A method and device for health monitoring and condition assessment of mechanical equipment with adaptive health threshold

Abstract

The present invention belongs to the technical field related to rotating machinery performance evaluation, and discloses a method and device for health monitoring and state evaluation of mechanical equipment with an adaptive health threshold, comprising the following steps: (1) using the Hilbert singular value decomposition algorithm to convert a vibration signal representing the operating state of the mechanical equipment into a singular value feature sequence; inputting the singular value feature sequence into a nonlinear state estimation model, and the nonlinear state estimation model outputting a health indicator for nonlinear state estimation reconstruction error; (2) selecting a health indicator in a normal stage as a health threshold benchmark, and introducing a peak over-threshold algorithm to dynamically update the health threshold; (3) using a K-means clustering algorithm based on the obtained health indicator to divide the operating stage of the mechanical equipment, and then evaluating the health state of the mechanical equipment based on a logic correction algorithm. The present invention improves the sensitivity and versatility of the health indicator.

Description

Technical Field

[0001] The present invention belongs to the technical field related to rotating machinery performance evaluation, and more specifically, relates to a method and device for health monitoring and status evaluation of mechanical equipment with an adaptive health threshold. Background Art

[0002] Equipment plays an increasingly important role in the operation of modern industrial systems, and prognostic and health management have become a research trend in academia and industry. Condition monitoring, as one of the main tasks of prognostic and health management, tracks equipment operating status by building health indicators. This allows for early detection of the transition from normal operation to incipient failures, triggering subsequent remaining useful life prediction. This is crucial for monitoring the safe operation of equipment and reducing major safety incidents in real-world applications.

[0003] With the increasing intelligence and rapid development of sensor technology, characteristic signals collected from sensors contain information reflecting equipment degradation. Consequently, numerous data-driven methods have emerged for assessing and predicting degradation trends in rotating machinery. On the one hand, health indicators can identify the onset of early faults, facilitating real-time monitoring of the rotating machinery's future operating status and performance degradation trends. Furthermore, early fault points also serve as a starting point for prediction. However, in most performance degradation assessment methods, the constructed health assessment indicators, such as traditional health indicators like root mean square (RMS) and kurtosis, are insensitive to early degradation information and cannot transiently detect early faults. These indicators are not universally applicable to most rotating machinery components. Health thresholds are typically set at fixed values, resulting in poor adaptability. Therefore, the problem of incipient fault detection remains to be addressed. On the other hand, health status assessments can provide more targeted monitoring of equipment components, which is crucial for guiding maintenance and replacement, ensuring safe operation. Summary of the Invention

[0004] In response to the above defects or improvement needs of the prior art, the present invention provides a method and device for mechanical equipment health monitoring and status assessment with adaptive health thresholds, which aims to solve the problems of low sensitivity and weak versatility of existing mechanical equipment health indicators.

[0005] To achieve the above objectives, according to one aspect of the present invention, a method for health monitoring and status assessment of mechanical equipment with an adaptive health threshold is provided, the method comprising the following steps:

[0006] Step 1: The vibration signal representing the operating state of the mechanical equipment is converted into a singular value feature sequence using the Hilbert singular value decomposition algorithm; the singular value feature sequence is input into a nonlinear state estimation model, and the nonlinear state estimation model outputs a health indicator for the nonlinear state estimation reconstruction error;

[0007] Step 2: Select the health indicator in the normal stage as the health threshold benchmark, and introduce the peak over-threshold algorithm to dynamically update the health threshold;

[0008] Step 3: Based on the obtained health indicators, the K-means clustering algorithm is used to divide the operation stages of the mechanical equipment, and then the health status of the mechanical equipment is evaluated based on the logic correction algorithm.

[0009] Furthermore, based on the Hankel matrix constructed by the Hilbert singular value decomposition algorithm, the original signal X is decomposed into a singular value sequence S=diag(s1, s2, ..., s q ), select the singular value to perform singular value decomposition and reconstruction.

[0010] Furthermore, let the input process memory matrix of the nonlinear state estimation model be S, r be the number of input singular values, and m be the number of monitoring samples, then:

[0011]

[0012] The input of the nonlinear state estimation model is the future test data S of the mechanical equipment obs , the output is for S obs The predicted value S est .

