Gearbox abnormity identification method

Abstract

The invention provides a gearbox abnormity identification method. The gearbox abnormity identification method comprises the following steps: a model training process: obtaining training data of a model; calculating the rotating speed of each gear step by step; determining the order of each gear; separating a time sequence signal of each gear; judging whether the rotating speed of the output shaft meets a preset condition; intercepting a time sequence signal of each gear in the period of time; performing equiangular domain resampling on the intercepted time sequence signals to obtain angular domain signals; performing fast Fourier transform on the angular domain signal to obtain an order spectrum; calculating a statistical variable corresponding to each order; inputting the statistical variables into the single classification model to train the single classification model to obtain a trained model; using preset test data to test the trained model; and judging whether the gearbox is abnormal or not by using the data acquired in real time and the tested model. The method can judge whether the gearbox is abnormal or not in real time, and improves the efficiency of judging whether the gearbox is abnormal or not.

Description

Technical Field

[0001] The present invention relates to the technical field of gear boxes, and in particular to a method for identifying abnormalities in gear boxes. Background Art

[0002] Excavators, aerial work platforms and other construction machinery are equipped with gearboxes, which can also be called transmissions. Gearboxes include an input shaft, multiple gears and an output shaft. Multiple gears are meshed and connected step by step and then meshed and connected with the input shaft and output shaft. Gearbox failures include gear failures, input shaft failures and output shaft failures. Gear failures include tooth tip fracture, tooth cavitation, tooth cracking and tooth wear. Input shaft failures and output shaft failures include bearing seat cracking, bearing inner ring failures, bearing outer ring failures and rolling element failures. These failures will cause abnormal operation or damage to the gearbox.

[0003] As a key component to cope with various harsh working conditions, the gearbox has high design requirements at the beginning of the design, and a series of reliability tests will be carried out to meet the needs of industrial production. Therefore, the probability of failure of the gearbox in actual work is low, and the time, hardware and manpower costs are high if you want to collect failure data of the actual vehicle.

[0004] In construction machinery, due to frequent switching of working conditions and large speed fluctuations, more signal interference is caused. At the same time, the speed sensor has a large error at low speeds, which will bring more interference and noise to signal analysis. Therefore, the accuracy of judging whether the gearbox is abnormal based on speed is currently low. Summary of the invention

[0005] The present invention provides a gearbox abnormality identification method to solve the technical problems of high cost and low accuracy in gearbox abnormality identification.

[0006] In order to solve the above technical problems, the present invention provides a method for identifying abnormalities in a gearbox, wherein the gearbox comprises a plurality of gears and an output shaft, and the method comprises the following steps:

[0007] Model training process:

[0008] Acquire training data of the model, the training data including the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, the model is a single classification model, and the training data is the operation data of the gearbox in a normal state;

[0009] Calculate the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft;

[0010] Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft;

[0011] Based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, a time series signal of each gear is separated by using a Vold-Kalman algorithm;

[0012] Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range;

[0013] If the rotation speed of the output shaft meets the preset condition, intercepting the timing signal of each gear within the period of time;

[0014] Resampling the intercepted time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra;

[0015] Calculate the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range;

[0016] Inputting the statistical variables into the single classification model to train the single classification model to obtain a trained model;

[0017] Model testing process:

[0018] Using preset test data to test the trained model and determine a test threshold, the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state;

[0019] Model usage process:

[0020] During the operation of the gearbox, the real-time data and the tested model are used to determine whether the gearbox is abnormal.

[0021] Preferably, the preset condition also includes whether the speed increase rate of the output shaft is less than a first threshold and whether the speed fluctuation of the output shaft is less than a second threshold.

