A method for diagnosing a fault of a gyro motor bearing

Abstract

The present application relates to a gyro motor bearing, in particular to a gyro motor bearing fault diagnosis method based on electric signal, which is used to solve the problem that the existing bearing fault diagnosis method is based on vibration signal, which has the disadvantages of high extraction cost, difficulty in batch signal collection, and lack of iterative optimization learning ability. The gyro motor bearing fault diagnosis method identifies sensitive characteristic parameters through effective electric signal, and respectively establishes HMM qua and HMM fau based on hidden Markov model (HMM) for the qualified state and the fault state of the bearing of the gyro motor, and further diagnoses the bearing fault of the gyro motor; the present application can assist the technical personnel to judge the performance of the product, eliminate unqualified products early, avoid waste caused by transfer to subsequent processes, and further provide data support for qualified products.

Description

Technical Field

[0001] The invention relates to a gyro motor bearing, and in particular to a gyro motor bearing fault diagnosis method based on electrical signals. Background Art

[0002] Gyro motors are widely used in aerospace and other fields, assembling liquid-floating, semi-liquid-floating, flexible, and torsion-bar rate gyros. Their rotors maintain high-speed, constant rotation, providing the gyro with constant angular momentum. Bearings are key mechanical components in gyro motors, and failures in them are crucial to the gyro motor and, ultimately, the gyro itself. Abnormal impact or friction in the bearings can lead to premature failure, directly impacting the gyro's operating life and posing a potential risk. Therefore, gyro motor bearing fault diagnosis technology is a crucial research topic for ensuring product quality and reliability.

[0003] Existing gyro motor bearing fault diagnosis technologies mostly use vibration signals. However, extracting vibration signals requires the purchase of specialized vibration signal measurement equipment and the design and commissioning of high-precision vibration signal measurement process equipment suitable for specific motor models. Sound and electrical signals, two emerging signals used in bearing fault diagnosis, have been initially applied in engineering practice. However, accurate acquisition of sound signals requires the construction of a dedicated soundproof room, which increases the difficulty and cost of bearing fault diagnosis. Sound signals are also easily affected by background noise, resulting in very limited fault diagnosis accuracy. In comparison, electrical signals, as a high-efficiency signal with no additional load and low cost, only require a power meter to collect the motor's current and power for subsequent analysis, making them more convenient and quicker.

[0004] The Hidden Markov Model (HMM) is a tool for identifying and classifying "process states." It has been applied to various fields of natural science, social science, and cognitive science, achieving promising results. Its pattern recognition capabilities can be used to identify qualified and faulty bearings as two distinct patterns, thus enabling fault diagnosis.

[0005] Chinese patent CN201610456903.X discloses a rolling bearing fault pattern recognition method and device based on stacked convolutional networks. Figure 1The method extracts time-frequency domain features of a vibration signal of a known state rolling bearing; the obtained time-frequency domain features of the known state rolling bearing are normalized into feature map elements conforming to an input format of a CNN network; the feature map elements are input into the CNN network, and model parameters of the CNN network are adjusted through forward self-learning and gradient descent-based back propagation processing in the CNN network, so that a trained CNN network is obtained; when identifying a fault mode of an actual rolling bearing, time-frequency domain features of a vibration signal of the unknown state rolling bearing are taken as input, and the trained CNN network is used to extract high-order features reflecting intrinsic information layer by layer, and the self-learned results of the features layer by layer are input into a top classifier, so that the fault mode recognition of the rolling bearing under multiple working conditions and strong noise is realized.

[0006] The bearing fault diagnosis signal used by the method is a vibration signal, which is not only high in extraction cost, but also difficult to realize batch signal collection (more than 20 pieces). The method for bearing fault diagnosis relying on the CNN network directly uses the bearing data of the University of Western Michigan for verification in the specific example description, which can only indicate that it has a certain effect in the fault diagnosis of the specific single bearing. When the bearing is assembled in a specific product, especially in a precision instrument such as a gyroscope, the effectiveness of the method needs to be further verified. The method described in the patent has no iterative optimization learning ability, and the accuracy of fault diagnosis is directly related to whether the model parameters of the CNN network are reasonably set, and the specific values of the parameters need a large amount of engineering practice results to support to ensure their rationality. SUMMARY

[0007] The purpose of the present application is to solve the problems of the prior art bearing fault diagnosis method based on vibration signals, which is not only high in extraction cost and difficult to realize batch signal collection, but also has no iterative optimization learning ability, and to provide a gyro motor bearing fault diagnosis method.

[0008] In order to solve the above-mentioned problems of the prior art, the present application provides the following technical solutions:

[0009] A gyro motor bearing fault diagnosis method, characterized in that it comprises the following steps:

[0010] Step 1, collect current signals and power signals of multiple gyro motors of the same type, and identify effective electrical signals through analysis;

[0011] The multiple gyro motors are divided into two states: qualified bearings and bearing faults. The qualified bearings account for 45% to 55%. The qualified bearings state means that the gyro assembled by the gyro motor meets the index requirements and has no abnormal noise, and no motor bearing-related faults occur during the subsequent gyro assembly, commissioning, acceptance delivery and the user's full life cycle. The bearing fault state means that the relevant technical indicators of the gyro motor exceed the tolerance and abnormal noise occurs during operation.

[0012] The effective electrical signal includes an effective current signal and an effective power signal. The effective current signal is the current signal from the current signal entering a stable state to before power failure. The effective power signal is the power signal from the power signal entering a stable state to before power failure.

