Health state evaluation system and method of decelerator gear box based on multi-feature parameter fusion

Abstract

The present application relates to a kind of multi-characteristic parameter fusion's reducer gear box health state evaluation system and method, system includes test bench, test bench is provided with input motor, output motor, gear box, input motor is connected with the input gear of being arranged in gear box by input shaft, output motor is connected with the output gear of being arranged in gear box by output shaft, input gear and output gear mesh transmission, gear box is provided with the input bearing for supporting input shaft and the output bearing for supporting output shaft in inside, gear box is communicated with sliding oil system, test bench is provided with photoelectric speed sensor, photoelectric speed sensor is directly opposite input shaft, gear box surface is provided with several vibration sensors, gear box is communicated with the oil return line of sliding oil system and is provided with sliding oil dust sensor, photoelectric speed sensor, sliding oil dust sensor, vibration sensor is connected with data acquisition system, data acquisition system is connected with host computer.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of helicopter health and usage monitoring, in particular to a health state evaluation system for a reducer gearbox with multi-feature parameter fusion. BACKGROUND

[0002] The transmission system of a conventional single-rotor tail rotor configuration helicopter mainly consists of a main reducer, an intermediate reducer, a tail reducer, a main transmission shaft and a tail transmission shaft, etc. The reducer is composed of shafts, gears (including bevel gears) and bearings, etc. It is responsible for transmitting the power of the engine to the rotor and tail rotor and various accessories, and simultaneously transmitting the load of the rotor to the airframe structure. It has high complexity and integration, and belongs to a single-channel complex structure without redundancy. Once a failure occurs, it will lead to disastrous consequences. According to statistics, about 68% of flight accidents caused by helicopter mechanical failures are due to engine and transmission system failures. Therefore, it is necessary to obtain the health state of the key moving parts of the helicopter transmission system as early as possible.

[0003] At present, the main method for fault diagnosis of helicopter transmission system gearbox at home and abroad is the damage diagnosis method based on structural vibration response analysis. This method can monitor relatively serious fault states of the transmission system gearbox, but due to the influence of strong background noise interference in the complex operating environment of the helicopter, the early fault symptom discovery ability is insufficient, and it is difficult to fully and effectively make maintenance decisions. With the demand for monitoring and diagnosis technology in the industry in the 1960s, especially in the early 1970s, the advent of ferrography analysis technology, the oil debris monitoring technology began to develop rapidly. Oil debris monitoring technology has become one of the main methods in the fault diagnosis technology system, which develops together with vibration monitoring, temperature monitoring and performance parameter monitoring, and is developing towards the integration of multiple methods, online and offline, integration of monitoring, diagnosis, maintenance and management, and instrument intelligence. This method can find early fault symptoms, but it is difficult to accurately locate the fault position and judge the fault degree. SUMMARY

[0004] The purpose of the present application is to provide a health state evaluation system and method for a reducer gearbox with multi-feature parameter fusion. The oil debris count signal and vibration signal of the reducer gearbox of the test bench are collected by a new type of oil debris sensor and a vibration sensor. The "small oil debris particle cumulative total amount g(t)" state index is selected to represent the fault characteristics of the oil debris count signal, and the "gear vibration root mean square value RMS gear ", "gear vibration kurtosis K", "bearing vibration root mean square value RMS bearingThe state indexes such as "bearing vibration statistics 6th order M6A", "shaft vibration first order harmonic OM1", and "shaft vibration second order harmonic OM2" are used to represent the fault features of the vibration signals, and proper fault features are selected as input signals to perform fusion analysis through a T-S fuzzy neural network model, so that early fault diagnosis of the bevel gear of the tail transmission system of the helicopter is realized.

[0005] The technical scheme of the present application is as follows:

[0006] A multi-feature parameter fusion reducer gear box health state evaluation system, comprising a test bench, wherein an input motor, an output motor and a gear box are arranged on the test bench, the input motor is connected with an input gear arranged in the gear box through an input shaft, the output motor is connected with an output gear arranged in the gear box through an output shaft, the input gear and the output gear are in meshing transmission, input bearings for supporting the input shaft and output bearings for supporting the output shaft are arranged in the gear box, the gear box is communicated with a lubricating oil system, an optical speed sensor is arranged on the test bench and faces the input shaft, a plurality of vibration sensors are arranged on the surface of the gear box, a lubricating oil wear debris sensor is arranged on an oil return pipeline communicated with the lubricating oil system, the optical speed sensor, the lubricating oil wear debris sensor and the vibration sensors are connected with a data acquisition system, and the data acquisition system is connected with an upper computer.

[0007] Further, the vibration sensors are six in number and are arranged on the surface of the gear box and correspond to the input bearings, the output bearings, the input gear, the output gear, the input shaft and the output shaft in the gear box respectively.