[0022] Preferably, the step of testing the trained model using preset test data specifically includes the following steps:

[0023] Acquire test data of the model, wherein the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state;

[0024] Calculating the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft in the test data, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft;

[0025] Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft;

[0026] Based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, a time series signal of each gear is separated by using a Vold-Kalman algorithm;

[0027] Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range;

[0028] If the rotation speed of the output shaft meets the preset condition, then the timing signal of each gear within the period of time is intercepted;

[0029] Resampling the intercepted time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra;

[0030] Calculate the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range;

[0031] The statistical variables are input into the trained model.

[0032] Preferably, the step of using the real-time acquired data and the tested model to determine whether the gearbox is abnormal specifically includes the following steps:

[0033] During the operation of the gearbox, the speed of the output shaft of the gearbox and the vibration data of the gearbox are obtained in real time;

[0034] Calculate the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft acquired in real time, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft;

[0035] Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft;

[0036] Based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, a time series signal of each gear is separated by using a Vold-Kalman algorithm;

[0037] Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range;

[0038] If the speed of the output shaft obtained in real time meets the preset condition, then the timing signal of each gear within the period of time is intercepted;

[0039] Resampling the intercepted time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra;

[0040] According to the order spectrum, the amplitude corresponding to each order and the preset interval range, respectively calculate the real-time statistical variables corresponding to each order;

[0041] The real-time statistical variables are input into the model after the test to determine whether the gearbox is abnormal. If the value output by the model after the test is greater than or equal to the test threshold, the gearbox is determined to be abnormal; if the value output by the model after the test is less than the test threshold, the gearbox is determined to be normal.

[0042] Preferably, the preset range is 30%r to 60%r, where r represents the maximum rotational speed of the output shaft of the gear box.

[0043] Preferably, the step of calculating the statistical variables corresponding to each order specifically includes the following steps: taking the ±1 order interval of each order as the target order interval, calculating the product between each order and the corresponding amplitude within the target order interval; and calculating the statistical variables corresponding to each order based on the product.

[0044] Preferably, the statistical variables include mean, variance, skewness and peak-to-peak value.

[0045] Preferably, the step of determining the order of each gear according to the ratio of the rotational speed of each gear to the rotational speed of the output shaft specifically includes the following steps: taking the ratio of the rotational speed of each gear to the rotational speed of the output shaft as the order of each gear.

[0046] Preferably, after the step of using the real-time acquired data and the tested model to determine whether the gearbox is abnormal, the method further includes the following step: displaying the determination result of the tested model.

[0047] The present invention also provides another gearbox abnormality identification method, wherein the gearbox comprises a plurality of gears and an output shaft, and the method comprises the following steps:

[0048] Model training process:

[0049] Acquire training data of the model, the training data including the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, the model is a single classification model, and the training data is the operation data of the gearbox in a normal state;

[0050] Calculate the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft;

[0051] Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft;

[0052] Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range;

[0053] If the rotation speed of the output shaft meets the preset condition, the vibration data of the gearbox, the order of each gear and the rotation speed of each gear within the period of time are intercepted;

[0054] Based on the intercepted vibration data of the gearbox, the order of each gear and the rotation speed of each gear, the time series signal of each gear is separated by using the Vold-Kalman algorithm;

[0055] Resampling the time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra;

[0056] Calculate the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range;

[0057] Inputting the statistical variables into the single classification model to train the single classification model to obtain a trained model;

[0058] Model testing process:

[0059] Using preset test data to test the trained model and determine a test threshold, the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state;

[0060] Model usage process:

[0061] During the operation of the gearbox, the real-time data and the tested model are used to determine whether the gearbox is abnormal.

[0062] The present invention provides a gearbox abnormality identification method, which can determine whether the gearbox is abnormal in real time. There is no need to wait until the gearbox cannot be used to manually inspect the gearbox, thereby improving the efficiency of determining whether the gearbox is abnormal and reducing the cost of determining whether the gearbox is abnormal. The training data is a zero-fault sample, which has a greater cost advantage when dealing with engineering machinery with a low failure rate and reduces the time for collecting fault samples. By using the data of the rising section as an analysis sample, the random noise caused by excessive or insufficient speed can be eliminated, and the uniformity and accuracy of the extracted statistical variables can be improved, thereby improving the accuracy of determining whether the gearbox is abnormal. The time series signal extracted using the vold-kalman algorithm is clearer than the traditional order analysis signal, and it solves the signal aliasing problem well. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 The present invention is a flowchart of a method for identifying an abnormality in a gearbox provided by an embodiment of the present invention.