[0013] Furthermore, for any input S obs , the nonlinear state estimation model generates an r-dimensional weight vector W = [w1 w2 … w n ] T , such that:

[0014] S est =SW=w1S(1)+w2S(2)+…+w r S(r) (2)

[0015] The predicted value X of the nonlinear state estimation model est is a linear combination of r historical observation vectors in S; the weight vector W is determined by the following method: construct the measured data S of the nonlinear state estimation model obs and the predicted value S est The residual between is:

[0016] ε=S obs -S est (3)

[0017] Choose W to minimize the residual sum of squares, the residual sum of squares is:

[0018]

[0019] Using partial derivative to solve equation (4) we can get the weight vector W = (ST ·S) T ·(S T ·S obs );Get the predicted value S nest :

[0020]

[0021] The process memory matrix S represents the entire dynamic process of normal operation of the mechanical equipment; the residual ε=S obs -S nest Health indicators are constructed to monitor the operating status of mechanical equipment. When the working status of mechanical equipment changes, the residual will increase.

[0022] Furthermore, let μ be the initial threshold, and the samples exceeding μ are recorded as N μ is the number of samples, the excess variable is recorded as y = ε - μ, and the corresponding excess distribution function is expressed as:

[0023]

[0024] F μ (y) seems to be a generalized Pareto distribution, that is:

[0025]

[0026] Where ξ is the scale parameter, β is the shape parameter; the estimated parameters and By using the maximum likelihood estimation method, we can obtain:

[0027]

[0028]

[0029] When y>0, formula (6) is rewritten as:

[0030]

[0031] According to the health index values ​​of the mechanical equipment in the early operation stage, the values ​​of 95% of the sequence are sorted from small to large and the values ​​of μ are taken as μ. In formula (10), F(μ) is given by Determine, n is the total number of health indicators.

[0032] Furthermore, the tail of F(ε) is estimated as:

[0033]

[0034] Under the premise that the tail probability q is 0.05, the adaptive threshold ε is obtained q :

[0035]

[0036] The present invention also provides a mechanical equipment health monitoring and status assessment system with adaptive health thresholds. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the mechanical equipment health monitoring and status assessment method with adaptive health thresholds as described above.

[0037] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the mechanical equipment health monitoring and status assessment method with adaptive health threshold as described above.

[0038] In general, compared with the prior art, the above technical solutions conceived by the present invention provide a method and device for health monitoring and status assessment of mechanical equipment with adaptive health thresholds, which has the following beneficial effects:

[0039] 1. The Hilbert singular value decomposition algorithm is used to convert the vibration signal representing the operating state of mechanical equipment into a singular value feature sequence. The singular value feature sequence is input into a nonlinear state estimation model, which outputs a health indicator for nonlinear state estimation reconstruction error. The obtained health indicator amplifies the difference between the initial fault sample and the normal sample, improving the sensitivity and versatility of the health indicator.

[0040] 2. The present invention makes full use of the advantage of singular value eigenvalues ​​in characterizing the degradation process. Useful information is reflected in the first few singular value sequences that are ranked high. Using them as input features of the nonlinear state estimation model can effectively solve the tedious problems of important information redundancy and feature screening.

[0041] 3. Constructing the health indicator of the nonlinear state estimation model amplifies the difference between the initial fault samples and normal samples; introducing the peak over-threshold algorithm to dynamically update the health threshold, realizing adaptive detection of initial faults.

[0042] 4. Based on the K-means clustering algorithm, the cluster centers of health indicators are calculated, and the logical correction algorithm is used to evaluate the health and sub-health of equipment and monitor the operating status, realizing the tracking and monitoring of degradation status. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is a flow chart of a method for health monitoring and status assessment of mechanical equipment with an adaptive health threshold provided by the present invention;

[0044] Figure 2 is a graph of IMS bearing full life degradation data collected by an embodiment of the present invention;

[0045] Figure 3 This is a graph of the full-life degradation data of an XJTU bearing collected in an embodiment of the present invention, where (a) and (b) correspond to XB3-2 and XB1-5, respectively;

[0046] Figure 4 is a diagram of the degradation process of IB1-4 represented by the health indicator proposed in the present invention, wherein (a) and (b) correspond to the degradation process of IB1-4 and the schematic diagram of the degradation process of IB1-4, respectively;

[0047] Figure 5 is a comparative method to characterize the degradation process of IB1-4, where (a) and (b) correspond to RMS and Kurtosis, respectively;

