into a one-dimensional array d = [d v , d i , d t , d a ], each parameter has n data, a total of 4*n data;

[0069] Cluster 4*n data using K-means, with the number of categories K set to 8;

[0070] After clustering is completed, select the cluster centers of speed, current, temperature, and vibration respectively, and select the cluster center containing the most corresponding parameters as the cluster center of the parameter. For example, the corresponding cluster center in the clustering result of the speed parameter contains the following data:

[0071] k_v={d v1 , d v2 , ..., d vi , d tk , ..., d tj}

[0072] Among them: kv is the final cluster center of the speed parameter, which contains several speed distance data and is the cluster center with the largest number of speed corresponding distances among the 2K cluster centers. This cluster contains a set of speed and other parameter corresponding data;

[0073] Remove the data corresponding to the speed corresponding cluster center and the non-speed parameter:

[0074] d′ v ={d v-1 d v-2 …d v-i}

[0075] Processing current, temperature, and vibration parameters in the same way, we get

[0076] d′=[d′ v , d′ i , d′t , d′ a ]

[0077] Based on the latest speed, current, temperature, and vibration parameters, softmax is used to calculate the weights:

[0078]

[0079] Calculate the corresponding weights of speed, current, temperature, and vibration parameters, and then calculate the overall health of the motor equipment based on these parameters:

[0080]

[0081] H is the overall health of the device, with a value range of [0, 100]. A larger value represents a higher health.

[0082] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A device health assessment method based on multi-parameter adaptive weights, characterized in that: The following steps are involved: S1. Control the operating state of the device under test, obtain at least two groups of test data of the device under test in different operating states, and obtain two groups of operating state data, each group of operating state data includes several health evaluation parameter sequences of the device under test; S2. For the two groups of operating status data, a curve graph of each health evaluation parameter is drawn with the measurement time sequence as the horizontal axis and the health evaluation parameter value as the vertical axis; S3, calculating the distance sequence of each health evaluation parameter according to the two obtained curve graphs to obtain a distance array; S4. Use K-Means algorithm to classify all data in the distance array to obtain classification results, and the initialized K value is twice the number of health evaluation parameters; perform data processing on the classification results to obtain an updated data sequence; S5, using the Softmax algorithm to calculate the updated data sequence to obtain the adaptive weight of each health evaluation parameter; S6. The health of the device under test is obtained by adaptively calculating the weight of each health evaluation parameter, thereby obtaining a health evaluation result.

2. According to claim 1, a method for evaluating equipment health based on multi-parameter adaptive weights is characterized in that: In S4, the classification results are processed to obtain an updated data sequence, including: In the classification results, the valid results of each health evaluation parameter are obtained according to the validity of the results. The valid results corresponding to each health evaluation parameter include the data of its own health evaluation parameter and the data of other health evaluation parameters; The data of other health evaluation parameters are removed from the valid results corresponding to each health evaluation parameter, the data of its own health evaluation parameter is retained, the mean is recalculated and the absolute value is taken to obtain the updated data sequence.

3. The device health assessment method based on multi-parameter adaptive weights according to claim 2 is characterized in that: In the classification results, the valid results of each health evaluation parameter are obtained according to the validity of the results, including: retaining the cluster center to which each health evaluation parameter belongs, and selecting the cluster center containing the most corresponding health evaluation parameters among the 2K results as the cluster center of the health evaluation parameter.

4. The device health assessment method based on multi-parameter adaptive weights according to claim 1 is characterized in that: The calculation formula of the adaptive weight of each health evaluation parameter in S5 is: Where, d′=[d′1,d′2,...,d′ m ], is the updated data sequence, d′1, d′2, ..., d′ m are the means of m health evaluation parameters respectively; 1≤k≤m, p(k) is the weight corresponding to the kth health evaluation parameter; m is the number of health evaluation parameters.

5. The device health assessment method based on multi-parameter adaptive weights according to claim 1 is characterized in that: The calculation formula for the health of the device under test in S6 is: Among them, p(k) is the weight corresponding to the kth health evaluation parameter, h(k) is the health value of the kth health evaluation parameter itself, 1≤k≤m, and m is the number of health evaluation parameters.