[0008] A method for the multi-feature parameter fusion reducer gear box health state evaluation system, comprising the following steps:

[0009] Step 1: start the multi-feature parameter fusion reducer gear box health state evaluation system, after stable operation for a period of time, the optical speed sensor collects the speed signal of the input shaft, the lubricating oil wear debris sensor collects the lubricating oil wear debris particle signal of the oil return pipeline, the vibration sensors collect the vibration signals of the input bearings, the output bearings, the input gear, the output gear, the input shaft and the output shaft, until the multi-feature parameter fusion reducer gear box health state evaluation system appears obvious abnormality, stop and disassemble the gear box to check the health state of the gear box;

[0010] Step 2: extract the feature parameters of the lubricating oil wear debris particle signal collected by the lubricating oil wear debris sensor of the oil return pipeline;

[0011] Step 3: perform time domain synchronous average pretreatment on the vibration signals of the input bearings, the output bearings, the input gear, the output gear, the input shaft and the output shaft collected by the vibration sensors;

[0012] Step 4: Extract the characteristic parameters of the vibration signal preprocessed by time synchronous average;

[0013] Step 5: Construct a "T-S fuzzy neural network" model, and define the input space, output space and fuzzy logic model rules of the "T-S fuzzy neural network" model;

[0014] Step 6: Train and optimize the "T-S fuzzy neural network" model, and input the data for fault diagnosis analysis of the gearbox.

[0015] Further, step 2 is specifically, first, according to the sensing ability of the oil debris sensor, the oil debris particles are classified according to the size, defined as: large oil debris particles, particle size greater than 401 microns; medium oil debris particles, particle size between 201-400 microns; small oil debris particles, particle size between 125-200 microns; then calculate and extract the "small oil debris particle cumulative total g(t)" characteristic parameter from the oil debris particle signal collected by the oil debris sensor in the oil return pipeline.

[0016] Further, step 3 is specifically using the "time synchronous average TSA" method to preprocess the vibration signals collected by the 6 vibration sensors, first, according to the rotation speed signal of the input shaft collected by the photoelectric speed sensor, the vibration signal x(t) is divided according to the rotation period, and the vibration signal of each rotation period is obtained. The number of vibration signal data points in each rotation period is different, and the vibration data in each rotation period needs to be resampled, and then time synchronous average processing can be performed to obtain the time synchronous average signal, and the calculation process is shown in formula (1):

[0017]

[0018] In the formula, j is the number of time synchronous average signal data points, x TSA (j) is the time synchronous average signal, N is the rotation period, that is, the average number, L is the number of vibration data points contained in each rotation period, T s is the sampling interval, i is the number, i=1, 2, 3, …, N-1.

[0019] Further, step 4 extracts the characteristic parameters of the vibration signal preprocessed by time synchronous average, including:

[0020] (a) Calculate the "gear vibration root mean square value RMS TSA " characteristic parameter of the input gear and the output gear using the time synchronous average signal x gear obtained by formula (1), and the calculation formula is shown in formula (2):

[0021]

[0022] In the formula, j is the number of time-domain synchronous average signal data points, and m is the length of the time-domain synchronous average signal data.

[0023] (b) The time-domain synchronous average signal x obtained using equation (1) TSA (j) Calculate the characteristic parameter of "gear vibration kurtosis value K" for the input gear and the output gear. The calculation formula is shown in equation (3):

[0024]

[0025] In the formula, j is the number of time-domain synchronous average signal data points, and m is the length of the time-domain synchronous average signal data. The time-domain synchronous average signal x TSA The mean of σ is the standard deviation.

[0026] (c) The time-domain synchronous average signal x obtained using equation (1) TSA (j) Calculate the root mean square (RMS) value of bearing vibration for the input and output bearings. bearing The characteristic parameters are calculated using the formula shown in equation (4):

[0027]

[0028] In the formula, j is the number of time-domain synchronous average signal data points, and m is the length of the time-domain synchronous average signal data.

[0029] (d) The time-domain synchronous average signal x obtained using equation (1) TSA (j) The characteristic parameters of the "bearing vibration statistics of the 6th order M6A" of the input bearing and the output bearing are calculated. The calculation formula is shown in Equation (5):

[0030]

[0031] In the formula, j is the number of data points of the time-domain synchronous average signal, m is the length of the time-domain synchronous average signal data, and d is the time-domain synchronous average signal x. TSA The differential signal, d j Let be the value of the j-th data point of the differential signal d. Let d be the mean of the differential signal d.

[0032] (e) Using equation (1), the time-domain synchronous average signal x TSA (j) Perform a fast Fourier transform to calculate the characteristic parameters of the first harmonic component OM1 of the shaft vibration of the input and output shafts. The formula for calculating OM1 is shown in equation (6):

[0033]

[0034] In the formula, n is the rotational speed of the shaft at the input or output end, and f1 is the first harmonic frequency.