[0048] Figure 6 1 is a diagram of the degradation process of XB3-2 characterized by the health indicators proposed in an embodiment of the present invention, wherein (a) and (b) correspond to the degradation process diagram of XB3-2 and the enlarged diagram of the degradation process of XB3-2, respectively;

[0049] Figure 7 1 is a diagram showing the degradation process of XB1-5 using the health indicators proposed in an embodiment of the present invention, wherein (a) and (b) correspond to the degradation process diagram of XB1-5 and an enlarged diagram of the degradation process, respectively;

[0050] Figure 8 Schematic diagram of the XB3-2 status evaluation result of the K-means clustering logic correction algorithm according to an embodiment of the present invention;

[0051] Figure 9 Schematic diagram of the XB1-5 status evaluation result of the K-means clustering logic correction algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for the purpose of explaining the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

[0053] The present invention provides a method for health monitoring and status assessment of mechanical equipment with an adaptive health threshold, the method mainly comprising the following steps:

[0054] Step 1: The vibration signal representing the operating state of the mechanical equipment is converted into a singular value feature sequence using the Hilbert singular value decomposition algorithm; the singular value feature sequence is input into a nonlinear state estimation model, and the nonlinear state estimation model outputs a health indicator for the nonlinear state estimation reconstruction error.

[0055] Specifically, a health indicator is created based on a nonlinear state estimation model. Vibration signals representing the operating state of mechanical equipment are collected as raw features. Since the singular value feature sequence of singular value decomposition has the advantage of representing degradation processes, the Hilbert singular value decomposition algorithm is used to extract the high-dimensional feature sequence of the vibration signal into a singular value feature sequence. A nonlinear state estimation model is built, and the extracted singular value feature sequence is input to construct a health indicator for the nonlinear state estimation reconstruction error. This health indicator amplifies the difference between the initial faulty samples and normal samples, addressing the undersensitivity of existing health indicators.

[0056] In one embodiment, a health index is created based on a nonlinear state estimation model. The vibration signal is used as the original feature, and a singular value feature sequence is constructed based on the Hilbert singular value decomposition algorithm as the input of the nonlinear state estimation model to create a health index. The Hankel matrix constructed based on the Hilbert singular value decomposition algorithm can decompose the original signal X into a singular value sequence S = diag (s1, s2, ..., s q ), selecting appropriate singular values ​​for singular value decomposition and reconstruction can eliminate noise in the original signal. Related research shows that useful information is reflected in the top-ranked singular value sequences. Selecting the top n singular value sequences as input features for the nonlinear state estimation model based on the singular value distribution trend of the specific monitoring object can effectively solve the tedious problems of important information redundancy and feature screening. Let the input process memory matrix of the nonlinear state estimation model be S, r be the number of input singular values, and m be the number of monitoring samples, then:

[0057]

[0058] The input of the nonlinear state estimation model is the future test data S of the mechanical equipment obs , the output is for S obs The predicted value S est For any input S obs , the nonlinear state estimation model generates an r-dimensional weight vector W = [w1 w2 …w n ] T , such that:

[0059] S est =SW=w1S(1)+w2S(2)+…+w r S(r) (2)

[0060] The predicted value X of the nonlinear state estimation model est is a linear combination of r historical observation vectors in S. The weight vector W is determined by the following method: Construct the measured data S of the nonlinear state estimation model obs and the predicted value S estThe residual between is:

[0061] ε=S obs -S est (3)

[0062] Choose W to minimize the residual sum of squares, the residual sum of squares is:

[0063]

[0064] Using partial derivative to solve equation (4) we can get the weight vector W = (S T ·S) T ·(S T ·S obs ). Furthermore, the Euclidean distance is used to replace the dot multiplication operation to make the nonlinear state estimation model have a more intuitive physical meaning, thereby obtaining the predicted value S nest :

[0065]

[0066] The process memory matrix S represents the entire dynamic process of the normal operation of the mechanical equipment. When the future test data S obs Similar to some historical observation vectors in S, that is, when the mechanical equipment is always in normal operation, the predicted value S nest Therefore, the residual ε=S obs -S nest Health indicators are constructed to monitor the operating status of mechanical equipment. When the working status of mechanical equipment changes, the residual will increase.

[0067] Step 2: Select the health indicators in the normal stage as the health threshold benchmark, and introduce the peak over-threshold algorithm to dynamically update the health threshold.

