Motor Shaft Fault Risk Pattern Identification Method Based on Interval Evidence Fusion of Normal Cloud Model

Abstract

The present invention discloses a method for identifying the fault risk mode of a motor shafting based on the interval evidence fusion of a normal cloud model. This method is based on the analysis of typical fault data to construct cloud models of various fault template modes; according to the data of different fault characteristics of the online monitored motor shafting, a cloud model of the fault to-be-detected mode is constructed; the cloud model of the to-be-detected mode is matched with the cloud models of various fault template modes to obtain the matching degree interval of the to-be-detected mode for different fault template modes; the obtained matching degree interval is normalized to obtain the interval evidence for fusion; the interval evidence is fused, and according to certain identification criteria, the fault risk mode is determined. The present invention utilizes the advantages of the normal cloud model in describing the fuzziness and randomness of objective objects, and obtains more accurate interval evidence by constructing the cloud models of the to-be-detected mode and the fault template mode, so that the identification result of the fault risk mode obtained based on the interval evidence fusion is more in line with the actual situation.

Description

Technical Field

[0001] The present invention relates to a method for identifying the fault risk mode of a motor shafting based on interval evidence fusion of a normal cloud model, belonging to the technical field of fault diagnosis and maintenance of mechanical equipment. Background Art

[0002] With the development of science and technology, motor shafting equipment has been gradually applied to some complex, high-speed and precision mechanical motion places. Due to the increasingly close connection between mechanical components, with the increase of the service life, the motor shafting equipment will inevitably fail. Once a failure occurs, it may directly cause the entire mechanical system to fail to operate normally, thereby affecting work efficiency, causing economic losses, and even triggering related safety problems. Therefore, using fault risk mode identification technology to accurately judge the occurrence of fault risks in a timely manner has important practical significance for improving the safety of equipment.

[0003] When identifying the fault risk mode of a motor shafting system, often due to the influence of the measurement environment and the real-time change of the operating state of the equipment itself, there is rarely a clear one-to-one correspondence between the fault risk mode generated by the motor shafting system and a single fault feature. Through traditional single-sensor, single-factor monitoring and single-model processing and identification methods, it is no longer possible to accurately and efficiently judge the corresponding fault risk mode of the motor shafting system. In order to increase the accuracy of fault risk identification, through multi-source information fusion technology, multiple sensors are used to collect different fault feature information, and according to these information, a certain rule is used for fusion, so as to obtain a more accurate description of the fault risk mode, and then improve the reliability of identification. Summary of the Invention

[0004] The purpose of the present invention is to propose a method for identifying the fault risk mode of a motor shafting based on interval evidence fusion of a normal cloud model. Through the description of different fault risk modes by various fault feature information, interval evidence is obtained according to the normal cloud model, and the interval evidence under different fault features obtained is fused by using the evidence reasoning rule. Based on the fused interval belief, the determination of the fault risk mode of the motor shafting is obtained according to the identification criterion.

[0005] The method for identifying the fault risk mode of a motor shafting based on interval evidence fusion of a normal cloud model proposed by the present invention includes the following steps:

[0006] (1) Set the rotational speed of the shafting to ρ revolutions per minute, 1000 ≤ ρ ≤ 2500, and let the fault set composed of the fault risk modes of the shafting system be Θ = {X, Y, Z}, where X represents the unbalance fault in the motor shafting, Y represents the misalignment fault in the motor shafting, and Z represents the foundation looseness fault in the motor shafting;

[0007] (2) By vibration acceleration sensors and vibration displacement sensors respectively installed in the horizontal and vertical directions of the shafting support seats, continuously collect the time-domain vibration sequence at set time intervals, obtain the amplitudes at 1 to 3 times the frequency of vibration acceleration through Fourier transform, and the average amplitude of time-domain vibration displacement, and use them as fault characteristics;

[0008] Set that the shafting system is in the fault risk mode U, U ∈ {X, Y, Z}, and sequentially collect the data of 4 fault characteristics including the amplitudes of 1 times frequency, 2 times frequency, and 3 times frequency of vibration acceleration and the average amplitude of time-domain vibration displacement, denoted as where i = 1, 2, 3, 4 represent the above 4 fault characteristics in sequence, N represents the number of collected data samples, and 50 ≤ N ≤ 500;

[0009] (3) Set the sample pattern cloud model of the fault risk mode U under the i-th fault characteristic as The cloud droplets constituting the sample pattern cloud model are denoted as where g = 1, 2, … G, G represents the total number of generated cloud droplets, 1000 ≤ G ≤ 5000, and construct the sample pattern cloud models of 3 fault risk modes under 4 fault characteristics;

[0010] (4) Online obtain a set of samples to be tested, denoted as F i , where N1 ≤ N, representing the total number of online data samples;