[0013] Step 2: Extract characteristic parameters from the effective electrical signal identified in step 1 one by one, and compare and analyze the characteristic parameters corresponding to the qualified state of the bearing and the characteristic parameters corresponding to the faulty state of the bearing. If a characteristic parameter has a significant difference, identify the characteristic parameter as a sensitive characteristic parameter; the characteristic parameters include mean value, standard deviation, crest ratio, kurtosis, skewness, and spectral density;

[0014] Step 3: Adaptively design the HMM based on the number of effective electrical signals, the number of sensitive characteristic parameters, and the characteristic factors of the sensitive characteristic parameters, and implement the algorithm function through MATLAB;

[0015] The designed HMM is used to model the qualified state and fault state of the gyro motor bearing, which are defined as HMM qua and HMM fau ; The HMM qua and HMM fau The dimension is the same as the number D of sensitive feature parameters identified in step 2;

[0016] Step 4: According to the method described in step 1 and step 2, collect the effective electrical signal of the gyro motor to be diagnosed in the same model as step 1, extract the sensitive characteristic parameters, and use the HMM established in step 4 qua and HMM fau Perform fault diagnosis and finally obtain the fault diagnosis result of the gyro motor.

[0017] Furthermore, the step 1 is specifically as follows:

[0018] Step 1.1: Collect current signals and power signals from multiple gyro motors of the same model, where the number of the multiple gyro motors is not less than 30; the current signals and power signals cover the startup phase, the working phase, and the stop phase of the gyro motor, and the collection time is not less than 40 minutes;

[0019] Step 1.2: Smoothing the current signal and power signal collected in step 1.1, dividing the current signal and power signal into multiple equal parts according to time, obtaining the average value of the current signal and the average value of the power signal in each equal part, and making a time-current signal average value line graph and a time-power signal average value line graph;

[0020] Step 1.3, based on the time-current signal average value line graph and the time-power signal average value line graph obtained in step 1.2, respectively obtain the corresponding time when each gyro motor current signal and power signal enters a stable state;

[0021] Step 1.4: Since the time it takes for the collected motor electrical signals to enter a stable state conforms to the normal distribution, the average time required for the current signal and power signal of all gyro motors to enter a stable state is calculated. and standard deviation σ, then we assume that The time required for the same model of gyro motor to enter a stable state;

[0022] Step 1.5: Extract all gyro motors The current signal and power signal before power failure are regarded as effective electrical signals.

[0023] Furthermore, the step 3 is specifically as follows:

[0024] Step 3.1, extracting sensitive characteristic parameters corresponding to the gyro motor in the bearing qualified state and the bearing fault state from the effective electrical signal;

[0025] Step 3.2: normalize each sensitive feature parameter;

[0026] Step 3.3: Initialize HMM qua and HMM fau Parameters Π of each dimension z 、A z 、B z , z = 0;

[0027] Among them, Π z is the hidden state probability matrix after z iterations; A z is the state transfer matrix after z iterations; B z The observation state generation matrix after z iterations; z is the number of iterations;

[0028] Step 3.4, according to π z 、A z 、B z The value of each element in , calculate HMM qua and HMM fau The forward probability α at each time point t in each dimension t ;

[0029] Step 3.5: According to π z 、A z 、B z The value of each element in , calculate HMM qua and HMM fau The backward probability β at each time point t in each dimension t ;

[0030] Step 3.6, according to the forward probability α t , backward probability β t , calculate HMM qua and HMM fau The probability γ that the hidden state is i at time t in each dimension t (i);

[0031] Here, i represents the hidden state of the gyro motor, which includes two states: qualified and faulty.

[0032] Step 3.7, according to the forward probability α t , backward probability β t , calculate HMM qua and HMM fau The probability that the hidden state at time t is i and the hidden state at time t+1 is j in each dimension is ξ t (i,j);

[0033] Among them, i and j represent the hidden status of the gyro motor, which includes qualified and faulty states;

[0034] Step 3.8, according to the probability γ t (i) Probability ξ t (i,j), recalculate HMM qua and HMM fau Parameters Π of each dimension z 、A z 、B z , and let z=z+1;

[0035] Step 3.9: Calculate HMM qua and HMM fau The corresponding π z and Π z-1 , A z and A z-1 , B z and B z-1 The difference between the elements at each corresponding position in the algorithm is defined as the threshold indicator for termination of the algorithm. When the absolute value of any of the differences is greater than ε, return to step 3.4. When the absolute value of the differences is not greater than ε, the HMM is obtained. qua and HMM fau Parameters Π of each dimensionz 、A z 、B z , that is, complete HMM qua and HMM fau The establishment of.

[0036] Furthermore, the step 4 is specifically as follows:

[0037] Step 4.1, according to the method described in step 1, collect the effective electrical signal of the gyro motor to be diagnosed of the same model as in step 1;

[0038] Step 4.2: Extract sensitive characteristic parameters from the effective electrical signal collected in step 4.1 according to the method described in step 2, and perform normalization processing on each sensitive characteristic parameter;

[0039] Step 4.3, calculate the sensitive feature parameters in HMM qua and HMM fau The forward probability at each time point t is

[0040] Step 4.4: Calculate the sensitive feature parameters in HMM qua and HMM fau The probability of appearing in

[0041] Step 4.5: Determine the weight coefficient of each sensitive feature parameter as ρ d , d=1,2,…,D;

[0042] Step 4.6: Compare the following equations: If so, the bearing of the gyro motor is diagnosed as qualified; if not, the bearing of the gyro motor is diagnosed as faulty.