[0064] Figure 2 The present invention provides a waveform diagram of the rotation speed of an output shaft of a gear box and a timing signal of a gear provided in one embodiment. DETAILED DESCRIPTION

[0065] In order to make the purpose, advantages and features of the present invention clearer, a gearbox abnormality identification method proposed by the present invention is further described in detail below in conjunction with the accompanying drawings. It should be noted that the accompanying drawings are all in a very simplified form and use non-precise proportions, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.

[0066] In the description of the present invention, the terms "first", "second", etc. are added for the convenience of description and reference, and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined by the terms "first", "second", etc. may explicitly or implicitly include one or more of the features.

[0067] like Figure 1 As shown, this embodiment provides a method for identifying abnormalities in a gearbox, wherein the gearbox includes a plurality of gears and an output shaft, and the method includes the following steps:

[0068] S1. Model training process:

[0069] S11, obtaining training data of the model, wherein the training data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, the model is a single classification model, and the training data is the operation data of the gearbox in a normal state. The single classification model may be an auto-encoder model. In order to fully train the model, training data corresponding to construction machinery of different life cycles may be selected, for example, training data corresponding to new vehicles that have just left the factory, construction machinery that has worked for 500 hours, 1000 hours, 1500 hours, 2000 hours, 3000 hours or 5000 hours may be selected; at the same time, training data corresponding to different working scenes such as sand fields, coal fields, quarries and earthwork sites may be selected. Training data may be collected using vibration sensors and speed sensors. In the process of collecting training data, it may be ensured that the construction machinery is in a normal state by means of human monitoring. When the construction machinery is in a normal state, the gearbox contained in the construction machinery is also in a normal state.

[0070] S12, according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft, the rotation speed of each gear is calculated step by step; the output shaft can be used as a reference axis to calculate the rotation speed of the previous gear according to Formula 1:

[0071]

[0072] Among them, n1 is the number of teeth on the output shaft, n2 is the number of teeth on the previous gear, f1 is the speed of the output shaft, and f2 is the speed of the previous gear.

[0073] The principle is the same as that of formula 1, and the speed of the k-th gear can be calculated according to formula 2:

[0074]

[0075] Among them, n k-1 is the number of teeth of the k-1th gear, n k is the number of teeth of the k-th gear, f k-1 is the speed of the k-1th gear, f k is the speed of the kth gear.

[0076] S13, determining the order of each gear according to the ratio of the rotation speed of each gear to the rotation speed of the output shaft; the ratio of the rotation speed of each gear to the rotation speed of the output shaft can be used as the order of each gear, that is, the order of each gear is calculated according to Formula 3. In other embodiments, the result calculated by Formula 3 can be multiplied by a preset coefficient to be used as the order of the corresponding gear.

[0077]

[0078] Where, f1 is the speed of the output shaft, f k is the speed of the k-th gear, Q k is the order of the kth gear.

[0079] S14, based on the vibration data of the gearbox, the order of each gear and the speed of each gear, the vold-kalman algorithm is used to separate the timing signal of each gear; the vibration data of the gearbox, the order of each gear and the speed of each gear are input into the vold-kalman algorithm to separate the timing signal of each gear. The original timing signal corresponding to all the gears of the gearbox contains the vibration data and noise data of all the gears. The timing signal s extracted by the vold-kalman algorithm k (t) can accurately reflect the vibration of the gear, and the noise data it contains is less than the noise data contained in the original timing signal.