[0011] (5) Use F i to replace the data of the fault risk mode U under the fault characteristic i in step (3), and then repeat the entire calculation process of step (3) to obtain the cloud model to be tested under 4 fault characteristics: {C1, C2, C3, C4}, C i = {d i,1 ,..., d i,g ,..., d i,G |d i,g = {c i,g , μ(c i,g )}};

[0012] (6) Under the fault characteristic i, calculate the matching degree between the sample pattern cloud model and the cloud model C to be tested i , and further obtain the interval evidence that the cloud model C to be tested i supports the occurrence of 3 fault risk modes;

[0013] (7) Fuse the interval evidence obtained in step (6) using the evidence reasoning rule;

[0014] (8) According to step (7), obtain H4 The secondary fusion result, and calculate the interval evidence of q obtained after fusion as

[0015] M(q) = [a θ , b θ (1)

[0016]

[0017]

[0018] According to formulas (1)-(3), the interval evidence of the fused normal cloud model can be obtained as shown in Table 1:

[0019] Table 1 Fusion result table

[0020]

[0021]

[0022] (9) Give the fault risk mode identification criterion according to the interval evidence shown in Table 1, that is, the fault risk mode pointed to by the sample F to be detected can be determined by meeting the following two conditions: i q, θ ∈ {X, Y, Z, Θ}:

[0023] ① The left and right endpoints of the interval evidence of M(θ) (θ ≠ Θ) are respectively greater than the left and right endpoints of the interval evidence corresponding to other fault risk modes;

[0024] ② The right endpoint of M(Θ) is less than 0.3.

[0025] Advantages of the present invention:

[0026] First, the present invention utilizes the advantages of the normal cloud model in describing the fuzziness and randomness of objective objects, and obtains more accurate interval evidence by constructing the cloud models of the pattern to be detected and the fault template pattern, making the fault risk identification result obtained based on the interval evidence fusion more in line with the actual situation.

[0027] Second, since there is no clear mapping relationship between a single fault feature and the fault risk mode of the motor shafting, therefore, the present invention synthesizes the information under different fault features, adopts a variety of information fusion methods, and performs fusion according to the evidence reasoning rule, thereby increasing the accuracy of the fault risk mode identification.

[0028] Third, due to the disadvantages of the single-value evidence in the incomplete, rough measurement of uncertain information or fuzzy information and the possible loss of a lot of useful information in the measurement process, etc., the present invention adopts interval evidence, thus ensuring the integrity of the uncertain information measurement and making the fusion identification result more comprehensive and reliable. Description of the drawings

[0029] Figure 1 is the flowchart of the method of the present invention.

[0030] Figure 2 is the matching description diagram of the pattern to be detected and the cloud models of three kinds of fault sample patterns in the embodiment of the method of the present invention.

[0031] Figure 3 is the structure diagram of the motor shafting fault risk pattern identification system in the embodiment of using the method of the present invention.

[0032] Figure 4 is the cloud model matching diagram of the pattern to be detected and the cloud models of three kinds of fault sample patterns under the fault feature of "vibration acceleration 1x frequency amplitude" in the embodiment of the present invention.

[0033] Figure 5 is the cloud model matching diagram of the pattern to be detected and the cloud models of three kinds of fault sample patterns under the fault feature of "vibration acceleration 2x frequency amplitude" in the embodiment of the present invention.

[0034] Figure 6 is the cloud model matching diagram of the pattern to be detected and the cloud models of three kinds of fault sample patterns under the fault feature of "vibration acceleration 3x frequency amplitude" in the embodiment of the present invention.

[0035] Figure 7 is the cloud model matching diagram of the pattern to be detected and the cloud models of three kinds of fault sample patterns under the fault feature of "average amplitude of time-domain vibration displacement" in the embodiment of the present invention. Detailed implementation manners

[0036] The present invention relates to a method for identifying motor shafting fault risk patterns based on normal cloud model interval evidence fusion. Based on the analysis of typical fault data, cloud models of various fault sample patterns are constructed under different fault features; according to the data of the motor shafting under different fault features monitored online, a cloud model of the pattern to be detected for faults is constructed; the cloud model of the pattern to be detected is matched with the cloud models of various fault sample patterns to obtain the matching degree interval of the pattern to be detected for different fault sample patterns; the obtained matching degree interval is normalized to obtain the interval evidence for fusion; the interval evidence is fused using the evidence reasoning rule, and according to a certain identification criterion, the fault risk pattern is determined. The present invention specifically includes the following steps:

[0037] (1) Set the rotational speed of the shafting to ρ revolutions per minute, 1000 ≤ ρ ≤ 2500, and let the fault set composed of the fault risk patterns of the shafting system be Θ = {X, Y, Z}, where X represents the unbalance fault in the motor shafting, Y represents the misalignment fault in the motor shafting, and Z represents the foundation looseness fault in the motor shafting.