[0043] Furthermore, the method further includes step 5, optimizing and adjusting HMM qua and HMM fau The weight coefficient of

[0044] Step 5.1, conduct a comprehensive assessment of the gyro motor that has completed step 4;

[0045] Step 5.1.1: Perform a first-level judgment on the gyro motor that has completed step 4. If the gyro motor's indicators do not meet the requirements or there is an abnormal sound, the gyro motor has a first-level fault and proceed to step 5.1.2; otherwise, proceed to step 5.1.3;

[0046] Step 5.1.2: Disassemble the gyro motor for secondary diagnosis. If any of the gyro motor bearing parts (outer ring raceway, inner ring roller, roller surface, cage surface) are found to be abnormally worn under microscope observation, the gyro motor is comprehensively diagnosed as faulty. Otherwise, proceed to step 5.1.3.

[0047] Step 5.1.3: The gyro assembled with the gyro motor undergoes subsequent production processes. After all indicators are inspected and qualified, the gyro is delivered to the user for use. If a fault related to the motor bearing occurs during use, the gyro motor is comprehensively judged to be faulty; otherwise, the gyro motor is comprehensively judged to be qualified.

[0048] Step 5.2: Count the diagnostic results of the gyro motor in step 4 and the comprehensive judgment results in step 5.1. The diagnostic results include the product number and the D sensitive feature parameters obtained in step 4 in the HMM. qua and HMM fau The probability of appearing in Where d = 1, 2, ..., D;

[0049] Step 5.3: For each gyro motor counted in step 5.2, compare them one by one The size of The diagnosis result of the sensitive characteristic parameter item d is qualified; if Then the diagnosis result of the sensitive characteristic parameter item d is fault;

[0050] Step 5.4: If the diagnosis result of the dth sensitive characteristic parameter obtained in step 5.3 is consistent with the comprehensive judgment result of the motor bearing, then the sensitive characteristic parameter is scored as 1; otherwise, the sensitive characteristic parameter is scored as 0;

[0051] Step 5.5: Calculate the scores of D sensitive feature parameters in turn, where the score of the dth sensitive feature parameter is G d , where d = 1, 2, ..., D;

[0052] Step 5.6: Calculate the weight coefficients of D sensitive characteristic parameters in turn, and obtain the weight coefficients of each sensitive characteristic parameter as follows: Where d = 1, 2, ..., D, which means the optimization adjustment of the weight coefficient is completed;

[0053] If the diagnosis of all gyro motors to be diagnosed has been completed, the process ends; otherwise, the process returns to step 4.

[0054] Furthermore, the step 2 is specifically as follows:

[0055] Step 2.1, the effective electrical signal identified in step 1 is divided into segments according to time, various characteristic parameters are extracted from each segment of the effective electrical signal, and a time-characteristic parameter line graph is drawn. The time-characteristic parameter line graph is divided into a bearing qualified state group and a bearing fault state group;

[0056] Step 2.2: Compare the qualified bearing state group and the faulty bearing state group. If a certain characteristic parameter has a significant difference, identify the characteristic parameter as a sensitive characteristic parameter. The significant difference is defined as a characteristic parameter of the qualified bearing state group being stable or concentrated near a certain value in multiple gyro motors, and a characteristic parameter of the faulty bearing state group being abrupt or concentrated near another different value in multiple gyro motors.

[0057] Furthermore, the step 3.2 is specifically as follows: performing local normalization on the average value and the spectral density as a type of sensitive characteristic parameter, wherein the local normalization is to compare the sensitive characteristic parameter only with itself;

[0058] The standard deviation, kurtosis, crest ratio, and skewness are used as the second-category sensitive feature parameters for global normalization, and the global normalization is the comparison of this sensitive feature parameter with all the second-category sensitive feature parameters.

[0059] Compared with the prior art, the present invention has the following beneficial effects:

[0060] (1) A gyro motor bearing fault diagnosis method of the present invention identifies sensitive characteristic parameters through effective electrical signals, and establishes HMMs for the gyro motor bearing qualified state and bearing fault state respectively based on the hidden Markov model (HMM). qua and HMM fau , and then diagnose the bearing fault of the gyro motor; the present invention can assist technical personnel in judging the performance of the product, eliminate unqualified products in advance, avoid waste caused by transferring to subsequent processes, and further provide data support for qualified products.

[0061] (2) The present invention provides a method for diagnosing gyro motor bearing faults by adaptively designing the HMM, specifically including normalization design, multi-objective design, weight design, etc., so that the established HMM can meet the requirements of diagnosing the faults of gyro motor bearings of specific models.

[0062] (3) The present invention provides a method for diagnosing gyro motor bearing faults with an iterative optimization function for weight coefficients. The method statistically analyzes actual production data of a specific model of gyro motor and feeds back the data to the weight coefficient setting step, thereby adjusting the weight coefficient to achieve the optimal diagnostic accuracy.

[0063] (4) The present invention defines a "valid electrical signal" concept for a gyro motor bearing fault diagnosis method. A test is designed to accurately identify the time required for a specific model of gyro motor to reach a stable state from an unstable state. The current and power signals of the gyro motor in the stable state are defined as "valid electrical signals." Using valid electrical signals for bearing fault diagnosis can essentially eliminate other interfering factors and improve the accuracy of fault diagnosis.

