A Simple Gearbox Fault Detection Method

Abstract

This invention proposes a simplified gearbox fault detection method to address the technical problems of high complexity and poor accuracy in existing gear fault detection methods. The invention involves continuously acquiring m sets of vibration signals from a gearbox at equal intervals and preprocessing them to identify sensitive features for gear faults. The data for these sensitive features is then preprocessed. The first q sets of data are divided into p intervals, and the slope value of the sensitive feature in the i-th interval is calculated using the least squares method. A data matrix is ​​constructed using the slope value. The weight value of the sensitive feature is calculated using the entropy method. The gear deterioration index in the i-th interval is calculated using the sum-product method with the weight value and the slope value. An alarm value for the gear deterioration index is calculated based on the central limit theorem. The gear deterioration index is calculated for the remaining r sets of data. If the gear deterioration index ≥ w, the gear is considered faulty; otherwise, the gear is not faulty. This invention can automatically detect the health status of the gearbox, significantly improving the accuracy of gear fault identification.

Description

Technical Field

[0001] This invention relates to the technical field of equipment condition monitoring and fault diagnosis, and in particular to a simple method for detecting gearbox faults. Background Technology

[0002] Gearboxes are critical components in mechanical equipment, widely used in pillar industries such as manufacturing, coal, petrochemicals, power, and water conservancy. Health monitoring of gearboxes can detect early-stage faults, improve intelligent operation and maintenance, prevent unplanned downtime, reduce the frequency of malfunctions, increase equipment operating efficiency, and ensure safe and stable operation.

[0003] Typically, when bearings or gears in a gearbox fail, it's difficult for humans to differentiate and diagnose the problem based on experience alone. Experts usually determine the fault type by analyzing the impact components of vibration spectra or pinpointing the exact frequency and location of the fault. In the field of intelligent manufacturing, intelligent operation and maintenance of gearboxes and unattended operation scenarios urgently require an accurate and automatic monitoring method for identifying gearbox faults. Therefore, it is necessary to propose a simple and accurate automatic method for identifying gearbox faults, automatically detecting gearbox health status, promoting the implementation of predictive maintenance models, and helping enterprises reduce costs, improve quality, and increase efficiency.

[0004] The gear fault diagnosis method based on the combination of VMD entropy method and VPMCD, with application number 202010270912.6, combines variational mode decomposition (VMD) and variable prediction model pattern recognition (VPMCD) to purify gear vibration signals, filter out most useless noise interference components, and highlight the signal's intrinsic information, resulting in higher fault identification accuracy and efficiency. However, its processing method is complex and computationally intensive, and it cannot perform real-time fault identification. Summary of the Invention

[0005] To address the technical problems of high complexity and poor accuracy in existing gear fault detection methods, this invention proposes a simplified gearbox fault detection method that can automatically detect the health status of the gearbox and significantly improve the accuracy of gear fault identification.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: a simple gearbox fault detection method, comprising the following steps:

[0007] Step 1: Under rated operating conditions, continuously collect m sets of vibration signals x(t) from the gearbox at equal intervals, preprocess the vibration signals x(t), calculate the feature values ​​in the time domain, frequency domain, and time-frequency domain respectively, and screen out s sensitive features of gear faults;

[0008] Step 2: Perform data preprocessing on each of the m sets of data for the s sensitive features to obtain n sets of data for the s sensitive features;

[0009] Step 3: Divide the first q sets of data for the s sensitive features into p equal intervals, with interval length L = q / p. Use the least squares method to calculate the slope value k of the j-th sensitive feature in the i-th interval. ij Using the slope value k ij Construct the data matrix K for the j-th sensitive feature; where i = 1, 2, ..., p; j = 1, 2, ..., s;

[0010] Step 4: Based on the data matrix K, calculate the weight value of the j-th sensitive feature using the entropy method.