[0035] i) using the time domain synchronous average signal x of formula (1) TSA (j) performing fast Fourier transform to calculate the "shaft vibration second harmonic component OM2" characteristic parameter of the input shaft and output shaft, and the OM2 calculation formula is shown in formula (7):

[0036]

[0037] In the formula, n is the rotating speed of the input shaft or output shaft, and f2 is the second harmonic frequency.

[0038] Further, in step 5:

[0039] The input and output spaces of the "T-S fuzzy neural network" model are determined, and the input space membership degree is defined as:

[0040] 1) low damage, represented as DL: the state of the input bearing, output bearing, input gear, output gear, input shaft and output shaft and their connected accessories inside the gearbox after the running-in period, which does not affect the use;

[0041] 2) general damage, represented as DM: the input bearing, output bearing, input gear, output gear, input shaft and output shaft and their connected accessories inside the gearbox have slight scratch damage characteristics after a long period of operation, which does not affect the use;

[0042] 3) severe damage, represented as DH: the input bearing, output bearing, input gear, output gear, input shaft and output shaft and their connected accessories inside the gearbox have damage deformation; cannot continue to run.

[0043] According to the characteristic parameter change history diagram, the membership degrees of the following characteristic parameters are calculated:

[0044] 1) the total amount of small oil abrasive particles g(t) containing DL, DM and DH damage characteristics;

[0045] 2) gear vibration root mean square value RMS gear containing DL and DH damage characteristics;

[0046] 3) gear vibration kurtosis value K containing DL and DH damage characteristics;

[0047] 4) bearing vibration root mean square value RMS bearing containing DL and DH damage characteristics;

[0048] 5) bearing vibration statistics 6th order M6A containing DL and DH damage characteristics;

[0049] 6) shaft vibration first harmonic component OM1 containing DL and DH damage characteristics;

[0050] 7) The shaft vibration second harmonic component OM2 contains the damage feature of DL, DH.

[0051] The input data of the input space selects the total amount of small oil abrasion particles g(t) and two most sensitive vibration feature parameters of the vibration signal in step 4.

[0052] The membership of the output space is defined as the state of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft and their connecting accessories inside the gearbox, and the state recognition membership of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft and their connecting accessories inside the gearbox is predicted by using the "T-S fuzzy neural network" model:

[0053] 1) The gearbox has no gear damage, which is represented as OK;

[0054] 2) The gearbox includes initial / mild scratch / damaging pit damage, which is represented as check;

[0055] 3) The gearbox includes severe wear / damaging pit damage, which is represented as stop.

[0056] Further, the rule definition of the "T-S fuzzy neural network" model is defined, in which the two most sensitive vibration feature parameters are set as V1 and V2, and the total amount of small oil abrasion particles is set as g(t), and the rule definition is as follows:

[0057] 1) If V1 is DL, V2 is DL, and g(t) is DL, then the output is OK;

[0058] 2) If V1 is DH, V2 is DH, and g(t) is DH, then the output is stop;

[0059] 3) If V1 is DL, V2 is DL, and g(t) is DM, then the output is check;

[0060] 4) If V1 is DL, V2 is DH, and g(t) is DL, then the output is OK;

[0061] 5) If V1 is DL, V2 is DL, and g(t) is DH, then the output is check;

[0062] 6) If V1 is DH, V2 is DL, and g(t) is DL, then the output is OK;

[0063] 7) If V1 is DH, V2 is DL, and g(t) is DM, then the output is check;

[0064] 8) If V1 is DH, V2 is DH, and g(t) is DL, then the output is check;

[0065] 9) If V1 is DH, V2 is DL, g(t) is DH, then the output is stop;

[0066] 10) If V1 is DH, V2 is DH, g(t) is DM, then the output is check;

[0067] 11) If V1 is DL, V2 is DH, g(t) is DH, then the output is stop;

[0068] 12) If V1 is DL, V2 is DH, g(t) is DM, then the output is check.

[0069] Further, step 6: the T-S fuzzy neural network model is trained and optimized, and input data are subjected to fault diagnosis analysis of the gearbox, and specifically:

[0070] After the T-S fuzzy logic model is constructed, the input data of the input space are input to perfect the T-S fuzzy neural network model, and after the model is perfected, the model can be used for input data to perform fault diagnosis analysis of the gearbox.