[0064] (5) The present invention defines a "sensitive characteristic parameter" concept for a gyro motor bearing fault diagnosis method. A test is designed to accurately identify characteristic parameters (such as the mean and standard deviation of the electrical signal) that significantly change when a bearing fault occurs in a specific model of gyro motor. These parameters are referred to as "sensitive characteristic parameters." Using sensitive characteristic parameters for bearing fault diagnosis can substantially avoid misjudgments due to characteristic parameters failing to accurately reflect bearing status, thereby improving the accuracy of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A flowchart of an existing rolling bearing fault pattern recognition method;

[0066] Figure 2 This is a flow chart of an embodiment of a method for diagnosing a gyro motor bearing fault according to the present invention;

[0067] Figure 3 This is a flowchart of step 1 of an embodiment of the present invention;

[0068] Figure 4 for Figure 3 Flowchart of step 2 of the embodiment;

[0069] Figure 5 for Figure 3 Flowchart of step 3 of the embodiment;

[0070] Figure 6 for Figure 3 Flowchart of step 4 of the embodiment;

[0071] Figure 7 for Figure 3 Flowchart of Example step 5.1;

[0072] Figure 8 for Figure 3 A line graph showing the average value of the time-current signal of a gyro motor in a qualified bearing state obtained in step 1.2 of the embodiment;

[0073] Figure 9 for Figure 3 A line graph showing the average value of the time-current signal of a gyro motor in a bearing fault state obtained in step 1.2 of the embodiment;

[0074] Figure 10 for Figure 3 A line graph showing the standard deviation of the time-current signal of a gyro motor in a qualified bearing condition obtained in step 2.1 of the embodiment;

[0075] Figure 11 for Figure 3 A line graph showing the standard deviation of the time-current signal of a gyro motor in a bearing fault state obtained in step 2.1 of the embodiment;

[0076] Figure 12 for Figure 3 A line graph showing batch group numbers versus weight coefficients in the embodiment. DETAILED DESCRIPTION

[0077] The present invention will be further described below with reference to the accompanying drawings and exemplary embodiments.

[0078] A gyro motor bearing fault diagnosis method comprises the following steps:

[0079] Step 1: Collect the current signal and power signal of 30 gyro motors of the same model, and identify the effective electrical signal through analysis, such as Figure 3 As shown;

[0080] The multiple gyro motors are divided into two states: qualified bearings and bearing faults. The qualified bearings account for 45% to 55%. The qualified bearings state means that the gyro assembled by the gyro motor meets the index requirements and has no abnormal noise, and no motor bearing-related faults occur during the subsequent gyro assembly, commissioning, acceptance delivery and user use throughout its life. The bearing fault state means that during the operation of the gyro motor, relevant technical indicators exceed the tolerance and abnormal noise occurs; the relevant technical indicators include inertia time and friction torque;

[0081] The effective electrical signal includes an effective current signal and an effective power signal. The effective current signal is the current signal from the current signal entering a stable state to before power failure. The effective power signal is the power signal from the power signal entering a stable state to before power failure.

[0082] Step 1.1, collecting current signals and power signals of 30 gyro motors of the same model, wherein the current signals and power signals cover the startup phase, the working phase, and the stop phase of the gyro motors, the collection time is 40 minutes, and the sampling frequency is 100 ms;

[0083] Step 1.2: Smoothing the current signal and power signal collected in step 1.1, dividing the current signal and power signal into 40 equal parts according to time, obtaining the average value of the current signal and the average value of the power signal in each equal part, and making a time-current signal average value line graph and a time-power signal average value line graph;

[0084] The time-current signal average value line graph of a gyro motor in a qualified bearing state is as follows: Figure 8 As shown in the figure, the average value of the time-current signal of a gyro motor in a bearing fault state is shown in the figure. Figure 9 As shown;

[0085] Step 1.3, based on the time-current signal average value line graph and the time-power signal average value line graph obtained in step 1.2, respectively obtain the corresponding time when each gyro motor current signal and power signal enters a stable state;

[0086] Step 1.4: Since the time it takes for the collected motor electrical signals to enter a stable state conforms to the normal distribution, the average time required for the current signal and power signal of all gyro motors to enter a stable state is calculated. and standard deviation σ, then we assume that is the time required for the same type of gyro motor to enter a stable state; in this embodiment 15 minutes;

[0087] Step 1.5: Extract all gyro motors The current signal and power signal before power failure are regarded as effective electrical signals;

[0088] Step 2: Reference Figure 4 , extracting characteristic parameters from the effective electrical signal identified in step 1 one by one, and identifying sensitive characteristic parameters by comparing and analyzing the characteristic parameters corresponding to the qualified state of the bearing and the characteristic parameters corresponding to the faulty state of the bearing; the characteristic parameters include mean value, standard deviation, crest ratio, kurtosis, skewness, and spectral density;

[0089] Theoretical analysis believes that:

[0090] The corresponding friction torque of a bearing in a qualified state is small, the corresponding current average is small, and it is far away from the starting voltage; the gyro motor has stable operating performance, the current is relatively stable and concentrated, and the standard deviation is small; the gyro motor will not have instantaneous impact vibration during operation, the corresponding current is relatively stable, and the crest ratio of continuous time periods is a stable curve with a small variation amplitude; in an ideal model where the stator current change is only affected by the single factor of bearing operation vibration, when the bearing is operating normally, the distribution of the measured stator current value should be close to the normal distribution, with a kurtosis of approximately 3 and a skewness of 0; there is no abnormal vibration, the energy at each frequency is stable, and the sum of its absolute values ​​is small and stable;

