A hierarchical progressive mechanical equipment health management method

Abstract

The present invention relates to the field of health prediction and management of mechanical equipment, and discloses a hierarchical progressive mechanical equipment health management method, including: 1) collecting vibration monitoring signals during the operation of the equipment to provide a data basis for constructing a health management model; 2) setting the value of a single scale factor, extracting the fixed-scale dispersion entropy features of the vibration signal samples, and realizing the online detection of the equipment health status based on the statistical dispersion entropy threshold; 3) setting the values of multiple scale factors, and combining with the CEEMDAN method to extract the multi-scale dispersion entropy features of the fault signal samples; 4) realizing the offline identification of the equipment fault types based on a fault classifier. The present invention can meet the requirements for quickly and accurately detecting the health status of mechanical equipment under the background of data-driven, accurately locate the abnormal or fault occurrence parts and severity for the abnormal operation state of the equipment, and provide a necessary basis for formulating a reasonable and feasible equipment maintenance plan and ensuring the safe and reliable operation of the equipment.

Description

Technical Field

[0001] The present invention relates to the field of mechanical equipment health prediction and management, and in particular to a hierarchical and progressive mechanical equipment health management method. Background Art

[0002] Ensuring the safe, stable and reliable operation of mechanical equipment is of great significance to modern industrial production capacity and corporate economic benefits. Considering the need for production site operation and maintenance personnel to pay attention to the real-time operation status of equipment, fast and accurate equipment health status detection results can provide necessary data support for operation and maintenance personnel to formulate reasonable maintenance plans and implement effective health management. In addition, for equipment operating in abnormal or faulty states, studying effective equipment fault identification model construction methods to achieve accurate and efficient fault type discrimination has become a technical problem that needs to be solved urgently.

[0003] The existing mechanical equipment health management model mainly adopts a single-layer, single-step structure mode, that is, the collected signal samples are directly input into the state type discrimination model to complete the one-step judgment of the equipment health status. The above method ignores the characteristic boundaries of normal or faulty states, does not take into account the urgent needs of operation and maintenance personnel, and significantly reduces the efficiency of using the iconic information in a large number of normal signal samples that are easy to collect. Therefore, studying hierarchical and progressive mechanical equipment health management models and methods can provide a new idea for solving the above problems. At present, there has been no systematic and in-depth research on hierarchical and progressive equipment health management models and methods, and there is a lack of a theoretical system for constructing health management models with adaptability and strong robustness, which cannot meet the functional requirements of online detection of equipment health status and offline identification of fault types at the same time. Summary of the invention

[0004] Purpose of the invention: In view of the problems existing in the prior art, the present invention provides a hierarchical and progressive method for the health management of mechanical equipment, establishes a method system for simultaneously realizing online detection of the health status of equipment and offline identification of fault types, and solves the technical difficulties of the prior art that ignores the characteristic boundaries of normal or faulty states, has low adaptability, insufficient analysis accuracy, and is difficult to complete accurate prediction and reliable management of the health of equipment.

[0005] Technical solution: The present invention provides a hierarchical and progressive mechanical equipment health management method, comprising the following steps:

[0006] Step 1: Collect vibration monitoring signals during equipment operation to provide a data basis for building a health management model;

[0007] Step 2: Set a single scale factor value to extract the fixed-scale spread entropy features of the vibration signal samples, and realize online detection of the equipment health status based on the statistical spread entropy threshold;

[0008] Step 3: Based on the health status detection results in Step 2, set the values of multiple scale factors, and combine with the CEEMDAN method to extract the multi-scale dispersion entropy features of the fault signal samples;

[0009] Step 4: Based on the fault classifier, use the multi-scale dispersion entropy features of the fault signal samples described in Step 3 as the input to achieve the offline identification of the equipment fault types.

[0010] Furthermore, Step 1 includes the following sub-steps:

[0011] (1-1) According to the structural composition and layout type of the mechanical equipment, arrange several vibration signal monitoring points at different parts of the equipment;

[0012] (1-2) Using the arranged vibration signal monitoring points, collect the vibration signal sample set {X1, X2, L X n}, where X i = x i1 , x i2 , L, x il , l represents the length of a single signal sample, and n represents the number of samples.