[0068] An adaptive health threshold is constructed based on the peak threshold exceeding algorithm. The peak threshold exceeding algorithm is incorporated into the health indicator. The health indicator during the normal phase is selected as the health threshold benchmark. This is then used as the basis for the peak threshold exceeding algorithm to dynamically update the health threshold, resolving the issue of fixed health threshold settings. Based on the created health indicator, initial fault detection is performed under the adaptive health threshold.

[0069] An adaptive health threshold is constructed based on the peak threshold exceeding algorithm. The peak threshold exceeding algorithm is introduced into the health index, and the health index in the normal stage is selected as the health threshold benchmark. The peak threshold exceeding algorithm is then used to dynamically update the health threshold.

[0070] According to the peak over-threshold algorithm principle, let μ be a sufficiently large initial threshold, and the samples exceeding μ are recorded as N μ is the number of samples, the excess variable is recorded as y = ε - μ, and the corresponding excess distribution function is expressed as:

[0071]

[0072] F μ (y) can be approximated as a generalized Pareto distribution, that is:

[0073]

[0074] Where ξ is the scale parameter and β is the shape parameter. Estimated parameters and Obtained by maximum likelihood estimation method.

[0075]

[0076] When y>0, formula (6) is rewritten as:

[0077]

[0078] According to the health index values ​​of the mechanical equipment in the early operation stage, the values ​​of 95% of the sequence are sorted from small to large and the values ​​of μ are taken as μ. In formula (10), F(μ) is given by Determine, n is the total number of health indicators, then the tail estimate of F(ε) is:

[0079]

[0080] Under the premise that the tail probability q is 0.05, the adaptive threshold ε can be obtained q :

[0081]

[0082] Step 3: Based on the obtained health indicators, the K-means clustering algorithm is used to divide the operation stages of the mechanical equipment, and then the health status of the mechanical equipment is evaluated based on the logic correction algorithm.

[0083] The created health indicators are used as the input of the K-means clustering algorithm. The health indicators corresponding to the normal operation, early degradation and severe degradation stages are clustered and the cluster centers are calculated respectively. Then, a logical correction algorithm is used to realize the equipment health, sub-health and monitoring operation status evaluation, and then track the degradation status.

[0084] The K-Means clustering algorithm is used to divide the operating stages of mechanical equipment, and the health status is then assessed using a logic correction algorithm. The created health indicators are used as input to the K-Means clustering algorithm to cluster the health indicators corresponding to normal operation, early degradation, and severe degradation stages, and calculate the cluster centers. A logic correction algorithm is used to evaluate the equipment's health, sub-health, and monitoring operating status, thereby tracking degradation status. The K-Means algorithm is a cluster analysis algorithm that primarily calculates data clusters by continuously selecting the nearest mean to a seed point. The K-Means algorithm steps are: a) select k objects from the data as initial cluster centers; b) calculate the distance from each cluster object to the cluster center to divide the cluster; c) recalculate each cluster center; d) calculate the standard measure function until the maximum number of iterations is reached, then stop; otherwise, continue the operation; e) determine the optimal cluster center.

[0085] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0086] like Figure 1 As shown, an embodiment of the present invention provides a method for health monitoring and status assessment of mechanical equipment with an adaptive health threshold, comprising the following steps:

[0087] Step 1: First, the vibration signal of the mechanical equipment is collected based on the vibration sensor. The vibration signal is converted into a singular value feature sequence based on the Hilbert singular value decomposition algorithm, and used as the input of the nonlinear state estimation model.

[0088] Based on the Hankel matrix constructed by the Hilbert singular value decomposition algorithm, the vibration signal X is decomposed into a singular value sequence S=diag(s1, s2, ..., s q ), selecting appropriate singular values ​​for singular value decomposition and reconstruction can eliminate noise from the original signal. Useful information is reflected in the top n singular value sequences. Based on the singular value distribution trend of the specific monitored object, the corresponding singular value sequence is selected as the input feature of the nonlinear state estimation model. The input process memory matrix of the nonlinear state estimation model is denoted by S, and r is the number of input singular values.

[0089]

[0090] Step 2: Construct a health indicator of the nonlinear state estimation reconstruction error to amplify the degree of difference between the initial fault samples and normal samples.