[0091] A bearing in a bearing failure state has abnormal friction, a large friction torque, and may even be unable to reach the synchronous speed. The average current is large and close to the starting current. The gyro motor has unstable operating performance, with a relatively discrete current and a large standard deviation. The bearing will cause instantaneous impact vibration during operation, which will cause large spikes in the stator current and a large curve variation. The current will produce convex jumps and spikes, and the kurtosis should be greater than 3. The instantaneous current greater than the average should be greater than the instantaneous current less than the average, so the skewness should be greater than 0. Due to abnormal vibration, the energy at certain frequencies will have sudden changes, suddenly increasing and decreasing, and the sum of its absolute values ​​is large and has sudden changes.

[0092] Step 2.1, the effective electrical signal identified in step 1 is divided into segments according to time, various characteristic parameters are extracted from each segment of the effective electrical signal, and a time-characteristic parameter line graph is drawn. The time-characteristic parameter line graph is divided into a bearing qualified state group and a bearing fault state group;

[0093] The standard deviation line graph of the time-current signal of a gyro motor with a qualified bearing is as follows: Figure 10 As shown in the figure, the standard deviation line graph of the time-current signal of a gyro motor in a bearing fault state is as follows: Figure 11 As shown;

[0094] Step 2.2: Compare the qualified bearing state group and the faulty bearing state group. If a characteristic parameter has a significant difference, for example, a characteristic parameter of the qualified bearing state group is stable or concentrated near a certain value in multiple gyro motors, while a characteristic parameter of the faulty bearing state group is abruptly changed or concentrated near another different value in multiple gyro motors, then the characteristic parameter is identified as a sensitive characteristic parameter.

[0095] Step 3, reference Figure 5 ,According to factors such as the number of effective electrical signals, the number of sensitive characteristic parameters, and the characteristics of sensitive characteristic parameters (such as dimensionless or dimensional), the hidden Markov model (HMM) is adaptively designed and the algorithm function is implemented through MATLAB;

[0096] The designed HMM is used to model the qualified state and fault state of the gyro motor bearing, which are defined as HMM qua and HMM fau ; The HMM qua and HMM fau The dimension of and the number of sensitive feature parameters D identified in step 2 are both 4;

[0097] Step 3.1, extracting sensitive characteristic parameters corresponding to the gyro motor in the bearing qualified state and the bearing fault state from the effective electrical signal;

[0098] Step 3.2: Normalize each sensitive characteristic parameter. Use the mean value and spectral density as the first-class sensitive characteristic parameters for local normalization. Local normalization means that the sensitive characteristic parameter is only compared with itself. Use the standard deviation, kurtosis, crest ratio, and skewness as the second-class sensitive characteristic parameters for global normalization. Global normalization means that the sensitive characteristic parameter is compared with all the second-class sensitive characteristic parameters.

[0099] Step 3.3: Initialize HMM qua and HMM fau Parameters Π of each dimension z 、A z 、B z , z = 0;

[0100] Among them, Π z is the hidden state probability matrix after z iterations; A z is the state transfer matrix after z iterations; B z The observation state generation matrix after z iterations; z is the number of iterations;

[0101] Step 3.4, according to π z 、A z 、B z The value of each element in , calculate HMM qua and HMM fau The forward probability α at each time point t in each dimension t ;

[0102] Step 3.5: According to π z 、A z 、B z The value of each element in , calculate HMM qua and HMM fau The backward probability β at each time point t in each dimension t ;

[0103] Step 3.6, according to the forward probability α t , backward probability β t , calculate HMM qua and HMM fau The probability γ that the hidden state is i at time t in each dimension t (i);

[0104] Here, i represents the hidden state of the gyro motor, which includes two states: qualified and faulty.

[0105] Step 3.7, according to the forward probability α t , backward probability β t , calculate HMM qua and HMM fauThe probability that the hidden state at time t is i and the hidden state at time t+1 is j in each dimension is ξ t (i,j);

[0106] Among them, i and j represent the hidden status of the gyro motor, which includes qualified and faulty states;

[0107] Step 3.8, according to the probability γ t (i) Probability ξ t (i,j), recalculate HMM qua and HMM fau Parameters Π of each dimension z 、A z 、B z , and let z=z+1;

[0108] Step 3.9: Calculate HMM qua and HMM fau The corresponding π z and Π z-1 , A z and A z-1 , B z and B z-1 The difference between the elements at each corresponding position in the algorithm is defined as the threshold indicator for termination of the algorithm. When the absolute value of any of the differences is greater than ε, return to step 3.4. When the absolute value of the differences is not greater than ε, the HMM is obtained. qua and HMM fau Parameters Π of each dimension z 、A z 、B z , that is, complete HMM qua and HMM fau the establishment of

[0109] In this embodiment, HMM qua for:

[0110] HMM qua d-dimensional hidden state probability matrix; HMM qua d-dimensional state transition matrix; B (d) HMM qua The observation state generation matrix of d dimensions;

[0111] When d=1, the corresponding sensitive characteristic parameter is the standard deviation.

[0112] When d=2, the corresponding sensitive characteristic parameter is the peak ratio.