[0011] Step 5: Calculate the gear deterioration index for the i-th interval using the sum-product method. Based on the central limit theorem, the alarm value w of the gear deterioration index is calculated;

[0012] Step 6: For the s sensitive features and r = nq sets of data, when the data length reaches L, calculate the gear deterioration index based on the slope and weight values ​​of each sensitive feature. If the gear deterioration index Determine if the gear is faulty; otherwise, the gear is not faulty.

[0013] Preferably, the sensitive feature includes a single-peak P k Envelope value E v and impact meshing index value I m The impact meshing index value I m for:

[0014]

[0015] In the formula, the root mean square value R v >0.1, 1 times the meshing frequency value G m >0.1, 100>I m >0; W v P is a waveform indicator. k It is a single peak value, and k, v, and m are subscripts.

[0016] Preferably, the preprocessing in step 1 involves removing random noise signals from the vibration signal x(t) using a white noise test method; the data preprocessing in step 2 includes removing equipment shutdown data and alarm data, and calculating the median; removing equipment shutdown data refers to deleting the feature values ​​calculated when the rotational speed is 0; removing equipment alarm data refers to deleting the feature values ​​calculated when the single peak value exceeds the preset alarm value; calculating the median means obtaining a median for every 10 data sets after removing shutdown data from m data sets.

[0017] Preferably, the least squares method is used to calculate the slope value k of the j-th sensitive feature in the i-th interval. ij The method is as follows:

[0018] For the j-th sensitive feature, construct the data matrix Y within the i-th interval. ij Using the least squares method to perform univariate linear fitting of Xβ ij =Y ij ,and:

[0019]

[0020] In the formula, y1-y L For the j-th sensitive feature, there are L sets of data within the i-th interval;

[0021] According to formula β ij =(X T X) -1 X T Y ij Find the slope value k of the fitted curve within this interval. ij .

[0022] Preferably, the entropy method is as follows: the entropy value of the j-th sensitive feature is:

[0023]

[0024] Among them, k ij Let be the slope value of the j-th sensitive feature in the i-th interval, and p be the number of intervals;

[0025] The weight value of the j-th sensitive feature for:

[0026]

[0027] Preferably, if the slope value is less than 0, the absolute value of the slope value is taken; if the slope value of a column is equal to 0, a value is added to the data in that column, which is 0.01 by default.

[0028] Preferably, the method for calculating the gear deterioration index is as follows: the gear deterioration index for the i-th interval is...

[0029]

[0030] Where s is the total number of sensitive features.

[0031] Preferably, the implementation method of the central limit theorem is as follows: calculating the gear degradation index. The mean μ and standard deviation σ are given; according to the 3σ principle, the alarm value w of the gear deterioration index is: w = μ + 3σ.

[0032] The beneficial effects of this invention are as follows: By calculating the gear deterioration index of sensitive features using the sum-product method, multiple sensitive features of gear faults are processed together, resulting in higher diagnostic accuracy compared to a single indicator. The gear deterioration index alarm value is calculated based on normal data using the central limit theorem. According to the 3 Sigma principle, the gear deterioration index will exceed this alarm value when a fault occurs. This invention, combined with the characteristics of the gearbox itself, automatically detects the health status of the gearbox, significantly improving the accuracy of gear fault identification, promoting the implementation of predictive maintenance models for equipment, and helping enterprises reduce costs, improve quality, and increase efficiency. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram of the process of the present invention.

[0035] Figure 2 The digital model of the device provided in the embodiments of the present invention.

[0036] Figure 3 The above are waveform spectrum diagrams of the gearbox before and after the actual failure of the gearbox according to an embodiment of the present invention, where (a) is before the failure and (b) is after the failure.

[0037] Figure 4 This is a trend diagram of sensitive features before and after an actual gearbox failure occurs, according to an embodiment of the present invention.

[0038] Figure 5 This is a field diagram of a broken gear in a gearbox according to an embodiment of the present invention. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] like Figure 1 As shown, a simple gearbox fault detection method includes the following steps:

[0041] Step 1: Under rated operating conditions, continuously collect m sets of original vibration signals x(t) of the gearbox at equal intervals. Preprocess each set of vibration signals, calculate the feature values ​​in the time domain, frequency domain, and time-frequency domain respectively, and screen out s gear fault sensitive features.