[0071] The present application has the beneficial effects that: the present application provides a reducer gearbox health state evaluation system and method with multi-feature parameter fusion, adopts a time domain synchronous average method, can enhance the vibration signal features related to the target gear, and simultaneously reduce the interference of background noise and other components; multi-dimensional feature parameters of the oil debris signal and the vibration signal are extracted from the time domain, the frequency domain and the complexity measurement angle, the sensitivity of different feature parameters to different component fault modes is not the same, and the feature parameters that can better represent the health state of the gearbox are selected as input parameters of the T-S fuzzy logic model; through the model, the data measured by different sensors and various diagnostic professional knowledge are integrated into a system, the health state of the transmission system is judged through fusion analysis, and finally clear decision information can be provided for the end user. BRIEF DESCRIPTION OF DRAWINGS

[0072] Figure 1 It is a model test bench reducer gearbox health state evaluation system structure schematic diagram;

[0073] Figure 2 It is a full-size oil debris particle accumulation total amount change curve diagram with running time;

[0074] Figure 3 It is an output end gear vibration root mean square value RMS gear Feature parameter change curve diagram with running time;

[0075] Figure 4 It is an output end gear vibration kurtosis K feature parameter change curve diagram with running time;

[0076] Figure 5 RMS is the vibration root mean square value of the output end bearing bearing the characteristic parameter curve diagram with the running time;

[0077] Figure 6 M6A is the vibration statistical quantity 6th order of the output end bearing, and the characteristic parameter curve diagram with the running time;

[0078] Figure 7 OM1 is the first order harmonic of the output end shaft vibration, and the characteristic parameter curve diagram with the running time;

[0079] Figure 8 OM2 is the second order harmonic of the output end shaft vibration, and the characteristic parameter curve diagram with the running time;

[0080] Figure 9 the membership diagram of the input space input parameter small oil abrasion cumulative total g(t);

[0081] Figure 10 the membership diagram of the input space input parameter gear vibration root mean square value RMS gear ;

[0082] Figure 11 the membership diagram of the input space input parameter shaft vibration first order harmonic OM1;

[0083] Figure 12 the reducer gearbox state membership diagram of the output space;

[0084] Figure 13 the T-S fuzzy neural network model output result diagram.

[0085] 1, model test bench, 2, input motor, 3, output motor, 4, reducer gearbox, 5, photoelectric speed sensor, 6, vibration sensor, 7, oil abrasion sensor, 8, oil inlet pipeline, 9, oil return pipeline, 10, data acquisition system. DETAILED DESCRIPTION

[0086] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0087] As Figure 1As shown, a multi-feature parameter fusion reducer gearbox health state evaluation system, including test bench 1, the test bench 1 is provided with input motor 2, output motor 3, gearbox 4, the input motor 2 is connected with the input gear arranged in the gearbox 4 through the input shaft, the output motor 3 is connected with the output gear arranged in the gearbox 4 through the output shaft, the input gear is engaged with the output gear transmission, the gearbox 4 is provided with the input bearing for supporting the input shaft and the output bearing for supporting the output shaft in the inside, the gearbox is communicated with the oil system, the test bench 1 is provided with photoelectric speed sensor 5, the photoelectric speed sensor 5 is opposite to the input shaft, the gearbox 4 surface is provided with several vibration sensors 6, the gearbox 4 is provided with oil sludge sensor 7 on the oil return line 9 communicated with the oil system, photoelectric speed sensor 5, oil sludge sensor 7, vibration sensor 6 and data acquisition system 10 are connected, data acquisition system 10 is connected with host computer. Wherein, the vibration sensor 6 has six, is arranged on the gearbox 4 surface respectively, and is one-to-one corresponding to the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft in the gearbox.

[0088] The method of the multi-feature parameter fusion reducer gearbox health state evaluation system comprises the following steps:

[0089] Step 1: start the multi-feature parameter fusion reducer gearbox health state evaluation system, after stable operation for a period of time, the photoelectric speed sensor 5 collects the speed signal of the input shaft, the oil sludge sensor 7 collects the oil sludge particle signal of the oil return line 9, the vibration sensor 6 collects the vibration signal of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft, until the multi-feature parameter fusion reducer gearbox health state evaluation system runs obviously abnormally, stop and disassemble to check the health condition of the gearbox 4;

[0090] Specifically, in step one, the input shaft speed is controlled to be stable at 3450rpm / min, the output torque is controlled to be 95% of the total output torque, the reducer gearbox 4 selects a perfect test piece, each time the machine is started, the stable state is run for 120min, the vibration data is collected every 10min, each time the data is collected for 2min, the sampling rate is 20KHz, the data acquisition system 10 collects data throughout the process, the sampling rate is 1Hz, the machine does not stop in the middle, disassembles and checks when abnormal situation occurs, and disassembles and checks regularly for 20 hours.