[0080] S15. Determine whether the speed of the output shaft meets the preset conditions, wherein the preset conditions include that the speed of the output shaft is continuously increasing and is within the preset range; the speed of the output shaft of the gearbox is continuously increasing, indicating that the gearbox is loaded, and the use of data in this interval can make the judgment result of the trained model more accurate. Setting the preset range can eliminate the data of the output shaft under low-speed rotation and high-speed rotation, so that the training data only contains the data of the output shaft under stable rotation, thereby making the judgment result of the trained model more accurate.

[0081] S16: If the speed of the output shaft meets the preset condition, intercept the timing signal of each gear within the period; if the speed of the output shaft does not meet the preset condition, return to step S11. Figure 2 As shown, if the maximum speed of the output shaft of the construction machinery is 800rpm, the preset range can be set to 240rpm~480rpm, and the timing signal of the gear within the time corresponding to 240rpm~480rpm, that is, about 0.8s~1.4s, is the timing signal to be intercepted.

[0082] S17, performing equal angular domain resampling on the intercepted time series signals to obtain angular domain signals; and then performing fast Fourier transform on the angular domain signals to obtain order spectra.

[0083] S18. Calculate the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range; the statistical variables may include parameters such as mean, variance, skewness and peak-to-peak value.

[0084] S19, inputting the statistical variables into the single classification model to train the single classification model to obtain a trained model;

[0085] S2. Model testing process:

[0086] The trained model is tested using preset test data to determine a test threshold, wherein the test data includes the speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operating data of the gearbox in normal and abnormal states; by testing the trained model using the test data, a test threshold for distinguishing between normal and abnormal gearboxes can be determined. This step has the same principle as step S1, except that the data used is different. Step S1 uses training data, while this step uses test data.

[0087] S3. Model usage process:

[0088] During the operation of the gearbox, the real-time acquired data and the tested model are used to determine whether the gearbox is abnormal. The principle of this step is the same as that of step S2, but the data used is different. Step S2 uses the test data, and this step uses the real-time acquired data.

[0089] The present embodiment provides a method for identifying gearbox abnormalities, which can determine in real time whether a gearbox is abnormal. There is no need to wait until the gearbox cannot be used to manually inspect the gearbox, thereby improving the efficiency of determining whether the gearbox is abnormal and reducing the cost of determining whether the gearbox is abnormal. The training data is a zero-fault sample, which has a greater cost advantage when dealing with engineering machinery with a low failure rate and reduces the time for collecting fault samples. By using the data of the rising section as an analysis sample, the random noise caused by excessive or insufficient speed can be eliminated, and the uniformity and accuracy of the extracted statistical variables can be improved, thereby improving the accuracy of determining whether the gearbox is abnormal. The time series signal extracted using the vold-kalman algorithm is clearer than the traditional order analysis signal, and can solve the signal aliasing problem well.

[0090] Preferably, the preset condition also includes whether the speed increase rate of the output shaft is less than a first threshold value and whether the fluctuation of the speed of the output shaft is less than a second threshold value. The preset condition can be used to screen the data of the stable increase of the speed of the output shaft of the gearbox as an analysis sample, thereby eliminating the random noise caused by the unstable speed. The speed increase rate of the output shaft is equal to the difference between the speeds at two time points divided by a preset time period. The kurtosis of the speed of the output shaft can be used as the fluctuation of the speed of the output shaft.

[0091] Preferably, the step of testing the trained model using preset test data in step S2 specifically includes the following steps:

[0092] S21, acquiring test data of the model, wherein the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state;

[0093] S22, calculating the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft in the test data, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft;

[0094] S23, determining the order of each gear according to the ratio of the rotation speed of each gear to the rotation speed of the output shaft;

[0095] S24, based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, separate the timing signal of each gear by using the Vold-Kalman algorithm;

[0096] S25, determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range;

[0097] S26, if the rotation speed of the output shaft meets the preset condition, intercepting the timing signal of each gear within the period of time;

[0098] S27, respectively performing equal angular domain resampling on the intercepted time series signals to obtain angular domain signals; and then performing fast Fourier transform on the angular domain signals to obtain an order spectrum;

[0099] S28, calculating the test statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range;

[0100] S29, inputting the test statistical variables into the trained model.