[0042] Under rated operating conditions, the original vibration signals x(t) of m gearbox groups are continuously collected at equal intervals, with a sampling frequency of f. s With N sampling points, signal preprocessing is performed on each set of vibration signals x(t), and then corresponding characteristic values ​​are calculated in the time domain, frequency domain, and time-frequency domain. The signal preprocessing process removes random noise signals and retains normal signals using a white noise test method. The time-frequency domain characteristic values ​​are obtained by passing the vibration signal x(t) through a series of processes including Butterworth filter, Fourier transform and inverse transform, and Hilbert-Huang transform. Among these, the time-domain characteristics include the single-peak value P. k Root mean square value R v Waveform Indicator W v Frequency domain characteristics include a meshing frequency value G. m The time-frequency domain features include the envelope value E. v and impact meshing index value I m .

[0043] Impact meshing index value I m The calculation formula is:

[0044]

[0045] In the formula, the root mean square value R v >0.1, 1 times the meshing frequency value G m >0.1, 100>I m >0. W v P is a waveform indicator. k It is a single peak value. The subscripts k, v, and m represent the differences between peak value, envelope value, and meshing frequency, respectively.

[0046] Selecting a single peak value P k Envelope value E v and impact meshing index value I m This is a sensitive characteristic of gear failure, where s = 3.

[0047] Under rated operating conditions, m sets of original vibration signals x(t) were continuously collected at equal intervals from measuring point 5 of the mill gearbox (sampling frequency f). s =5120Hz, number of sampling points N=8192), the digital model of the mill is as follows Figure 2 As shown, the motor is connected to the coal mill via a gearbox, and they are connected by gears and gear shafts. From Figure 2 The information displayed includes the device structure, the types of sensors installed, and the locations of the measurement points. Here, m = 2098.

[0048] The gear parameters and meshing frequency of the gearbox are shown in Table 1.

[0049] Table 1. Gear parameters and meshing frequency (Hz) of the gearbox

[0050]

[0051] like Figure 3 As shown, by comparing the waveform spectrum diagrams before and after the gearbox failure, it can be seen that the waveform diagrams have obvious impact characteristics and significant changes in single peak values; the sideband amplitude at the gear meshing frequency in the spectrum diagrams changes significantly, that is, the envelope value changes significantly.

[0052] like Figure 4 As shown, the single peak value P k Envelope value E v and impact meshing index I m The amplitude of the value changes significantly with the occurrence of gear failure; therefore, a single peak value P was selected. k Envelope value E v and impact meshing index value I m For gear fault-sensitive characteristics, s = 3 here.

[0053] Step 2: Perform data preprocessing on each of the m sets of data for the s sensitive features to obtain n sets of data for the s sensitive features.

[0054] The data preprocessing process includes removing equipment shutdown data and alarm data, and calculating the median. Removing equipment shutdown data refers to deleting the feature values ​​calculated when the speed is 0. Removing equipment alarm data refers to deleting the feature values ​​calculated when the single peak value exceeds the preset alarm value. Calculating the median means obtaining a median from every 10 data sets after removing shutdown data from m data sets. The purpose of calculating the median is to improve the accuracy of gear fault identification and avoid misjudgments caused by poor signal acquisition quality at certain moments.

[0055] Single peak P k Envelope value E v and impact meshing index value I m The median trend chart is as follows Figure 4 As shown, here, n = 210. The first 90 sets of data are selected as normal characteristic data of the gearbox, used to calculate the baseline value and alarm value of the gear deterioration index. The data after the first 90 sets are used to verify the actual application effect of the gear deterioration index value.

[0056] Step 3: Divide the first q sets of data for the s sensitive features into p equal intervals, and use the least squares method to calculate the slope value k of the i-th interval. i Construct the data matrix k of the j-th sensitive feature. ij Where i = 1, 2, ..., p; j = 1, 2, ..., s.