[0091] Step 2: Extract the characteristic parameters of the oil abrasive particle signal collected by the oil abrasive sensor 7 from the oil return pipeline 9; specifically, first, according to the sensing ability of the oil abrasive sensor 7, classify the oil abrasive particles by size, defined as: large oil abrasive particles, particle size greater than 401 microns; medium oil abrasive particles, particle size between 201-400 microns; small oil abrasive particles, particle size between 125-200 microns; then calculate and extract the "small oil abrasive particle cumulative total g(t)" characteristic parameter from the oil abrasive particle signal collected by the oil abrasive sensor 7 from the oil return pipeline 9, the full-size oil abrasive particle cumulative total vs. running time curve is shown in FIG. Figure 2 .

[0092] Step 3: Time domain synchronous average preprocessing of the vibration signals collected by the vibration sensor 6 from the input bearing, output bearing, input gear, output gear, input shaft and output shaft; specifically, using the "time domain synchronous average TSA" method to preprocess the vibration signals collected by the six vibration sensors 6, first, according to the rotation speed signal of the input shaft collected by the photoelectric rotation speed sensor 5, divide the vibration signal x(t) by the rotation period to obtain the vibration signal of each rotation period, the number of vibration signal data points in each rotation period is different, the vibration data in each rotation period needs to be resampled, then the time domain synchronous average processing can be obtained, the calculation process is shown in formula (1):

[0093]

[0094] In the formula, j is the number of time domain synchronous average signal data points, x TSA (j) is the time domain synchronous average signal, N is the rotation period, i.e. the average number, L is the number of vibration data points contained in each rotation period, T s is the sampling interval, i is the number, i = 1, 2, 3, …, N-1.

[0095] Step 4: Extract the characteristic parameters of the vibration signal after time domain synchronous average preprocessing;

[0096] Step 4 extracts the characteristic parameters of the vibration signal after time domain synchronous average preprocessing, including:

[0097] (a) Calculate the "gear vibration root mean square value RMS TSA " characteristic parameter of the input gear and output gear using the time domain synchronous average signal x gear (j) obtained by formula (1), the calculation formula is shown in formula (2):

[0098]

[0099] In the formula, j is the number of time-domain synchronous average signal data points, m is the length of the time-domain synchronous average signal data, and the RMS value of the output gear vibration is... gear The curves showing the change of characteristic parameters over runtime, such as... Figure 3 As shown.

[0100] (b) The time-domain synchronous average signal x obtained using equation (1) TSA (j) Calculate the characteristic parameter of "gear vibration kurtosis value K" for the input gear and the output gear. The calculation formula is shown in equation (3):

[0101]

[0102] In the formula, j is the number of time-domain synchronous average signal data points, and m is the length of the time-domain synchronous average signal data. The time-domain synchronous average signal x TSA The mean value, σ is the standard deviation, and the curve of the characteristic parameter K of the output gear vibration changes with running time is shown in the figure. Figure 4 As shown.

[0103] (c) The time-domain synchronous average signal x obtained using equation (1) TSA (j) Calculate the root mean square (RMS) value of bearing vibration for the input and output bearings. bearing The characteristic parameters are calculated using the formula shown in equation (4):

[0104]

[0105] In the formula, j is the number of time-domain synchronous average signal data points, m is the length of the time-domain synchronous average signal data, and RMS value of bearing vibration is... bearing The curves showing the change of characteristic parameters over runtime, such as... Figure 5 As shown.

[0106] (d) The time-domain synchronous average signal x obtained using equation (1) TSA (j) The characteristic parameters of the "bearing vibration statistics of the 6th order M6A" of the input bearing and the output bearing are calculated. The calculation formula is shown in Equation (5):

[0107]

[0108] In the formula, j is the number of data points of the time-domain synchronous average signal, m is the length of the time-domain synchronous average signal data, and d is the time-domain synchronous average signal x. TSA The differential signal, d j Let be the value of the j-th data point of the differential signal d. The graph shows the variation of the 6th-order M6A characteristic parameter of the output bearing vibration statistic with operating time, representing the mean of the differential signal d. Figure 6 As shown.

[0109] (e) using the time domain synchronous average signal x of formula (1) TSA (j) performing fast Fourier transform, calculating the "shaft vibration first harmonic component OM1" characteristic parameters of the input shaft and the output shaft, and the OM1 calculation formula is shown in formula (6):

[0110]

[0111] In the formula, n is the speed of the input shaft or the output shaft, f1 is the first harmonic frequency, and the output shaft vibration first harmonic OM1 characteristic parameter curve with the change of the running time is shown in Figure 7 .

[0112] i) using the time domain synchronous average signal x of formula (1) TSA (j) performing fast Fourier transform, calculating the "shaft vibration second harmonic component OM2" characteristic parameters of the input shaft and the output shaft, and the OM2 calculation formula is shown in formula (7):

[0113]

[0114] In the formula, n is the speed of the input shaft or the output shaft, f2 is the second harmonic frequency, and the output shaft vibration second harmonic OM2 characteristic parameter curve with the change of the running time is shown in Figure 8 .