[0038] (2) By vibration acceleration sensors and vibration displacement sensors respectively installed in the horizontal and vertical directions of the shafting support seats, continuously collect the time-domain vibration sequences at set time intervals, and obtain the amplitudes at 1-3 times the frequency of vibration acceleration and the average amplitude of time-domain vibration displacement through Fourier transform, and use them as fault characteristics.

[0039] Set the shafting system to be in the fault risk mode U, U ∈ {X, Y, Z}, and sequentially collect the data of 4 fault characteristics, namely the amplitudes of 1 times frequency, 2 times frequency, 3 times frequency of vibration acceleration and the average amplitude of time-domain vibration displacement, denoted as where i = 1, 2, 3, 4 represent the above 4 fault characteristics in sequence, N represents the number of collected data samples, and 50 ≤ N ≤ 500.

[0040] (3) Set the sample mode cloud model of the fault risk mode U under the i-th fault characteristic as The cloud droplets constituting the sample mode cloud model are denoted as where g = 1, 2, … G, G represents the total number of generated cloud droplets, 1000 ≤ G ≤ 5000, and construct the sample mode cloud models of 3 fault risk modes under 4 fault characteristics.

[0041] (4) Online obtain a set of samples to be tested, denoted as F i , where N1 ≤ N, representing the total number of online data samples.

[0042] (5) Use F i to replace the data of the fault risk mode U under the fault characteristic i in step (3), and then repeat the entire calculation process of step (3) to obtain the cloud models to be tested under 4 fault characteristics: {C1, C2, C3, C4}, C i = {d i,1 ,..., d i,g ,..., d i,G |d i,g = {c i,g , μ(c i,g )}}.

[0043] (6) Under the fault characteristic i, calculate the matching degree between the sample mode cloud model and the cloud model C to be tested i , and then obtain the interval evidence that the cloud model C to be tested i supports the occurrence of 3 fault risk modes.

[0044] (7) Fuse the interval evidence obtained in step (6) using the evidence reasoning rule.

[0045] (8) According to step (7), obtain H4 The secondary fusion result, and calculate the interval evidence of θ obtained after fusion as

[0046] M(θ) = [a θ , b θ (1)

[0047]

[0048]

[0049] According to equations (1)-(3), the interval evidence of the fused normal cloud model can be obtained, as shown in Table 2:

[0050] Table 2 Fusion result table

[0051]

[0052] (9) Give the fault risk mode identification criterion according to the interval evidence shown in Table 2, that is, the fault risk mode pointed to by the sample to be tested F can be determined by satisfying the following two conditions i is θ, θ ∈ {X, Y, Z, Θ}:

[0053] ① For the interval evidence of M(θ), (θ ≠ Θ), the left and right endpoints are respectively greater than the left and right endpoints of the corresponding interval evidence of other fault risk modes;

[0054] ② The right endpoint of M(Θ) is less than 0.3.

[0055] Furthermore, the said fundamental frequency is (ρ / 60) Hz, the second harmonic frequency is (ρ / 30) Hz, and the third harmonic frequency is (ρ / 15) Hz.

[0056] Furthermore, the vibration acceleration sensor and the vibration displacement sensor continuously collect the time-domain vibration sequence at a time interval of Δt = 16 s.

[0057] Furthermore, step (3) is specifically:

[0058] (3.1) From step (2), the data of the fault risk mode U under the fault feature i can be obtained Take its arithmetic mean as the expected value EX of the cloud model i,U

[0059]

[0060] (3.2) Set the entropy to be En, 0.05 ≤ En ≤ 0.1, the hyperentropy value to be He, 0.005 ≤ He ≤ 0.01, and

[0061] (3.3) Generate a normal random number with entropy En as the expectation and hyper-entropy He as the standard deviation Then, with EX i,U as the expectation, as the standard deviation, generate a normal random number as the cloud droplet

[0062] (3.4) Calculate the certainty degree corresponding to the cloud droplet and the certainty degree has the property of probability likelihood. The specific calculation method is as follows:

[0063]

[0064] (3.5) Let reflect the position information of the cloud droplet in the two-dimensional space. Repeat the calculation process of steps (3.3)-(3.4) G times until G cloud droplets and their respective certainty degrees are generated, then the template mode cloud model of the fault risk mode U under the fault feature i can be obtained Then, for 3 fault risk modes under 4 fault features, a total of 4×3 = 12 template mode cloud models can be generated:

[0065] The template mode cloud models of 3 fault risk modes under the 1x frequency amplitude feature are

[0066] The template mode cloud models of 3 fault risk modes under the 2x frequency amplitude feature are

[0067] The template mode cloud models of 3 fault risk modes under the 3x frequency amplitude feature are

[0068] And the template mode cloud models of 3 fault risk modes under the time-domain vibration displacement average amplitude feature are

[0069] Furthermore, step (6) is specifically:

[0070] (6.1) Construct the entropy-containing expectation curve of the template mode cloud model as The entropy-containing expectation curve of the cloud model C of the pattern to be inspected i is where and x i are random variables and follow the Gaussian distribution