[0091] For any input S obs , the nonlinear state estimation generates an r-dimensional weight vector W = [w1 w2 … w n ] TMake S est =SW=w1S(1)+w2S(2)+…+w r S(r), the weight vector W is determined by the following method: construct the measured data S of the nonlinear state estimation model obs and the predicted value S est The residual between them is ε=S obs -S est , choose W to minimize the residual sum of squares. The residual sum of squares is:

[0092]

[0093] The weight vector W = (S T ·S) T ·(S T ·S obs ). Further, the predicted value S is obtained nest :

[0094]

[0095] Using residual ε=S obs -S nest Health indicators are constructed to monitor the operating status of mechanical equipment. When the working status of mechanical equipment changes, the residual will increase.

[0096] Step 3: Construct an adaptive health threshold based on the peak threshold algorithm. Introduce the peak threshold algorithm into the health indicator, select the health indicator in the normal stage as the health threshold benchmark, and then use the peak threshold algorithm to dynamically update the health threshold.

[0097] According to the health index values ​​of the mechanical equipment in the early operation stage, sorted from small to large and taking 95% of the sequence value as μ, the tail estimate of F(ε) is:

[0098]

[0099] Under the premise that the tail probability q is very small, the adaptive threshold ε can be obtained q , q is taken as 0.05.

[0100]

[0101] Step 4: Use the K-means clustering algorithm to divide the mechanical equipment into different operating stages and evaluate its health status using a logic correction algorithm. The created health indicators are used as input to the K-means clustering algorithm to cluster the health indicators corresponding to normal operation, early degradation, and severe degradation stages, and calculate the cluster centers. A logic correction algorithm is used to evaluate the equipment's health, sub-health, and monitoring operating status, thereby tracking degradation status.

[0102] The K-Means algorithm is a cluster analysis algorithm that is primarily used to calculate data clusters by continuously taking the nearest mean from a seed point. The K-Means algorithm steps are: a) Select k objects from the data as initial cluster centers; b) Calculate the distance from each cluster object to the cluster center to divide it; c) Calculate each cluster center again; d) Calculate the standard measure function until the maximum number of iterations is reached, then stop; otherwise, continue the operation; e) Determine the optimal cluster center.

[0103] The present invention also provides a mechanical equipment health monitoring and status assessment system with adaptive health thresholds. The system includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, it executes the mechanical equipment health monitoring and status assessment method with adaptive health thresholds as described above.

[0104] The present invention also provides a computer-readable storage medium, which stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions prompt the processor to implement the mechanical equipment health monitoring and status assessment method with adaptive health threshold as described above.

[0105] Experimental Data 1: Data from the IMS rolling bearing accelerated life test dataset. This experiment was conducted three times with a sampling frequency of 20 kHz and a sampling interval of 10 minutes. This example uses the full life data of rolling bearing 4 (denoted as IB1-4) from the first experiment as the research object to verify the effectiveness of the adaptive nonlinear state estimation health indicator. Figure 2 The full life degradation signal diagram of the bearing is given.

[0106] Test data 2: from the XJTU-SY rolling bearing accelerated life test data set, the sampling frequency is set to 25.6kHz, and the sampling interval is set to 1min. 15 LDK UER204 rolling bearings were divided into three groups and required to undergo accelerated tests under three different working conditions. The horizontal vibration data of the tested bearings contains more state information. This embodiment selects the full life data of rolling bearing 2 (denoted as XB3-2) collected in the third experiment and the horizontal full life data of rolling bearing 5 (denoted as XB1-5) collected in the first experiment as research objects for validity verification and robustness verification. Figure 3 The full life degradation signal diagram of the bearing is given.

[0107] After the vibration signal is constructed into a singular value feature sequence, it is input into the nonlinear state estimation model to construct the health index. Figure 4The health indicators of IB1-4 are shown in Figure 1. The initial fault position is at the 1240th minute. In order to verify the advantages of the health indicators of this embodiment, traditional health indicators such as RMS and Kurtosis are used for comparison. Figure 5 The health index and Kurtosis proposed in this embodiment have a more stable ability than RMS in characterizing the normal stage of the bearing. At the same time, in the later stage, the health index proposed in this invention has a stronger state characterization ability. Figure 6 The health indicators and adaptive health thresholds of XB3-2 can be obtained, and its initial fault starting position is at the 1230th minute. Figure 7 The initial fault starting position of XB1-5 is at the 35th minute.