[0057] Given n sets of data with a certain sensitive feature, divide them into p equal intervals, with interval length L = n / p. Construct a data matrix within the i-th interval and perform a univariate linear fit Xβ = Y using the least squares method. That is, given matrices X and Y, find the parameter β, where...

[0058]

[0059] In the formula, y1-y L Let L be the data sets within the i-th interval of this sensitive feature.

[0060] According to the formula β=(X T X) -1 X T Y calculates the slope k = β of the fitted curve within this interval, and so on, calculates the slope value k of the feature data in all intervals of this sensitive feature. i Let i = 1, 2, ..., p. The purpose of dividing the curve into intervals is to find the changes between intervals.

[0061] Repeat the above process to calculate all slope values ​​k of the s sensitive features in p intervals. ij Construct data matrix k ij (i = 1, 2, ..., p; j = 1, 2, ..., s). Here, s = 3, q ​​= 90, p = 18, and L = 5.

[0062] Step 4: Based on data matrix k ij The weight value of the j-th sensitive feature is calculated using the entropy method.

[0063] For data matrix k ij The entropy method is used to calculate the weights of s sensitive features. The entropy method calculation formula is as follows:

[0064]

[0065] Among them, e j Let k be the entropy value of the j-th sensitive feature. ij Let be the slope value of the j-th sensitive feature in the i-th interval, p be the number of intervals, and s be the number of sensitive features.

[0066] If a slope value is less than 0, then the absolute value of the slope value is taken; if the slope value of a column is equal to 0, then a value is added to the data in that column. In the entropy method calculation, there cannot be negative numbers, and the default value is 0.01.

[0067] The formula for calculating the weight value of the j-th sensitive feature is:

[0068]

[0069] in, Let be the weight value of the j-th sensitive feature.

[0070] As shown in Table 2, the single-peak P k Envelope value E v and impact meshing index value I m The weights of the three sensitive features are 0.213, 0.422, and 0.366, respectively.

[0071] Table 2 Sensitive Feature Index Parameters and Calculated Values

[0072]

[0073]

[0074] Step 5: Calculate the gear deterioration index for the i-th interval using the sum-product method. Based on the central limit theorem, the alarm value w of the gear deterioration index is calculated.

[0075] Gear deterioration index The calculation formula used is:

[0076]

[0077] in, Let be the gear deterioration index value for the i-th interval.

[0078] Calculate the gear degradation index based on the central limit theorem. The mean μ and standard deviation σ are used to calculate the alarm value w of the gear deterioration index according to the 3σ principle. The calculation formula is as follows:

[0079] w=μ+3σ

[0080] As shown in Table 2, the baseline and alarm values ​​for the gear deterioration index are 0.065 and 0.163, respectively. The baseline value for the gear deterioration index is the mean μ, which is used to calculate the alarm value for the deterioration index.

[0081] Step 6: For s sensitive features followed by r sets of data, r = nq, when the data length reaches L, determine the slope k of each sensitive feature. ij Combining the weight values ​​of each sensitive feature Calculate the gear deterioration index If the gear deterioration index This indicates a gear malfunction; otherwise, the gear is not malfunctioning. If the data length has not reached L, continue waiting until the data length reaches L.

[0082] contrast Figure 4 As shown in Table 2, the calculated gear deterioration index changes significantly when the trend worsens. Figure 4 The gear deterioration index values ​​for ranges 101-105 and 176-180 are 0.687 and 0.415 respectively, both exceeding the baseline and alarm values. The range 101-105 has the highest calculated deterioration index value. Figure 5 As shown, manual on-site shutdown and inspection confirmed that the gearbox had a broken tooth fault. After comparing with the time of the fault occurrence on-site, the above interval exactly corresponds to the time of the equipment fault, proving that the gear deterioration index can accurately identify faults such as broken teeth and tooth meshing in the gearbox.