[0101] By testing the trained model with test data, the test threshold for distinguishing between normal and abnormal gearbox conditions can be determined.

[0102] Preferably, the step of using the real-time acquired data and the tested model to determine whether the gearbox is abnormal in step S3 specifically includes the following steps:

[0103] S31, during the operation of the gearbox, obtaining the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox in real time;

[0104] S32, calculating the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft acquired in real time, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft;

[0105] S33, determining the order of each gear according to the ratio of the rotation speed of each gear to the rotation speed of the output shaft;

[0106] S34, based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, separate the timing signal of each gear by using the Vold-Kalman algorithm;

[0107] S35, determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range;

[0108] S36, if the speed of the output shaft obtained in real time meets the preset condition, intercepting the timing signal of each gear within the period of time;

[0109] S37, respectively performing equal angular domain resampling on the intercepted time series signals to obtain angular domain signals; and then performing fast Fourier transform on the angular domain signals to obtain an order spectrum;

[0110] S38. Calculate the real-time statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range.

[0111] S39. Input the real-time statistical variables into the tested model to determine whether the gearbox is abnormal. If the value output by the trained model is greater than or equal to the test threshold, the gearbox is determined to be abnormal; if the value output by the trained model is less than the test threshold, the gearbox is determined to be normal.

[0112] The principles of step S3 and step S1 are the same, except that step S3 uses data acquired in real time, while step S1 uses training data.

[0113] Preferably, the preset range is 30%r to 60%r, where r represents the maximum speed of the output shaft of the gearbox. 30%r to 60%r means that the gearbox is in a stable speed range, and using data in this range can reduce random noise of the data.

[0114] Preferably, step S18, i.e., the step of calculating the statistical variables corresponding to each order, specifically comprises the following steps: taking the ±1 order interval of each order as the target order interval, calculating the product between each order and the corresponding amplitude within the target order interval; and calculating the statistical variables corresponding to each order according to the product. Parameters such as the mean, variance, skewness and peak-to-peak value corresponding to the product can be used as statistical variables.

[0115] Preferably, step S13, i.e., the step of determining the order of each gear according to the ratio of the rotation speed of each gear to the rotation speed of the output shaft, specifically comprises the following steps: taking the ratio of the rotation speed of each gear to the rotation speed of the output shaft as the order of each gear. The ratio of the rotation speed of each gear to the rotation speed of the output shaft can be taken as the order of each gear, i.e., the order of each gear is calculated according to Formula 3. In other embodiments, the result calculated by Formula 3 can be multiplied by a preset coefficient as the order of the corresponding gear.

[0116] Preferably, after the step of using the real-time acquired data and the tested model to determine whether the gearbox is abnormal, the following step is further included: displaying the judgment result of the tested model. The judgment result can be displayed on a display screen for personnel to view.

[0117] Preferably, after the step of displaying the judgment result of the tested model, the method further comprises the following step: storing the judgment result of the tested model. The judgment result can be used as training data for subsequent models after being stored.

[0118] Based on Figure 1 The same technical concept as the above-mentioned method for identifying abnormality of a gearbox, this embodiment provides another method for identifying abnormality of a gearbox, wherein the gearbox comprises a plurality of gears and an output shaft, and the method comprises the following steps:

[0119] S1. Model training process:

[0120] S11, acquiring training data of the model, wherein the training data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, the model is a single classification model, and the training data is the operation data of the gearbox in a normal state;

[0121] S12, calculating the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft;

[0122] S13, determining the order of each gear according to the ratio of the rotation speed of each gear to the rotation speed of the output shaft;

[0123] S15, determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range;