[0071] (6.2) Under the fault feature i, obtain the of the 3 template mode cloud models and the V of the cloud model of the pattern to be inspectedi (x i ) Its coordinates in two-dimensional space are The specific calculation method is as follows:

[0072]

[0073]

[0074] (6.3) From step (3.5), we can get the coordinates of cloud droplets in two-dimensional space of three sample cloud models under fault feature i: set up The distance intersection The nearest K cloud droplets, and 100≤K≤250, the certainty of these cloud droplets μ(c i,k ),μ(c i,k )∈[0,1] is the cloud model C of the pattern to be tested i Cloud Model with Sample Pattern The matching degree of Pick The minimum value is a i,U , The maximum value is b i,U , Compose matching interval

[0075] I i (U) = [a i,U ,b i,U ] (8)

[0076] (6.4) Will I i (U) is normalized to obtain the normalized matching degree interval:

[0077]

[0078]

[0079]

[0080]

[0081] (6.5) The confidence interval of the complete set of faults Θ = {X, Y, Z} is calculated as

[0082]

[0083]

[0084]

[0085] Mi (Θ) measures the overall uncertainty of the match in step (6.4);

[0086] (6.6) The normal cloud model interval evidence is formed by equation (9) and equation (13), as shown in Table 3

[0087] Table 3 Interval evidence table obtained through the cloud model

[0088]

[0089]

[0090] In order to deepen the understanding of matching the cloud model of the mode to be inspected with the sample cloud model of the three fault risk modes to form interval evidence, an example is given here. Since the entropy expectation curve reflects the contour characteristics of the cloud model and is the skeleton of the cloud droplet set, all cloud droplets fluctuate randomly near the entropy expectation curve. The matching degree of the fault sample cloud model and the cloud model of the mode to be inspected can be determined according to the intersection of the entropy expectation curve. Assume that the sample cloud model of the three fault risk modes of the motor shaft equipment under the 1-fold frequency characteristic The parameters (expectation, entropy, and super entropy) of the cloud model C1 of the test pattern are shown in Table 4:

[0091] Table 4. Parameters of cloud model under 1-fold frequency characteristics

[0092]

[0093] Through the method described in step (3), it is possible to obtain Figure 2 The cloud model of the to-be-tested mode and the sample mode cloud models of the three fault risk modes are shown in the figure. The intersection coordinates of the entropy expectation curves of the to-be-tested mode cloud model and the sample mode cloud model can be obtained through formulas (6)-(7): Right now Figure 2 Then select the 200 cloud droplets closest to the corresponding intersection in the sample pattern cloud model of each fault risk pattern. The certainty corresponding to these 200 cloud droplets is the matching degree of the test pattern to the fault sample pattern. The maximum and minimum values ​​are selected to form the matching degree interval as shown in the following table:

[0094] Table 5 Matching interval table under 1-fold frequency characteristics

[0095]

[0096] According to formulas (10)-(12) for normalization, the confidence interval of the fault set Θ can be obtained by formulas (14)-(15). The interval evidence under the 1-fold frequency characteristic is shown in the following table:

[0097] Table 6 Evidence interval table under 1-fold frequency amplitude characteristics

[0098]

[0099] According to the same steps above, interval evidences under the characteristics of the second - harmonic amplitude, third - harmonic amplitude, and average vibration displacement amplitude can be obtained. The interval evidences under the four fault characteristics are shown in Table 7 as follows:

[0100] Table 7 Interval Evidence Table

[0101]

[0102] Furthermore, step (7) is specifically as follows:

[0103] (7.1) Take H (H≥2) points respectively within the interval θ∈{X,Y,Z,Θ}, and record the point values taken as and satisfy the following conditions:

[0104] ①

[0105] ②

[0106] (7.2) According to the method described in step (7.1), H point - value evidences under the fault characteristic i can be obtained, denoted as E i ={e i,1 ,...,e i,h ,...,e i,H}, where

[0107] (7.3) Set the reliability factor of the evidence e i,h as r λ , satisfying 0≤r λ ≤1, λ = 1, 2, 3, 4, and the importance weight w λ =r λ ;

[0108] (7.4) Through step (7.2), 4H point - value evidences under the four fault characteristics can be obtained. Select any one point - value evidence under each fault characteristic for combination, and each combination contains point - value evidences of the four fault characteristics. Then there are H 4 combinations. Use the evidence reasoning rule to fuse the 4 point - value evidences in each combination, and l fusion results (l = 1, 2,…,H 4 ) are obtained, specifically as follows:

[0109]

[0110]

[0111]

[0112]

[0113] where \(P(\Theta)\) is the power set of \(\Theta\).