[0113] When d=3, the corresponding sensitive characteristic parameter is kurtosis.

[0114] When d=4, the corresponding sensitive characteristic parameter is skewness.

[0115] HMM fau for:

[0116] HMM qua The hidden state probability matrix of dimension d; HMM qua d-dimensional state transition matrix; HMM qua The observation state generation matrix of d dimensions;

[0117] When d=1, the corresponding sensitive characteristic parameter is the standard deviation.

[0118] When d=2, the corresponding sensitive characteristic parameter is the peak ratio.

[0119] When d=3, the corresponding sensitive characteristic parameter is kurtosis.

[0120] When d=4, the corresponding sensitive characteristic parameter is skewness.

[0121] Step 4: Reference Figure 6 According to the method described in step 1 and step 2, the effective electrical signal of the gyro motor of the same model to be diagnosed in step 1 is collected, the sensitive characteristic parameters are extracted, and the HMM established in step 4 is used. qua and HMM fau Perform fault diagnosis;

[0122] Step 4.1, according to the method described in step 1, collect the effective electrical signal of the gyro motor to be diagnosed of model A;

[0123] Step 4.2: Extract sensitive characteristic parameters from the effective electrical signal collected in step 4.1 according to the method described in step 2, and perform normalization processing on each sensitive characteristic parameter;

[0124] Step 4.3, calculate the sensitive feature parameters in HMM qua and HMM fau The forward probability at each time point t is

[0125] Step 4.4: Calculate the sensitive feature parameters in HMM qua and HMM fau The probability of appearing in

[0126] Step 4.5: Determine the weight coefficient of each sensitive feature parameter as ρ d , d=1,2,…,D;

[0127] The weight coefficients of each sensitive characteristic parameter can be set to the same in the initial stage, or slightly adjusted according to the actual situation of the sensitive characteristic parameter identification process; the weight coefficients cannot differ too much, and the subsequent iterative optimization of the weight coefficients mainly depends on the actual production results of the statistics. In this embodiment, the weight coefficients of standard deviation, kurtosis, crest ratio, and skewness are all set to 1;

[0128] HMM qua middle:

[0129] When d=1, the corresponding sensitive characteristic parameter is the standard deviation.

[0130] When d=2, the corresponding sensitive characteristic parameter is the peak ratio.

[0131] When d=3, the corresponding sensitive characteristic parameter is kurtosis.

[0132] When d=4, the corresponding sensitive characteristic parameter is skewness.

[0133] HMM fau middle:

[0134] When d=1, the corresponding sensitive characteristic parameter is the standard deviation.

[0135] When d=2, the corresponding sensitive characteristic parameter is the peak ratio.

[0136] When d=3, the corresponding sensitive characteristic parameter is kurtosis.

[0137] When d=4, the corresponding sensitive characteristic parameter is skewness.

[0138] Step 4.6: Compare the following equations: If so, the gyro motor bearing is diagnosed as qualified; if not, the gyro motor bearing is diagnosed as faulty;

[0139] This embodiment is established, and the bearing of the gyro motor is diagnosed as qualified;

[0140] Step 5: Optimize and adjust HMM qua and HMM fau The weight coefficient of

[0141] Step 5.1, reference Figure 7 , make a comprehensive judgment on the gyro motor that has completed step 4;

[0142] Step 5.1.1: Perform a first-level judgment on the gyro motor that has completed step 4. If the gyro motor's indicators do not meet the requirements or there is an abnormal sound, the gyro motor has a first-level fault and proceed to step 5.1.2; otherwise, proceed to step 5.1.3;

[0143] Step 5.1.2: Disassemble the gyro motor for secondary diagnosis. If any of the gyro motor bearing parts (outer ring raceway, inner ring roller, roller surface, cage surface) are found to be abnormally worn under microscope observation, the gyro motor is comprehensively diagnosed as faulty. Otherwise, proceed to step 5.1.3.

[0144] Step 5.1.3: The gyro assembled with the gyro motor undergoes subsequent production processes. After all indicators are inspected and qualified, the gyro is delivered to the user for use. If a fault related to the motor bearing occurs during use, the gyro motor is comprehensively judged to be faulty; otherwise, the gyro motor is comprehensively judged to be qualified.

[0145] Step 5.2: Count the diagnostic results of the gyro motor in step 4 and the comprehensive judgment results in step 5. The diagnostic results include the product number and the D sensitive feature parameters obtained in step 4 in the HMM. qua and HMM fau The probability of appearing in Where d = 1, 2, ..., D;

[0146] Step 5.3: For each gyro motor counted in step 5.2, compare them one by one The size of The diagnosis result of the sensitive characteristic parameter item d is qualified; if Then the diagnosis result of the sensitive characteristic parameter item d is fault;

[0147] Step 5.4: If the diagnosis result of the dth sensitive characteristic parameter obtained in step 5.3 is consistent with the comprehensive judgment result of the motor bearing, then the sensitive characteristic parameter is scored as 1; otherwise, the sensitive characteristic parameter is scored as 0;

[0148] Step 5.5: Calculate the scores of D sensitive feature parameters in turn, where the score of the dth sensitive feature parameter is G d , where d = 1, 2, ..., D;

[0149] Step 5.6: Calculate the weight coefficients of D sensitive characteristic parameters in turn, and obtain the weight coefficients of each sensitive characteristic parameter as follows: Where d = 1, 2, ..., D;

[0150] If the diagnosis of all gyro motors to be diagnosed has been completed, the process ends; otherwise, the process returns to step 4.