[0115] Step 5: Constructing a "T-S fuzzy neural network" model, and defining the input space, output space and fuzzy logic model rules of the "T-S fuzzy neural network" model.

[0116] Determine the input and output space of the "T-S fuzzy neural network" model, wherein the input space membership degree is defined as:

[0117] 1) Low damage, represented as DL: the state of the input bearing, output bearing, input gear, output gear, input shaft and output shaft inside the gearbox and their connected accessories after the running-in period, which does not affect the use;

[0118] 2) General damage, represented as DM: the input bearing, output bearing, input gear, output gear, input shaft and output shaft inside the gearbox and their connected accessories have slight scratch damage after a long period of operation, which does not affect the use;

[0119] 3) Serious damage, represented as DH: the input bearing, output bearing, input gear, output gear, input shaft and output shaft inside the gearbox and their connected accessories have damage deformation; cannot continue to run.

[0120] According to the characteristic parameter curve with time, the membership degrees of the following characteristic parameters are calculated:

[0121] 1) Damage characteristics of the total cumulative amount of small lubricating oil wear particles (g(t)) including DL, DM, and DH;

[0122] 2) Root mean square value (RMS) of gear vibration gear Damage characteristics containing DL and DH;

[0123] 3) Damage characteristics of gear vibration kurtosis value K including DL and DH;

[0124] 4) Root mean square value (RMS) of bearing vibration bearing Damage characteristics containing DL and DH;

[0125] 5) Damage characteristics of bearing vibration statistics, order 6 M6A including DL and DH;

[0126] 6) Damage characteristics of the first harmonic component OM1 of shaft vibration containing DL and DH;

[0127] 7) Damage characteristics of the second harmonic component OM2 of shaft vibration containing DL and DH.

[0128] The input data for the input space are selected from the total cumulative amount of small lubricating oil wear particles g(t) and the two most sensitive vibration characteristic parameters among the characteristic parameters of the vibration signal mentioned in step 4.

[0129] The membership diagram of the input space parameter, the total cumulative amount of small lubricating oil wear debris g(t), is shown below. Figure 9 As shown, the input space parameter is the root mean square value (RMS) of gear vibration. gear Membership diagram as follows Figure 10 As shown, the membership diagram of the first harmonic OM1 of the input space axis vibration is as follows. Figure 11 As shown.

[0130] The membership degree of the output space is defined as the state of the input bearing, output bearing, input gear, output gear, input shaft, and output shaft and their connecting accessories inside the gearbox. The membership degrees of the input bearing, output bearing, input gear, output gear, input shaft, and output shaft and their connecting accessories inside the gearbox, predicted using the "T-S fuzzy neural network" model, are as follows:

[0131] 1) No gear damage in the gearbox indicates OK;

[0132] 2) Gearboxes with initial / minor scratches / destructive dents are indicated for inspection;

[0133] 3) Gearboxes with severe wear / destructive pitting damage are indicated as stopped.

[0134] The state membership diagram of the output space reducer gearbox is as follows: Figure 12 As shown.

[0135] Further, the "T-S fuzzy neural network" model rule definition, wherein the two most sensitive vibration characteristic parameters are selected as V1 and V2, and the small oil abrasive particle accumulation total amount characteristic parameter is g(t), and the rule definition is as follows:

[0136] 1) If V1 is DL, V2 is DL, and g(t) is DL, then the output is OK;

[0137] 2) If V1 is DH, V2 is DH, and g(t) is DH, then the output is stop;

[0138] 3) If V1 is DL, V2 is DL, and g(t) is DM, then the output is check;

[0139] 4) If V1 is DL, V2 is DH, and g(t) is DL, then the output is OK;

[0140] 5) If V1 is DL, V2 is DL, and g(t) is DH, then the output is check;

[0141] 6) If V1 is DH, V2 is DL, and g(t) is DL, then the output is OK;

[0142] 7) If V1 is DH, V2 is DL, and g(t) is DM, then the output is check;

[0143] 8) If V1 is DH, V2 is DH, and g(t) is DL, then the output is check;

[0144] 9) If V1 is DH, V2 is DL, and g(t) is DH, then the output is stop;

[0145] 10) If V1 is DH, V2 is DH, and g(t) is DM, then the output is check;

[0146] 11) If V1 is DL, V2 is DH, and g(t) is DH, then the output is stop;

[0147] 12) If V1 is DL, V2 is DH, and g(t) is DM, then the output is check.

[0148] Step 6: The "T-S fuzzy neural network" model is trained and optimized, and the input data is used for fault diagnosis analysis of the gearbox, which is specifically:

[0149] The T-S fuzzy logic model is built, the input data of the input space are trained, and the "T-S fuzzy neural network" model is perfected. After the model is perfected, it can be used for input data to perform fault diagnosis analysis of the gearbox. The T-S fuzzy neural network model output result graph is shown in Figure 13 .