[0114] To deepen the understanding of the fusion of interval evidence through the evidence reasoning rule, an example is given here. Interval evidence can be obtained from steps (1)-(6). Assume the interval evidence is as shown in Table 7. Arbitrarily select 2 points within the generated interval evidence as the point-valued evidence for evidence reasoning rule fusion. Then, there are 8 pieces of point-valued evidence under 4 fault characteristics, specifically as follows:

[0115] Evidence for the 1x frequency amplitude characteristic:

[0116] \(e\) 1,1 = \(\{(X, 0.3953); (Y, 0.3864); (Z, 0.1972); (\Theta, 0.0201)\}\)

[0117] \(e\) 1,2 = \(\{(X, 0.3959); (Y, 0.3872); (Z, 0.1986); (\Theta, 0.0183)\}\)

[0118] Evidence for the 2x frequency amplitude characteristic:

[0119] \(e\) 2,1 = \(\{(X, 0.5446); (Y, 0.2147); (Z, 0.1859); (\Theta, 0.0549)\}\)

[0120] \(e\) 2,2 = \(\{(X, 0.5445); (Y, 0.2127); (Z, 0.1828); (\Theta, 0.0601)\}\)

[0121] Evidence for the 3x frequency amplitude characteristic:

[0122] \(e\) 3,1 = \(\{(X, 0.3526); (Y, 0.2825); (Z, 0.3560); (\Theta, 0.0089)\}\)

[0123] \(e\) 3,2 = \(\{(X, 0.3529); (Y, 0.2821); (Z, 0.3559); (\Theta, 0.0092)\}\)

[0124] Evidence for the average amplitude characteristic of time-domain vibration displacement:

[0125] \(e\) 4,1={(X, 1.0000); (Y, 0.0000); (Z, 0.0000); (Θ, 0.0000)}

[0126] e 4,2 ={(X, 1.0000); (Y, 0.0000); (Z, 0.0000); (Θ, 0.0000)}

[0127] Set the reliability factor r λ and the importance weight w λ Both are 0.9. The evidence under different features is fused according to formulas (16)-(17) for evidence reasoning. That is, each time 4 pieces of evidence participate in the fusion, so there are a total of 2 4 = 16 fusion results. The specific results are shown in Table 8 as follows:

[0128] Table 8 Table of Fusion Results of Evidence Reasoning Rules

[0129]

[0130]

[0131] According to the results fused by the evidence reasoning rules, the maximum and minimum values of the credibility after the fusion of each failure risk mode are selected to form interval evidence. The fusion results are shown in Table 9 as follows:

[0132] Table 9 Table of Interval Evidence Results

[0133]

[0134] According to the identification method described in step (9), the failure risk mode can be determined as X (motor shafting imbalance failure).

[0135] The following combines with the accompanying drawings to introduce in detail the embodiments of the method of the present invention:

[0136] The flow block diagram of the method of the present invention is as Figure 1 shown. The core part is: constructing the template mode cloud model and the to-be-inspected mode cloud model of the failure risk mode; matching the template mode cloud model and the to-be-inspected mode cloud model to obtain the matching degree interval; processing the matching degree interval by the global normalization method to obtain interval evidence; fusing the interval evidence according to the evidence reasoning rules; and identifying the failure risk mode according to the identification criterion based on the fused interval evidence.

[0137] The following combines with Figure 3 the embodiments of the motor shafting failure risk mode identification system in to introduce in detail each step of the method of the present invention.

[0138] 1. Setting Example of Motor Shafting Failure Risk Mode Identification System

[0139] The experimental equipment is the ZHS-2 multi-functional motor flexible shafting system as Figure 3 described in Figure 3 . The vibration acceleration sensors and vibration displacement sensors installed respectively in the horizontal and vertical directions of the shafting support seat continuously collect the time-domain vibration sequences. The signals are transmitted to the computer through the HG-8902 acquisition box. Then, using the HG-8902 data analysis software in the Labview environment, the amplitudes at 1 to 3 times the frequency of the vibration acceleration of the shafting and the average amplitude of the time-domain vibration displacement are obtained, and these are used as the fault characteristics.

[0140] 2. Selection of Fault Risk Modes and Their Characteristic Parameters of the Motor Shafting

[0141] On the test bench, the fault risk modes are set as "X is unbalance in the motor shafting", "Y is misalignment in the motor shafting", and "Z is looseness of the base in the motor shafting", so the fault set is Θ = {X, Y, Z}; the rotational speed of the shafting is set as: 1500 revolutions per minute, the 1st harmonic frequency is 25 Hz, and the nth harmonic frequency is (n×25) Hz. When the shafting operates normally, the amplitude of each vibration acceleration frequency does not exceed 0.1 mm / s 2 . When a fault occurs, the increase in the frequency and its amplitude corresponding to different fault risk modes are also different. According to the analysis, the vibration energy of the 3 fault risk modes mostly concentrates on the 1st to 3rd harmonic frequencies. If only analyzing the amplitude of a single frequency, it is difficult to determine which fault risk has occurred, and the occurrence of a fault is a gradual process. Coupled with the fact that the sensor is easily affected by the working environment, the fault risk result cannot be identified based on a single fault characteristic. Therefore, the amplitudes of the vibration acceleration from the 1st to 3rd harmonic frequencies and the average amplitude of the time-domain vibration displacement (unit: mm) are selected as the fault characteristic parameters.