[0124] S160, if the rotation speed of the output shaft meets the preset condition, intercepting the vibration data of the gearbox, the order of each gear and the rotation speed of each gear during the period of time;

[0125] S140, based on the intercepted vibration data of the gearbox, the order of each gear and the rotation speed of each gear, separate the timing signal of each gear by using the Vold-Kalman algorithm;

[0126] S17, respectively performing equal angular domain resampling on the time series signals to obtain angular domain signals; and then performing fast Fourier transform on the angular domain signals to obtain an order spectrum;

[0127] S18, calculating the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range;

[0128] S19, inputting the statistical variables into the single classification model to train the single classification model to obtain a trained model;

[0129] S2. Model testing process:

[0130] Using preset test data to test the trained model and determine a test threshold, the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state;

[0131] S3. Model usage process:

[0132] During the operation of the gearbox, the real-time acquired data and the tested model are used to determine whether the gearbox is abnormal. Figure 1 The difference of the corresponding embodiments is that Figure 1 The order of steps S14, S15, and S16 in is modified to S15, S160, and S140. Since step S15 is advanced, the data objects processed by S160 and S140 are adjusted, and the other steps remain unchanged. Figure 1 The technical effects of the corresponding embodiments are the same.

[0133] In summary, the present invention provides a gearbox abnormality identification method that can determine in real time whether the gearbox is abnormal. There is no need to wait until the gearbox cannot be used to manually inspect the gearbox, thereby improving the efficiency of determining whether the gearbox is abnormal and reducing the cost of determining whether the gearbox is abnormal. The training data is a zero-fault sample, which has a greater cost advantage when dealing with engineering machinery with a low failure rate and reduces the time for collecting fault samples. By using the data of the rising section as an analysis sample, the random noise caused by excessive or insufficient speed can be eliminated, and the uniformity and accuracy of the extracted statistical variables can be improved, thereby improving the accuracy of determining whether the gearbox is abnormal. The time series signal extracted using the vold-kalman algorithm is clearer than the traditional order analysis signal, and can solve the signal aliasing problem well.

[0134] The above description is only a description of the preferred embodiments of the present invention, and is not intended to limit the scope of the present invention. Any changes or modifications made by a person skilled in the art based on the above disclosure shall fall within the protection scope of the present invention.

Claims

1. A method for identifying abnormalities in a gearbox, wherein the gearbox comprises a plurality of gears and an output shaft, characterized in that: The method comprises the following steps: Model training process: Acquire training data of the model, the training data including the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, the model is a single classification model, and the training data is the operation data of the gearbox in a normal state; Calculate the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft; Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft; Based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, a time series signal of each gear is separated by using a Vold-Kalman algorithm; Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range; If the rotation speed of the output shaft meets the preset condition, intercepting the timing signal of each gear within the period of time; Resampling the intercepted time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra; Calculate the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range; Inputting the statistical variables into the single classification model to train the single classification model to obtain a trained model; Model testing process: Using preset test data to test the trained model and determine a test threshold, the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state; Model usage process: During the operation of the gearbox, the real-time data and the tested model are used to determine whether the gearbox is abnormal.

2. A gearbox abnormality identification method according to claim 1, characterized in that: The preset condition also includes whether the speed increase rate of the output shaft is less than a first threshold and whether the fluctuation of the speed of the output shaft is less than a second threshold.

3. A gearbox abnormality identification method according to claim 1, characterized in that: The step of testing the trained model using preset test data specifically includes the following steps: Acquire test data of the model, wherein the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state; Calculating the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft in the test data, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft; Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft; Based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, a time series signal of each gear is separated by using a Vold-Kalman algorithm; Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range; If the rotation speed of the output shaft meets the preset condition, then the timing signal of each gear within the period of time is intercepted; Resampling the intercepted time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra; Calculate the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range; The statistical variables are input into the trained model.