[0150] The above merely describes specific embodiments of the present application, and the detailed description of the present application is not exhaustive of the conventional technology. However, the protection scope of the present application is not limited to this, and any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. The protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A multi-characteristic parameter fusion method for evaluating the health state of a reducer gear box, characterized in that, The health state evaluation system of the reducer gearbox using multi-feature parameter fusion is executed, and the system comprises a test bench, wherein an input motor, an output motor and a gearbox are arranged on the test bench; the input motor is connected with an input gear arranged in the gearbox through an input shaft; the output motor is connected with an output gear arranged in the gearbox through an output shaft; the input gear and the output gear are in mesh transmission; an input bearing for supporting the input shaft and an output bearing for supporting the output shaft are arranged in the gearbox; the gearbox is communicated with a lubricating oil system; an optical-electric rotating speed sensor is arranged on the test bench and faces the input shaft; a plurality of vibration sensors are arranged on the surface of the gearbox; a lubricating oil wear sensor is arranged on an oil return pipeline communicated with the lubricating oil system; the optical-electric rotating speed sensor, the lubricating oil wear sensor and the vibration sensors are connected with a data acquisition system; and the data acquisition system is connected with an upper computer. The evaluation method comprises the following steps: Step 1: start the health state evaluation system of the reducer gearbox using multi-feature parameter fusion; after stable operation for a period of time, the optical-electric rotating speed sensor collects the rotating speed signal of the input shaft, the lubricating oil wear sensor collects the lubricating oil wear particle signal of the oil return pipeline, the vibration sensors collect the vibration signals of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft, until the health state evaluation system of the reducer gearbox using multi-feature parameter fusion is abnormally operated, the test bench is stopped and the gearbox is disassembled and inspected to check the health state of the gearbox; Step 2: extract the characteristic parameters of the lubricating oil wear particle signal collected by the lubricating oil wear sensor on the oil return pipeline; Step 3: perform time domain synchronous average pretreatment on the vibration signals collected by the vibration sensors on the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft; Step 4: extract the characteristic parameters of the vibration signals after time domain synchronous average pretreatment; Step 5: construct a "T-S fuzzy neural network" model and define the input space, the output space and the fuzzy logic model rule of the "T-S fuzzy neural network" model, specifically: determine the input and output spaces of the "T-S fuzzy neural network" model, wherein the membership degree of the input space is defined as: 1) low damage, represented as DL: the state of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft in the gearbox and their connected accessories after the running-in period, which does not affect the use; 2) general damage, represented as DM: the state of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft in the gearbox and their connected accessories after a long period of operation, which has slight scratch damage and does not affect the use; 3) severe damage, represented as DH: the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft in the gearbox and their connected accessories are damaged and deformed; cannot continue to run; The membership degrees of the following characteristic parameters are calculated according to the time variation history diagram of each characteristic parameter: 1) Total amount of small oil sludge particles accumulated Damage signatures containing DL, DM, DH; 2) Gear vibration root mean square value Damage features containing DL, DH; 3) Gearbox vibration kurtosis Damage features containing DL, DH; 4) Bearing vibration root mean square value Damage signatures containing DL, DH; 5) Bearing vibration statistics 6th order Damage features containing DL, DH; 6) Shaft vibration first harmonic component Damage features containing DL, DH; 7) Shaft vibration second harmonic component Damage features containing DL, DH; Input data of the input space select small oil abrasion particle cumulative total amount and the two most sensitive vibration characteristic parameters of the vibration signal mentioned in step 4 The membership degree of the output space is defined as the state of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft and their connecting accessories inside the gearbox, and the state recognition membership degree of the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft and their connecting accessories inside the gearbox predicted by the T-S fuzzy neural network model is: 1) the gearbox has no gear damage, denoted as OK; 2) the gearbox includes initial / mild scratch / destroyed pit damage, denoted as check; 3) the gearbox includes severe wear / destroyed pit damage, denoted as stop; The model rule definition of "T-S fuzzy neural network", in which the two most sensitive vibration characteristic parameters are set as V1 and V2, and the total amount of small oil abrasive particles is set as The rule definition is as follows: 1) if V1 is DL, V2 is DL, is DL, then the output is OK; 2) if V1 is DH, V2 is DH, is DH, then the output is stop; 3) if V1 is DL, V2 is DL, is DM, then the output is check; 4) if V1 is DL, V2 is DH, is DL, then the output is OK; 5) if V1 is DL, V2 is DL, is DH, then the output is check; 6) if V1 is DH, V2 is DL, is DL, then the output is OK; 7) if V1 is DH, V2 is DL, is DM, then output check; 8) if V1 is DH, V2 is DH, is DL, then the output is check; 9) if V1 is DH, V2 is DL, is DH, then the output is stop; 10) if V1 is DH, V2 is DH, is DM, then the output is check; 11) if V1 is DL, V2 is DH, is DH, then the output is stop; 12) if V1 is DL, V2 is DH, is DM, then the output is check; Step 6: the T-S fuzzy neural network model is trained and optimized, and the input data are subjected to fault diagnosis analysis of the gearbox.