[0142] 3. Construction of the Template Mode Cloud Model of 3 Fault Risk Modes under 4 Fault Characteristics

[0143] Using the method described in step (2) of the present invention, the 3 fault risk modes of "X", "Y", and "Z" are set on the motor shafting. Through the vibration acceleration sensors and vibration displacement sensors installed respectively in the horizontal and vertical directions of the shafting support seat, with a time interval of Δt = 16 s, the time-domain vibration sequences are continuously collected. The sample data under 4 fault characteristics, namely the amplitudes at 1 to 3 times the frequency of the vibration acceleration and the average amplitude of the time-domain vibration displacement, are obtained through Fourier transform. A total of 200 sample data are collected under each fault characteristic. According to the method described in step (3), the template mode cloud model of 3 fault risk modes under 4 fault characteristics is constructed. The number of cloud droplets for each cloud model is set as 5000, and a total of 12 cloud models need to be established, respectively as Figures 4 - 7As shown, where the "X cloud model", "Y cloud model", and "Z cloud model" respectively represent the template cloud models of three fault risk modes, and the parameter values used to construct the cloud models are shown in Table 10:

[0144] Table 10 Parameter Table for Constructing the Fault Template Mode Cloud Model

[0145]

[0146] 4. Construct the cloud model of the mode to be inspected

[0147] Taking the occurrence of the fault risk mode X as an example, according to the fault feature acquisition method described in step (2), 40 observations are respectively made on the amplitudes at 1-3 times the frequency of the fault feature vibration acceleration and the average amplitude of the time-domain vibration displacement. 40 samples to be inspected are obtained online. According to the method described in step (3), the number of cloud droplets of each cloud model is set to 5000, and 4 cloud models of the modes to be inspected under 4 fault features are constructed as Figures 4 - 7 shown, and its parameters are shown in Table 11:

[0148] Table 11 Parameter Table for Constructing the Cloud Model of the Mode to be Inspected

[0149]

[0150] 5. Obtain the interval evidence according to the cloud models of the mode to be inspected and the fault template mode

[0151] According to the method described in step (6.1), the entropy-containing expectation curves of the fault template mode and the cloud models of the modes to be inspected are obtained. The intersection coordinates of the entropy-containing expectation curves of the cloud models of the modes to be inspected and the fault template mode under different features are shown in Table 12 by using formulas (6) and (7):

[0152] Table 12 Intersection Coordinate Table of the Entropy-containing Expectation Curves of the Cloud Models of the Modes to be Inspected and the Fault Template Mode

[0153]

[0154] Select 200 cloud droplets closest to the intersection point. The determination degrees of these cloud droplets are the matching degrees of the cloud model C of the mode to be inspected i and the template mode cloud model . Take the minimum value and the maximum value to form the matching degree interval as shown in Table 13:

[0155] Table 13 Matching Degree Interval Table

[0156]

[0157] Normalize the obtained matching degree interval globally according to formulas (10)-(15) to obtain the interval evidence as shown in Table 14:

[0158] Table 14 Interval Evidence Table

[0159]

[0160] 6. Use the evidence reasoning rule to fuse the obtained interval evidence to obtain the fused interval evidence

[0161] According to steps (7) to (8) of the method of the present invention, any 2 points are taken from the generated interval evidence as the point-value evidence for the evidence reasoning rule fusion. Then there are 8 pieces of point-value evidence under 4 fault characteristics, specifically as follows:

[0162] Evidence under the 1x frequency characteristic:

[0163] e 1,1 ={(X, 0.3957); (Y, 0.3867); (Z, 0.1978); (Θ, 0.0198)}

[0164] e 1,2 ={(X, 0.4065); (Y, 0.3872); (Z, 0.2001); (Θ, 0.0062)}

[0165] Evidence under the 2x frequency characteristic:

[0166] e 2,1 ={(X, 0.5446); (Y, 0.2147); (Z, 0.1859); (Θ, 0.0549)}

[0167] e 2,2 ={(X, 0.5656); (Y, 0.2421); (Z, 0.1903); (Θ, 0.0017)}

[0168] Evidence under the 3x frequency characteristic:

[0169] e 3,1 ={(X, 0.3526); (Y, 0.2825); (Z, 0.3560); (Θ, 0.0089)}

[0170] e 3,2 ={(X, 0.3585); (Y, 0.2889); (Z, 0.3520); (Θ, 0.0006)}

[0171] Evidence under the average amplitude characteristic of the time-domain vibration displacement:

[0172] e 4,1 ={(X, 1.0000); (Y, 0.0000); (Z, 0.0000); (Θ, 0.0000)}

[0173] e4,2 = {(X, 1.0000); (Y, 0.0000); (Z, 0.0000); (Θ, 0.0000)}

[0174] Set the reliability factor r λ and the importance weight w λ both to 0.9. Evidence reasoning fusion is performed on the evidence under different characteristics according to formulas (16)-(17). That is, each time 4 pieces of evidence participate in the fusion, so there are a total of 2 4 = 16 fusion results. According to the results fused by the evidence reasoning rule, the maximum and minimum values of the belief degree after fusion of each failure risk mode are selected to form interval evidence, and the final fusion results are shown in Table 15:

[0175] Table 15 Evidence interval table after fusion

[0176]

[0177] 7. Identify the failure risk mode according to the fusion result

[0178] According to the identification criterion in step (9) of the method of the present invention, it can be determined that the failure risk of the motor shafting is "X (unbalance of the motor shafting)", which is consistent with the actual situation.

Claims

1. A method for identifying the fault risk mode of a motor shafting based on interval evidence fusion of a normal cloud model, characterized in that The method comprises the following steps: (1) Assume that the shaft system speed is ρ rpm, 1000≤ρ≤2500, and let the fault set composed of the fault risk mode of the shaft system be Θ={X,Y,Z}, where X represents the unbalance fault in the motor shaft system, Y represents the misalignment fault in the motor shaft system, and Z represents the base loose fault in the motor shaft system; (2) Vibration acceleration sensors and vibration displacement sensors installed in the horizontal and vertical directions of the shaft support seat are used to continuously collect time domain vibration sequences at set time intervals. The amplitudes of vibration acceleration at 1 to 3 times the frequency and the average amplitude of vibration displacement in the time domain are obtained through Fourier transform, which are used as fault characteristics; Set the shafting system to be in the fault risk mode U, where U ∈ {X, Y, Z}, and sequentially collect the data of four fault characteristics: the amplitudes of the 1x, 2x, and 3x vibration accelerations and the average amplitude of the time-domain vibration displacement, denoted as F U , where i = 1, 2, 3, 4 sequentially represent the above four fault characteristics, N represents the number of collected data samples, and 50 ≤ N ≤ 500; (3) Set the sample mode cloud model of the failure risk mode U under the i-th failure feature as The cloud droplets constituting the sample mode cloud model are denoted as where g = 1, 2, …, G, G represents the total number of generated cloud droplets, 1000 ≤ G ≤ 5000, and construct the sample mode cloud models of 3 failure risk modes under 4 failure features; (4) Obtain a set of sample data to be inspected online, denoted as F i , where N1 ≤ N, representing the total number of online data samples; (5) Use F i to replace the data of the failure risk mode U under the failure feature i in step (3), and then repeat the entire calculation process of step (3) to obtain the cloud models of the patterns to be inspected under 4 failure features: {C1, C2, C3, C4}, C i = {d i,1 ,..., d i,g ,..., d i,G |d i,g = {c i,g , μ(c i,g )}}; (6) Calculate the matching degree between the template pattern cloud model and the cloud model C of the pattern to be inspected i , and then obtain the interval evidence i supported by the cloud model C of the pattern to be inspected for the occurrence of three failure risk patterns; (7)Fuse the interval evidence obtained in step (6); using the evidential reasoning rule. (8) Obtain H according to step (7). 4 The fusion result for the Μ(θ) = [a θ , b θ (1) According to equations (1)-(3), the interval evidence of the fused normal cloud model can be obtained as follows: Interval evidence: Μ(X), corresponding fusion result: [a X , b X ; Interval evidence: Μ(Y), corresponding fusion result: [a Y , b Y ; Interval evidence: Μ(Y), corresponding fusion result: [a Y , b Y ; Interval evidence: Μ(Θ), corresponding fusion result: [a Y ,b Y ; (9)Give the identification criterion for the fault risk mode according to the interval evidence, that is, it can be determined that the fault risk mode pointed to by the sample F to be tested meets the following two conditions: i is θ, θ ∈ {X, Y, Z, Θ}: ① The left and right endpoints of the M(θ) interval evidence are respectively greater than the left and right endpoints of the interval evidence corresponding to other fault risk modes, θ≠Θ; ②The right endpoint of Μ(Θ) is less than 0.

3.

2. The method for identifying the fault risk mode of a motor shafting based on interval evidence fusion of a normal cloud model according to claim 1, characterized in that: The 1st frequency is (ρ / 60) Hz, the 2nd frequency is (ρ / 30) Hz, and the 3rd frequency is (ρ / 15) Hz.

3. The method for identifying the fault risk mode of a motor shafting based on interval evidence fusion of a normal cloud model according to claim 1, characterized in that: The vibration acceleration sensor and the vibration displacement sensor continuously collect the time domain vibration sequence with a time interval of △t=16s.