[0151] Taking the actual situation of 8 batches, totaling 90 pieces of a certain type of miniaturized flexible gyroscope as an example, after each batch is completed, a line graph of the batch number-each weight coefficient is drawn, as shown in the following figure: Figure 12 As shown in the figure, it can be seen that each weight coefficient has a monotonic change trend and is relatively stable. At the same time, the accuracy of bearing fault diagnosis has increased slightly when compared between batches, indicating that with the continuous increase in the number of samples, the weight coefficients are constantly approaching the true value.

[0152] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. For ordinary professional and technical personnel in this field, the specific technical solutions recorded in the above embodiments can be modified, or some of the technical features therein can be replaced by equivalents. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions protected by the present invention.

Claims

1. A gyro motor bearing fault diagnosis method, characterized in that: The steps include: Step 1: Collect current signals and power signals of multiple gyro motors of the same model, and identify effective electrical signals through analysis; The multiple gyro motors are divided into two states: qualified bearings and bearing faults. The qualified bearings account for 45% to 55%. The qualified bearings state means that the gyro assembled by the gyro motor meets the index requirements and has no abnormal noise, and no motor bearing-related faults occur during the subsequent gyro assembly, commissioning, acceptance delivery and the user's full life cycle. The bearing fault state means that the relevant technical indicators of the gyro motor exceed the tolerance and abnormal noise occurs during operation. The effective electrical signal includes an effective current signal and an effective power signal. The effective current signal is the current signal from the current signal entering a stable state to before power failure. The effective power signal is the power signal from the power signal entering a stable state to before power failure. Step 2: Extract characteristic parameters from the effective electrical signal identified in step 1 one by one, and compare and analyze the characteristic parameters corresponding to the qualified state of the bearing and the characteristic parameters corresponding to the faulty state of the bearing. If a characteristic parameter has a significant difference, identify the characteristic parameter as a sensitive characteristic parameter; the characteristic parameters include mean value, standard deviation, crest ratio, kurtosis, skewness, and spectral density; Step 3: Adaptively design the HMM based on the number of effective electrical signals, the number of sensitive characteristic parameters, and the characteristic factors of the sensitive characteristic parameters, and implement the algorithm function through software; The designed HMM is used to model the qualified state and fault state of the gyro motor bearing, which are defined as HMM qua and HMM fau ; The HMM qua and HMM fau The dimension is the same as the number D of sensitive feature parameters identified in step 2; Step 4: According to the method described in step 1 and step 2, collect the effective electrical signal of the gyro motor to be diagnosed in the same model as step 1, extract the sensitive characteristic parameters, and use the HMM established in step 4 qua and HMM fau Perform fault diagnosis and finally obtain the fault diagnosis result of the gyro motor.

2. A gyro motor bearing fault diagnosis method according to claim 1, characterized in that: The step 1 is specifically as follows: Step 1.1: Collect current signals and power signals from multiple gyro motors of the same model, where the number of the multiple gyro motors is not less than 30; the current signals and power signals cover the startup phase, the working phase, and the stop phase of the gyro motor, and the collection time is not less than 40 minutes; Step 1.2: Smoothing the current signal and power signal collected in step 1.1, dividing the current signal and power signal into multiple equal parts according to time, obtaining the average value of the current signal and the average value of the power signal in each equal part, and making a time-current signal average value line graph and a time-power signal average value line graph; Step 1.3, based on the time-current signal average value line graph and the time-power signal average value line graph obtained in step 1.2, respectively obtain the corresponding time when each gyro motor current signal and power signal enters a stable state; Step 1.4: Calculate the average time required for all gyro motor current and power signals to enter a stable state. and standard deviation σ, will Determine the time required for the same model of gyro motor to enter a stable state; Step 1.5: Extract all gyro motors The current signal and power signal before power failure are regarded as effective electrical signals.

3. A gyro motor bearing fault diagnosis method according to claim 2, characterized in that: The step 3 is specifically as follows: Step 3.1, extracting sensitive characteristic parameters corresponding to the gyro motor in the bearing qualified state and the bearing fault state from the effective electrical signal; Step 3.2: normalize each sensitive feature parameter; Step 3.3: Initialize HMM qua and HMM fau Parameters Π of each dimension z 、A z 、B z , z = 0; Among them, Π z is the hidden state probability matrix after z iterations; A z is the state transfer matrix after z iterations; B z The observation state generation matrix after z iterations; z is the number of iterations; Step 3.4, according to π z 、A z 、B z The value of each element in , calculate HMM qua and HMM fau The forward probability α at each time point t in each dimension t ; Step 3.5: According to π z 、A z 、B z The value of each element in , calculate HMM qua and HMM fau The backward probability β at each time point t in each dimension t ; Step 3.6, according to the forward probability α t , backward probability β t , calculate HMM qua and HMM fau The probability γ that the hidden state is i at time t in each dimension t (i); Among them, i represents the hidden state of the gyro motor, which includes two states: qualified and faulty; Step 3.7, according to the forward probability α t , backward probability β t , calculate HMM qua and HMM fau The probability that the hidden state at time t is i and the hidden state at time t+1 is j in each dimension is ξ t (i,j); Among them, i and j represent the hidden status of the gyro motor, which includes qualified and faulty states; Step 3.8, according to the probability γ t (i) Probability ξ t (i,j), recalculate HMM qua and HMM fau Parameters Π of each dimension z 、A z 、B z , and let z=z+1; Among them, Π z is the hidden state probability matrix after z iterations; A z is the state transfer matrix after z iterations; B z Generate a matrix for the observed state after z iterations; Step 3.9: Calculate HMM qua and HMM fau The corresponding π z and Π z-1 , A z and A z-1 , B z and B z-1 The difference between the elements at each corresponding position in the algorithm is defined as the threshold indicator for termination of the algorithm. When the absolute value of any of the differences is greater than ε, return to step 3.