2. The method of claim 1, wherein: The six vibration sensors are arranged on the surface of the gearbox and correspond to the input bearing, the output bearing, the input gear, the output gear, the input shaft and the output shaft in the gearbox.

3. The method of claim 2, wherein, The step 2 is specifically, firstly, according to the sensing ability of the oil abrasive sensor, the oil abrasive particles are classified according to the size, defined as: large oil abrasive particles, particle size greater than 401 microns; medium oil abrasive particles, particle size between 201-400 microns; small oil abrasive particles, particle size between 125-200 microns; then the "small oil abrasive particle cumulative total amount ” characteristic parameters are calculated and extracted from the oil abrasive particle signal of the oil return pipeline collected by the oil abrasive sensor.

4. The method of claim 3, wherein, The step 3 is specifically to preprocess the vibration signals collected by the six vibration sensors by using a "time domain synchronous average (TSA)" method. First, according to the rotation speed signal of the input shaft collected by the photoelectric rotation speed sensor, the vibration signals are divided into different rotation periods, and the vibration signals in each rotation period are resampled and then subjected to time domain synchronous average processing to obtain time domain synchronous average signals. The calculation process is shown in formula (1): The vibration signals are divided according to the rotation period to obtain vibration signals in each rotation period. The number of data points of the vibration signals in each rotation period is different. The vibration data in each rotation period is resampled, and then time domain synchronous average processing is performed to obtain time domain synchronous average signals. The calculation process is shown in formula (1): （1） In the formula, is the number of time-domain synchronous average signal data points, is the time-domain synchronous average signal, is the rotation cycle, i.e., the average number of times, is the number of vibration data points contained in each rotation cycle, is the sampling interval, is the number of times, .

5. The method of claim 4, wherein, The characteristic parameters of the time domain synchronous average preprocessed vibration signal in step 4 include: (a) The time-domain synchronous average signal obtained using equation (1) Calculate the root mean square value of gear vibration for the input and output gears. The characteristic parameters are calculated using the formula shown in equation (2): （2） In the formula, is the number of time-domain synchronous average signal data points, is the length of the time-domain synchronous average signal data; (b) obtaining a time domain synchronous average signal from the signal of formula (1) The "gear vibration kurtosis value" of the input gear and the output gear characteristic parameter, and the formula is shown in formula (3): （3） wherein is the number of time-domain synchronous average signal data points, is the length of the time-domain synchronous average signal data, is the time-domain synchronous average signal is the mean value of the time-domain synchronous average signal, is the standard deviation; (c) the time domain synchronous averaged signal obtained from equation (1) The "bearing vibration root mean square value" characteristic parameter of the input bearing and the output bearing is calculated, and the formula is as shown in equation (4): ​ （4） In the formula, is the number of time-domain synchronous average signal data points, is the length of the time-domain synchronous average signal data; (d) using the time domain synchronous averaged signal of formula (1) The "bearing vibration statistics 6th order" characteristic parameter of the input bearing and the output bearing is calculated, and the calculation formula is shown in formula (5): "bearing vibration statistics 6th order" characteristic parameter of the input bearing and the output bearing is calculated, and the calculation formula is shown in formula (5): （5） wherein is the number of data points of the time domain synchronous average signal, is the length of the time domain synchronous average signal, is the time domain synchronous average signal is the difference signal of the time domain synchronous average signal, is the mean value of the difference signal is the value of the difference signal is the value of the difference signal is the mean value of the difference signal is the mean value of the difference signal (e) a time domain synchronous average signal of the form (1) The first order harmonic components of the shaft vibrations of the input and output shafts are calculated by performing a fast Fourier transform characteristic parameters, The calculation formula is shown in equation (6): （6） wherein is the rotational speed of the input or output shaft, is the first harmonic frequency; i) a time domain synchronous average signal of formula (1) a fast Fourier transform is performed to calculate the "shaft vibration second harmonic component" of the input shaft and the output shaft characteristic parameter, The calculation formula is shown in formula (7): （7） wherein is the rotational speed of the input or output shaft, is the second harmonic frequency.

6. The method of claim 5, wherein, Step 6: the T-S fuzzy neural network model is trained and optimized, and the input data are subjected to fault diagnosis analysis of the gearbox, specifically: After the T-S fuzzy logic model is constructed, the input data of the input space can be used to perfect the T-S fuzzy neural network model, and after the model is perfected, it can be used for input data to perform fault diagnosis analysis of the gearbox.

Images