4. The method for identifying the fault risk mode of a motor shafting based on interval evidence fusion of a normal cloud model according to claim 1, characterized in that: Step (3) is specifically: (3.1) The data F of the failure risk mode U under the failure feature i can be obtained from step (2). U , Take its arithmetic mean as the expected value EX of the cloud model. i,U (3.2) Setting has an entropy of En, where 0.05 ≤ En ≤ 0.1, and a hyperentropy value of He, where 0.005 ≤ He ≤ 0.01, and (3.3) Generate a normal random number with entropy En as the expectation and hyperentropy He as the standard deviation Then, with EX i,U as the expectation, as the standard deviation, generate a normal random number as the cloud droplet (3.4) Calculate cloud droplets The corresponding degree of certainty And the degree of certainty Has the property of probabilistic likelihood, and the specific calculation method is as follows: (3.5) Let reflect the position information of cloud droplets in two-dimensional space. Repeat the calculation process of steps (3.3)-(3.4) G times until G cloud droplets and their respective degrees of certainty are generated, and then the template mode cloud model of the fault risk mode U under the fault feature i can be obtained. Then, for 3 fault risk modes under 4 fault features, a total of 4×3 = 12 template mode cloud models can be generated: The sample mode cloud models of three fault risk modes under the 1-fold frequency amplitude characteristic are as follows: The sample mode cloud models of three fault risk modes under the 2-fold frequency amplitude characteristic are as follows: The sample mode cloud models of three fault risk modes under the 3-fold frequency amplitude characteristics are as follows: And the sample mode cloud models of the three fault risk modes under the characteristics of the average amplitude of vibration displacement in the time domain are as follows:

5. The method for identifying the fault risk mode of the motor shafting based on the interval evidence fusion of the normal cloud model according to claim 4, wherein: Step (6) is specifically: (6.1) Construct the sample pattern cloud model The entropy-containing expected curve of is the cloud model C of the pattern to be inspected i The entropy-containing expected curve of where and x i are random variables and follow a Gaussian distribution (6.2) Under the fault feature i, obtain the intersection points of the V i (x i ) of the three template mode cloud models and the cloud model of the pattern to be inspected The coordinates in the two-dimensional space are The specific calculation method is as follows: (6.3) From step (3.5), the coordinates of the cloud droplets of the three template mode cloud models in the two-dimensional space under the fault feature i are Set as the K cloud droplets closest to the intersection point , and 100 ≤ K ≤ 250. The degrees of certainty μ(c i,k ), μ(c i,k ) ∈ [0, 1] of these cloud droplets are the matching degrees between the cloud model C i to be detected and the template mode cloud model , and form a matching degree set Take the minimum value as a i,U , the maximum value as b i,U , to form a matching degree interval I i (U) = [a i,U , b i,U (8) (6.4) Normalize I i (U) to obtain the normalized matching degree interval: (6.5) The confidence interval of the complete set of faults Θ = {X, Y, Z} is calculated as Μ i (Θ) measures the overall uncertainty of the match in step (6.4); (6.6) The normal cloud model interval evidence is constructed from equations (9) and (13), as follows: Interval evidence: fundamental frequency amplitude, Μ i (X) corresponds to Μ i (Y) corresponds to Μ i (Z) corresponds to Μ i (Θ) corresponds to Interval evidence: 2x frequency amplitude, Μ i (X) corresponds to Μ i (Y) corresponds to Μ i (Z) corresponds to Μ i (Θ) corresponds to Interval evidence: 3x frequency amplitude, Μ i (X) corresponds to Μ i (Y) corresponds to Μ i (Z) corresponds to Μ i (Θ) corresponds to Interval evidence: average amplitude of vibration displacement, Μ i (X) corresponds to Μ i (Y) corresponds to Μ i (Z) corresponds to Μ i (Θ) corresponds to 6. The method for identifying the fault risk mode of the motor shafting based on the interval evidence fusion of the normal cloud model according to claim 5, wherein: Step (7) is specifically: (7.1) Perform H samplings within the interval respectively, where H ≥ 2. Denote the sampled point values as and they satisfy the following conditions: ① ② (7.2) According to the method described in step (7.1), H point-value evidences under the fault feature i can be obtained, denoted as E i ={e i,1 ,...,e i,h ,...,e i,H}, where Set the reliability factor of evidence e i,h as r λ , satisfying 0 ≤ r λ ≤ 1, λ = 1, 2, 3, 4, and the importance weight w λ = r λ ; (7.4) Four H point value type evidences under four fault characteristics can be obtained through step (7.2). Select any one point value type evidence under each fault characteristic for combination, and each combination contains point value type evidences of four fault characteristics. Then there are a total of H 4 combinations. The four point value type evidences in each combination are fused using the evidence reasoning rule to obtain l fusion results, where l = 1, 2,..., H 4 , specifically as follows: Where P(Θ) is the power set of Θ.

Images