[0013] Furthermore, Step 2 includes the following steps:

[0014] (2-1) Set the value of a single scale factor ξ0;

[0015] (2-2) Extract the fixed-scale dispersion entropy eigenvalue of the vibration signal sample X i to be detected. The feature extraction formula is as follows:

[0016]

[0017] Among them, s represents the embedding dimension, c represents the number of mapping categories, represents the dispersion pattern, and p(·) represents the relative probability calculation function;

[0018] (2-3) Based on the historical sample set, determine the statistical dispersion entropy threshold H θ , and the calculation formula is as follows:

[0019]

[0020] Among them, represents the mean value of the dispersion entropy features of the historical sample set, S t represents the standard deviation of the dispersion entropy distribution of the historical sample set, and η represents the confidence coefficient;

[0021] (2-4) If the following conditions are met: H(X i , ξ0) ≥ Hθ , the vibration signal sample X to be detected i is a sample collected under the normal operation state of the device; if the above conditions are not satisfied, the vibration signal sample X to be detected i is a sample collected under the abnormal or faulty state of the device.

[0022] Furthermore, step 3 includes the following sub-steps:

[0023] (3-1) Using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) method, decompose the abnormal or faulty vibration signal sample X j to obtain m Intrinsic Mode Function (IMF) components, expressed as follows:

[0024]

[0025] (3-2) Set the values of the multi-scale factors: ξ1, ξ2, …, ξ k , where k represents the number of scale factor values;

[0026] (3-3) For the set of faulty signal samples {X1, X2, …, X r}, calculate its multi-scale dispersion entropy feature space, and the form of feature space construction is as follows:

[0027] [Sample i (H(IMF j , ξ w ))] r×(m×k)

[0028] where IMF j represents the Intrinsic Mode Function component in (3-1), j takes values from 1 to m, w takes values from 1 to k, and i takes values from 1 to r.

[0029] Furthermore, step 4 includes the following sub-steps:

[0030] (4-1) Select the Softmax model as the device fault type classifier;

[0031] (4-2) Use the multi-scale dispersion entropy features of the set of faulty signal samples obtained in step 3 as the input of the Softmax model to realize the offline identification of the device fault type.

[0032] Beneficial effects:

[0033] (1) The present invention first introduces the hierarchical and progressive thinking into the field of mechanical equipment health management, comprehensively considers the engineering requirements of online detection of equipment health status and offline identification of fault types, overcomes the limitation that single-layer and single-step diagnosis strategies are difficult to guide maintenance personnel to effectively maintain equipment, avoids the problem of inefficient utilization of a large number of easily collected normal sample status flag information, fills the gap that there is no scientific theoretical basis and technical guidance for equipment to carry out hierarchical health management, effectively meets the requirements of detecting, identifying and managing the health status of mechanical equipment under the background of data-driven, and provides necessary support for promoting the transformation from after-failure maintenance to condition-based maintenance.

[0034] (2) The present invention first considers constructing a fixed-scale dispersion entropy feature and a multi-scale dispersion entropy feature space, which are respectively applied to the online detection of the health status of mechanical equipment and the offline identification of fault types, significantly improves the accuracy of equipment health status analysis, overcomes the theoretical problem that it is difficult to effectively and adaptively construct detection and identification models, effectively improves the level of assessability and judgment of equipment health status, and provides reliable theoretical guidance and technical support for implementing intelligent health management of equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The present invention provides a flowchart of a hierarchical and progressive mechanical equipment health management model and method;

[0036] Figure 2 It is a change curve of the vibration signal sequence of the rolling bearing of a harvester equipment in an embodiment of the present invention;

[0037] Figure 3 It is a diagram of the online detection result of the health status of the rolling bearing of a harvester equipment in an embodiment of the present invention;

[0038] Figure 4 It is a diagram of the decomposition result of the fault vibration signal sample of the inner ring of the rolling bearing of a harvester equipment in an embodiment of the present invention;

[0039] Figure 5 It is a diagram of the offline identification result of the fault type of the rolling bearing of a harvester equipment in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The present invention will be introduced in detail below with reference to the accompanying drawings.