4. When the absolute value of the differences is not greater than ε, the HMM is obtained. qua and HMM fau Parameters Π of each dimension z 、A z 、B z , that is, complete HMM qua and HMM fau The establishment of.

4. A gyro motor bearing fault diagnosis method according to claim 3, characterized in that: The step 4 is specifically as follows: Step 4.1, according to the method described in step 1, collect the effective electrical signal of the gyro motor to be diagnosed of the same model as in step 1; Step 4.2: Extract sensitive characteristic parameters from the effective electrical signal collected in step 4.1 according to the method described in step 2, and perform normalization processing on each sensitive characteristic parameter; Step 4.3, calculate the sensitive feature parameters in HMM qua and HMM fau The forward probability at each time point t is Step 4.4: Calculate the sensitive feature parameters in HMM qua and HMM fau The probability of appearing in Step 4.5: Determine the weight coefficient of each sensitive feature parameter as ρ d , d=1,2,…,D; Step 4.6: Compare the following equations: If so, the bearing of the gyro motor is diagnosed as qualified; if not, the bearing of the gyro motor is diagnosed as faulty.

5. A gyro motor bearing fault diagnosis method according to claim 4, characterized in that: Also includes step 5, optimizing and adjusting HMM qua and HMM fau The weight coefficient of Step 5.1, conduct a comprehensive assessment of the gyro motor that has completed step 4; Step 5.1.1: Perform a first-level check on the gyro motor that has completed step 4. If the gyro motor's indicators do not meet the requirements or there is an abnormal sound, the gyro motor has a first-level fault and proceed to step 5.1.

2. Otherwise, proceed to step 5.1.3; Step 5.1.2: Disassemble the gyro motor for secondary diagnosis. If any of the gyro motor bearing parts are found to be abnormally worn under a microscope, the gyro motor is comprehensively judged to be faulty. Otherwise, proceed to step 5.1.

3. Step 5.1.3: The gyro assembled with the gyro motor undergoes subsequent production processes. After all indicators are inspected and qualified, the gyro is delivered to the user for use. If a fault related to the motor bearing occurs during use, the gyro motor is comprehensively judged to be faulty; otherwise, the gyro motor is comprehensively judged to be qualified. Step 5.2: Count the diagnostic results of the gyro motor in step 4 and the comprehensive judgment results in step 5.

1. The diagnostic results include the product number and the D sensitive feature parameters obtained in step 4 in the HMM. qua and HMM fau The probability of appearing in Where d = 1, 2, ..., D; Step 5.3: For each gyro motor counted in step 5.2, compare them one by one. The size of The diagnosis result of the sensitive characteristic parameter item d is qualified; if Then the diagnosis result of the dth sensitive characteristic parameter is fault; Step 5.4: If the diagnosis result of the dth sensitive characteristic parameter obtained in step 5.3 is consistent with the comprehensive judgment result of the motor bearing, then the sensitive characteristic parameter is scored as 1; otherwise, the sensitive characteristic parameter is scored as 0; Step 5.5: Calculate the scores of D sensitive feature parameters in turn, where the score of the dth sensitive feature parameter is G d , where d = 1, 2, ..., D; Step 5.6: Calculate the weight coefficients of D sensitive characteristic parameters in turn, and obtain the weight coefficients of each sensitive characteristic parameter as follows: Where d = 1, 2, ..., D, which means the optimization adjustment of the weight coefficient is completed; If the diagnosis of all gyro motors to be diagnosed has been completed, the process ends; otherwise, the process returns to step 4.

6. A gyro motor bearing fault diagnosis method according to any one of claims 1 to 5, characterized in that: The step 2 is specifically as follows: Step 2.1, the effective electrical signal identified in step 1 is divided into segments according to time, various characteristic parameters are extracted from each segment of the effective electrical signal, and a time-characteristic parameter line graph is drawn. The time-characteristic parameter line graph is divided into a bearing qualified state group and a bearing fault state group; Step 2.2: Compare the qualified bearing state group and the faulty bearing state group. If a certain characteristic parameter has a significant difference, identify the characteristic parameter as a sensitive characteristic parameter. The significant difference is defined as a characteristic parameter of the qualified bearing state group being stable or concentrated near a certain value in multiple gyro motors, and a characteristic parameter of the faulty bearing state group being abrupt or concentrated near another different value in multiple gyro motors.

7. A gyro motor bearing fault diagnosis method according to any one of claims 3 to 5, characterized in that: The step 3.2 specifically includes: performing local normalization on the average value and the spectral density as a type of sensitive characteristic parameter, wherein the local normalization is to compare the sensitive characteristic parameter only with itself; The standard deviation, kurtosis, crest ratio, and skewness are used as the second-category sensitive feature parameters for global normalization, and the global normalization is the comparison of this sensitive feature parameter with all the second-category sensitive feature parameters.

Images