4. A gearbox abnormality identification method according to claim 1, characterized in that: The step of using the real-time acquired data and the tested model to determine whether the gearbox is abnormal specifically includes the following steps: During the operation of the gearbox, the speed of the output shaft of the gearbox and the vibration data of the gearbox are obtained in real time; Calculate the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft acquired in real time, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft; Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft; Based on the vibration data of the gearbox, the order of each gear and the rotation speed of each gear, a time series signal of each gear is separated by using a Vold-Kalman algorithm; Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range; If the speed of the output shaft obtained in real time meets the preset condition, then the timing signal of each gear within the period of time is intercepted; Resampling the intercepted time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra; According to the order spectrum, the amplitude corresponding to each order and the preset interval range, respectively calculate the real-time statistical variables corresponding to each order; The real-time statistical variables are input into the model after the test to determine whether the gearbox is abnormal. If the value output by the model after the test is greater than or equal to the test threshold, the gearbox is determined to be abnormal; if the value output by the model after the test is less than the test threshold, the gearbox is determined to be normal.

5. A gearbox abnormality identification method according to claim 1, characterized in that: The preset range is 30%r to 60%r, where r represents the maximum rotational speed of the output shaft of the gear box.

6. A gearbox abnormality identification method according to claim 1, characterized in that: The step of calculating the statistical variables corresponding to each order specifically includes the following steps: taking the ±1 order interval of each order as the target order interval, calculating the product between each order and the corresponding amplitude within the target order interval; and calculating the statistical variables corresponding to each order based on the product.

7. A gearbox abnormality identification method according to claim 6, characterized in that: The statistical variables include mean, variance, skewness and peak-to-peak value.

8. A gearbox abnormality identification method according to claim 1, characterized in that: The step of determining the order of each gear according to the ratio of the rotational speed of each gear to the rotational speed of the output shaft specifically includes the following steps: taking the ratio of the rotational speed of each gear to the rotational speed of the output shaft as the order of each gear.

9. A gearbox abnormality identification method according to claim 1, characterized in that: After the step of using the real-time acquired data and the tested model to determine whether the gearbox is abnormal, the method further includes the following steps: displaying the determination result of the tested model.

10. A method for identifying abnormalities in a gearbox, the gearbox comprising a plurality of gears and an output shaft, characterized in that: The method comprises the following steps: Model training process: Acquire training data of the model, the training data including the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, the model is a single classification model, and the training data is the operation data of the gearbox in a normal state; Calculate the rotation speed of each gear step by step according to the number of teeth of each gear, the number of teeth of the output shaft, the rotation speed of the output shaft, the meshing relationship between the plurality of gears, and the meshing relationship between the gears and the output shaft; Determining the order of each of the gears according to the ratio of the rotation speed of each of the gears to the rotation speed of the output shaft; Determining whether the rotation speed of the output shaft meets a preset condition, wherein the preset condition includes that the rotation speed of the output shaft continues to increase and is within a preset range; If the rotation speed of the output shaft meets the preset condition, the vibration data of the gearbox, the order of each gear and the rotation speed of each gear within the period of time are intercepted; Based on the intercepted vibration data of the gearbox, the order of each gear and the rotation speed of each gear, the time series signal of each gear is separated by using the Vold-Kalman algorithm; Resampling the time series signals in equal angle domains to obtain angle domain signals; and performing fast Fourier transform on the angle domain signals to obtain order spectra; Calculate the statistical variables corresponding to each order according to the order spectrum, the amplitude corresponding to each order and the preset interval range; Inputting the statistical variables into the single classification model to train the single classification model to obtain a trained model; Model testing process: Using preset test data to test the trained model and determine a test threshold, the test data includes the rotation speed of the output shaft of the gearbox and the vibration data of the gearbox, and the test data is the operation data of the gearbox in a normal state and an abnormal state; Model usage process: During the operation of the gearbox, the real-time data and the tested model are used to determine whether the gearbox is abnormal.

Images