[0041] In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0042] The present invention takes the construction of a health management model and method analysis of the rolling bearing of a certain harvester equipment as an example, Figure 1The following is a flowchart of a hierarchical progressive mechanical equipment health management method provided by the present invention, which specifically includes the following steps:

[0043] Step 1: Collect the vibration monitoring signals of the rolling bearings during the operation of the harvester equipment to provide a data basis for the construction of the health management model.

[0044] (1-1) Arrange several vibration signal monitoring points at the rolling bearings of the harvester equipment.

[0045] (1-2) Use the arranged vibration signal monitoring points to collect the vibration signal sample set {X1, X2, L X n}, where X i = x i1 , x i2 , L, x il , l represents the length of a single signal sample, and n represents the number of samples. Taking the construction and method analysis of the rolling bearing health management model of a certain harvester equipment as an example, the description of the bearing health state types and signal samples is shown in Table 1. As Figure 2 shown in the vibration signal sequence change curve of the rolling bearings of the harvester equipment. From the results shown in the figure, it can be seen that the vibration signals under different bearing health state types are all high-frequency and irregular sequence curves, and it is impossible to directly distinguish the bearing health states. Therefore, new exploration and attempts need to be made for the construction method of the bearing health state management model.

[0046] Table 1 Description of rolling bearing health state types and signal samples

[0047] Health status type Abbreviation Sample quantity Category label Normal Nor 20 1 Ball fault BF 20 2 Inner loop fault IF 20 3 Outer loop fault OF 20 4 Combined fault CF 20 5

[0048] Step 2: Set the value of a single scale factor, extract the fixed-scale dispersion entropy characteristics of the vibration signal samples, and realize the online detection of the equipment health state based on the statistical dispersion entropy threshold.

[0049] (2-1) Set the value of a single scale factor ξ0. In this example, the value of the single scale factor is as follows: ξ0 = 20.

[0050] (2-2) Extract the fixed-scale dispersion entropy characteristic values of the vibration signal samples X i to be detected. The feature extraction formula is as follows:

[0051]

[0052] where s represents the embedding dimension, c represents the number of mapping categories, represents the dispersion pattern, and p(·) represents the relative probability calculation function. The settings of the above parameters are as follows: the embedding dimension s is 4, and the number of mapping categories c is 6.

[0053] (2-3) Determine the statistical dispersion entropy threshold H based on the historical sample set θ , and the calculation formula is as follows:

[0054]

[0055] Where, represents the mean value of the dispersion entropy feature of the historical sample set, and S t represents the standard deviation of the dispersion entropy distribution of the historical sample set, and η represents the confidence coefficient. In this embodiment, the confidence coefficient η takes the value of 2.58. After calculation, the statistical dispersion entropy threshold H θ is: H θ = 1.397.

[0056] (2-4) If the following conditions are met: H(X i , ξ0) ≥ H θ , then the vibration signal sample X to be detected i is the sample collected under the normal operation state of the device; if the above conditions are not met, the vibration signal sample X to be detected i is the sample collected under the abnormal or faulty state of the device. As Figure 3 is the online detection result diagram of the health state of the rolling bearing of the harvester equipment. It can be seen that based on the calculated statistical dispersion entropy threshold H θ , all normal samples and samples of 4 different fault types can be completely distinguished, effectively realizing the online detection of the bearing health state.

[0057] Step 3: Based on the health state detection result of Step 2, set the values of the multi-scale factors, and combine the CEEMDAN method to extract the multi-scale dispersion entropy features of the fault signal samples.

[0058] (3-1) Use the CEEMDAN method to decompose the abnormal or faulty vibration signal sample X j , and m intrinsic mode function (IMFs) components can be obtained, which are expressed as follows:

[0059]

[0060] As Figure 4 is the decomposition result diagram of the inner ring fault vibration signal sample of the rolling bearing of the harvester equipment. It can be seen that after the fault vibration signal sample is decomposed by CEEMDAN, 15 IMFs components and a residual component r can be obtained.

[0061] (3-2) Set the values of the multi-scale factors: ξ1, ξ2, L, ξ k , where k represents the number of values of the scale factor. In this embodiment, the values of the multi-scale factors are as follows: ξ1 = 1, ξ2 = 2, L, ξ 20 = 20.