[0083] In summary, the method proposed in this invention can automatically detect the health status of the gearbox by combining the characteristics of the equipment itself, significantly improve the accuracy of gear fault identification, help enterprises achieve predictive maintenance of equipment, and has high engineering application value.

[0084] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A simple method for detecting gearbox faults, characterized in that, The steps are as follows: Step 1: Under rated operating conditions, continuously collect m sets of vibration signals x(t) from the gearbox at equal intervals, preprocess the vibration signals x(t), calculate the feature values ​​in the time domain, frequency domain, and time-frequency domain respectively, and screen out s sensitive features of gear faults; Step 2: Perform data preprocessing on each of the m sets of data for the s sensitive features to obtain n sets of data for the s sensitive features; Step 3: Divide the first q sets of data for the s sensitive features into p equal intervals, with interval length L = q / p. Use the least squares method to calculate the slope value k of the j-th sensitive feature in the i-th interval. ij Using the slope value k ij Construct the data matrix K for the j-th sensitive feature; where i = 1, 2, ..., p; j = 1, 2, ..., s; Step 4: Based on the data matrix K, calculate the weight value of the j-th sensitive feature using the entropy method. Step 5: Calculate the gear deterioration index for the i-th interval using the sum-product method. Based on the central limit theorem, the alarm value w of the gear deterioration index is calculated; Step 6: For the s sensitive features and r = nq sets of data, when the data length reaches L, calculate the gear deterioration index based on the slope and weight values ​​of each sensitive feature. If the gear deterioration index Determine if the gear is faulty; otherwise, the gear is not faulty.

2. The simplified gearbox fault detection method according to claim 1, characterized in that, The sensitive feature includes single-peak P k Envelope value E v and impact meshing index value I m The impact meshing index value I m for: In the formula, the root mean square value R v >0.1, 1 times the meshing frequency value G m >0.1, 100>I m >0;W v P is a waveform indicator. k It is a single peak value, and k, v, and m are subscripts.

3. The simplified gearbox fault detection method according to claim 1 or 2, characterized in that, The preprocessing in step 1 involves removing random noise signals from the vibration signal x(t) using a white noise test method. The data preprocessing in step 2 includes removing equipment shutdown data and alarm data, as well as calculating the median. Removing equipment shutdown data means deleting the feature values ​​calculated when the rotational speed is 0. Removing equipment alarm data means deleting the feature values ​​calculated when the single peak value exceeds the preset alarm value. Calculating the median means obtaining a median for every 10 data sets after removing shutdown data from m data sets.

4. The simplified gearbox fault detection method according to claim 3, characterized in that, The least squares method is used to calculate the slope value k of the j-th sensitive feature in the i-th interval. ij The method is as follows: For the j-th sensitive feature, construct a data matrix Y within the i-th interval. ij Using the least squares method to perform univariate linear fitting of Xβ ij =Y ij ,and: In the formula, y1-y L For the j-th sensitive feature, there are L sets of data within the i-th interval; According to formula β ij =(X T X) -1 X T Y ij Find the slope value k of the fitted curve within this interval. ij .

5. The simplified gearbox fault detection method according to claim 1 or 4, characterized in that, The entropy method is as follows: the entropy value of the j-th sensitive feature is: Where, k ij Let be the slope value of the j-th sensitive feature in the i-th interval, and p be the number of intervals; The weight value of the j-th sensitive feature for:

6. The simplified gearbox fault detection method according to claim 5, characterized in that, If the slope value is less than 0, the absolute value of the slope value is taken; if the slope value of a column is equal to 0, a value is added to the data in that column, which is 0.01 by default.

7. The simplified gearbox fault detection method according to claim 6, characterized in that, The method for calculating the gear deterioration index is as follows: the gear deterioration index for the i-th interval is... Where s is the total number of sensitive features.

8. The simplified gearbox fault detection method according to claim 1 or 7, characterized in that, The implementation method of the central limit theorem is as follows: calculate the gear degradation index. The mean μ and standard deviation σ are given; according to the 3σ principle, the alarm value w of the gear deterioration index is: w = μ + 3σ.

Images