[0108] In order to further track the degradation status of equipment, the health index is input into the K-means clustering logic correction algorithm to realize the equipment health status assessment. Figure 6 The XB3-2 health indicators are Figure 7 The corresponding relationship between the XB3-2 real-time health status level, and Figure 8 XB1-5 health indicators and Figure 9 From the correspondence between the XB1-5 real-time health status levels, it can be seen that the real-time health level evaluation of the equipment is divided into three stages: "healthy", "sub-healthy" and "monitoring operation", and the time period is consistent with the health index constructed by nonlinear state estimation.

[0109] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for health monitoring and status assessment of mechanical equipment with an adaptive health threshold, characterized by: The method comprises the following steps: Step 1: The vibration signal representing the operating state of the mechanical equipment is converted into a singular value feature sequence using the Hilbert singular value decomposition algorithm; the singular value feature sequence is input into a nonlinear state estimation model, and the nonlinear state estimation model outputs a health indicator for the nonlinear state estimation reconstruction error; Step 2: Select the health indicator in the normal stage as the health threshold benchmark, and introduce the peak over-threshold algorithm to dynamically update the health threshold; Step 3: Based on the obtained health indicators, the K-means clustering algorithm is used to divide the operation stages of the mechanical equipment, and then the health status of the mechanical equipment is evaluated based on the logic correction algorithm.

2. The method for health monitoring and status assessment of mechanical equipment with adaptive health thresholds according to claim 1, characterized in that: Based on the Hankel matrix constructed by Hilbert singular value decomposition algorithm, the original signal X is decomposed into a singular value sequence S=diag(s1, s2, ..., s q ), select the singular value to perform singular value decomposition and reconstruction.

3. The method for health monitoring and status assessment of mechanical equipment with adaptive health threshold according to claim 2, characterized in that: The input process memory matrix of the nonlinear state estimation model is denoted as S, r is the number of input singular values, and m is the number of monitoring samples, then: The input of the nonlinear state estimation model is the future test data S of the mechanical equipment obs , the output is for S obs The predicted value S est .

4. The method for health monitoring and status assessment of mechanical equipment with adaptive health threshold according to claim 3, characterized in that: For any input S obs , the nonlinear state estimation model generates an r-dimensional weight vector W = [w1 w2 … w n ] T , such that: S est =SW=w1S(1)+w2S(2)+…+w r S(r) (2) The predicted value X of the nonlinear state estimation model est is a linear combination of r historical observation vectors in S; the weight vector W is determined by the following method: construct the measured data S of the nonlinear state estimation model obs and the predicted value S est The residual between is: e=S obs -S est (3) Choose W to minimize the residual sum of squares, the residual sum of squares is: Using partial derivative to solve equation (4) we can get the weight vector W = (S T ·S) T ·(S T ·S obs );Get the predicted value S nest : The process memory matrix S represents the entire dynamic process of normal operation of the mechanical equipment; the residual ε=S obs -S nest Health indicators are constructed to monitor the operating status of mechanical equipment. When the working status of mechanical equipment changes, the residual will increase.

5. The method for health monitoring and status assessment of mechanical equipment with adaptive health threshold according to claim 1, characterized in that: Let μ be the initial threshold, and the samples exceeding μ are recorded as N μ is the number of samples, the excess variable is recorded as y = ε - μ, and the corresponding excess distribution function is expressed as: F μ (y) seems to be a generalized Pareto distribution, that is: Where ξ is the scale parameter, β is the shape parameter; the estimated parameters and The maximum likelihood estimation method is used to obtain: When y>0, formula (6) is rewritten as: According to the health index values ​​of the mechanical equipment in the early operation stage, the values ​​of 95% of the sequence are sorted from small to large and the values ​​of μ are taken as μ. In formula (10), F(μ) is given by Determine, n is the total number of health indicators.

6. The method for health monitoring and status assessment of mechanical equipment with adaptive health threshold according to claim 5, characterized in that: The tail estimate of F(ε) is: Under the premise that the tail probability q is 0.05, the adaptive threshold ε is obtained q :

7. A mechanical equipment health monitoring and status assessment system with adaptive health thresholds, characterized by: The system includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the method for health monitoring and status assessment of mechanical equipment with adaptive health threshold according to any one of claims 1 to 6 is executed.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by the processor, the machine-executable instructions prompt the processor to implement the mechanical equipment health monitoring and status assessment method with adaptive health threshold according to any one of claims 1 to 6.

Images