[0062] (3-3) For the set of fault signal samples {X1, X2, ..., X r}, calculate its multi-scale dispersion entropy feature space, and the form of feature space construction is as follows:

[0063] [Sample i (H(IMF j , ξ w )] r×(m×k)

[0064] Among them, IMF j represents the intrinsic mode function component in (3-1), the value of j ranges from 1 to 15, the value of w ranges from 1 to 20, and the value of i ranges from 1 to r.

[0065] Step 4: Based on the fault classifier, realize the offline identification of equipment fault types.

[0066] (4-1) Select the Softmax model as the equipment fault type classifier.

[0067] (4-2) Use the multi-scale dispersion entropy features of the fault signal sample set obtained in step 3 as the input of the Softmax model to realize the offline identification of equipment fault types.

[0068] As Figure 5 shown is the offline identification result diagram of the fault types of the rolling bearings of the harvester equipment. It can be seen from the shown results that based on the multi-scale dispersion entropy features of the extracted fault signal samples, accurate identification of different fault type samples can be realized, fully verifying the effectiveness of the above hierarchical progressive mechanical equipment health management method.

[0069] The above embodiments are only for illustrating the technical concept and features of the present invention, and the purpose is to enable those who are familiar with this technology to understand the content of the present invention and implement it accordingly, and it cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit and essence of the present invention should be covered within the protection scope of the present invention.

Claims

1. A hierarchical progressive mechanical equipment health management method, characterized in that The steps are as follows: Step 1: Collect vibration monitoring signals during the operation of the device to provide a data basis for constructing a health management model; (1-1) According to the structural composition and layout type of the mechanical equipment, arrange a number of vibration signal monitoring points at different parts of the equipment; (1-2) By using the arranged vibration signal monitoring measuring points, collect the vibration signal sample set {X1, X2, …, X n} during the operation of the device, where X i = x i1 , x i2 , …, x il , l represents the length of a single signal sample, and n represents the number of samples; Step 2: Set the value of a single scale factor, extract the fixed-scale dispersion entropy characteristics of the vibration signal samples, and based on the statistical dispersion entropy threshold, realize the online detection of the equipment health status; (2-1) Set the value of a single scale factor ξ0; (2-2) Extract the vibration signal sample X to be detected i of the fixed-scale dispersion entropy eigenvalue. The feature extraction formula is as follows: where s represents the embedding dimension and c represents the number of mapping categories, represents the scattering pattern, represents the relative probability calculation function; (2-3) Determine the statistical dispersion entropy threshold H based on the historical sample set θ , and the calculation formula is as follows: Among them, represents the mean of the historical sample set scatter entropy feature, and S t represents the standard deviation of the historical sample set scatter entropy distribution, and η represents the confidence coefficient; (2-4) If the following conditions are satisfied: H(X i , ξ0) ≥ H θ , then the vibration signal sample X to be detected i is a sample collected under the normal operating state of the device; if the above conditions are not satisfied, then the vibration signal sample X to be detected i is a sample collected under the abnormal or faulty state of the device; Step 3: Based on the health status detection results of Step 2, set the values of multiple scale factors, and combined with the adaptive noise complete ensemble empirical mode decomposition CEEMDAN method, extract the multi-scale dispersion entropy characteristics of the fault signal samples; (3-1) Using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) method, decompose the abnormal or fault vibration signal sample X j to obtain m intrinsic mode function components, which are expressed as follows: (3-2) Set the values of multiple scale factors: ξ1, ξ2, …, ξ k , where k represents the number of values of the scale factor; (3-3)For the set of fault signal samples {X1, X2, …, X r}}, calculate its multi-scale dispersion entropy feature space, and the construction form of the feature space is as follows: [Sample j (H(IMF i ,ξ w ))] r×(m×k) Among them, IMF i represents the inherent model function component in (3-1), where the value range of i is from 1 to m, the value range of w is from 1 to k, and the value range of j is from 1 to r; Step 4: Based on the fault classifier, use the multi-scale dispersion entropy characteristics of the fault signal samples described in Step 3 as the input to realize the offline identification of the equipment fault type.

2. The hierarchical progressive mechanical equipment health management method according to claim 1, wherein The said Step 4 includes the following sub-steps: (4-1) Select the Softmax model as the equipment fault type classifier; (4-2) Use the multi-scale dispersion entropy characteristics of the fault signal sample set obtained in Step 3 as the input of the Softmax model to realize the offline identification of the equipment fault type.

Images