CN102759448A - Gearbox fault detection method based on flexible time-domain averaging - Google Patents

Abstract

The invention discloses a gearbox fault detection method based on flexible time-domain averaging. The gearbox fault detection method comprises the steps of: 1, adsorbing an acceleration transducer on a bearing end cover of a gearbox to be detected; 2, acquiring vibration signals of the gearbox through data acquisition equipment to solve a frequency-domain discrete sampling value of the vibration signals; 3, calculating a continuous time expression of time-domain averaging; 4, converting the continuous time expression into a trigonometric function expansion form; 5, weighting ak to realize filtering, de-noising and order information extracting for the time-domain averaging; 6, carrying out discrete sampling to obtain average sequences of discrete time domains; and 7, observing periodical impacts occurring in the sequences obtained in 6, and judging the type and the severity of a gearbox fault. According to the invention, truncation errors are inhibited effectively, and the advantages of increasing the resolution ratio of the signals and increasing the detection efficiency for the gearbox are achieved while a de-noising effect is achieved.

Description

Gearbox fault detection method based on flexible time domain average

Technical field

The invention belongs to the fault diagnosis technology field of gear case, particularly based on the gearbox fault detection method of flexible time domain average.

Background technology

Gear case detection method based on time domain average is a kind of effective means of extracting cycle impact signal composition from the signal of noise.This method can be extracted the vibration information of certain pivoting part separately from the complicated mechanical vibration signal, therefore, this method is widely used in the fault diagnosis of gear case.

Yet there are following two kinds of defectives in tradition based on the gear case detection method of time domain average:

1) in the fault detect of gear case, the SF of vibration signal must be the integral multiple of gear gyrofrequency to be detected, otherwise will cause the truncation error of signal, influences the detection accuracy rate of gear distress.Yet this condition is difficult to satisfy in the reality test.

2) mathematical model of traditional time domain average fault detection method can use the comb filter model to characterize, the fundamental frequency that its time domain output waveform is a periodic signal and the comprehensive contribution of all order harmonic components.Therefore in application; Tradition time domain average fault detection method both can't extract the contribution of some order to time domain waveform separately according to the actual requirements; Also can't realize functions such as signal noise silencing, interpolation, filtering, and above-mentioned functions institute's active demand during gear distress detects exactly.

This two aspects problem largely limit the application of existing method in fields such as mechanical fault diagnosis.

From at present domestic and international present Research, still have nothing to do in the related patent U.S. Patent No. issue of time domain average method.Though have document that first kind of defective of traditional time domain average method improved,, all be based on the phase compensating method of polynomial interpolation from essence; In order to reach higher compensation precision; Often need high-order interpolation, its calculated amount is big, is difficult to handle in real time.In addition, there is not document all second kind of defective of traditional time domain average detection method not improved at present.

Summary of the invention

In order to overcome above-mentioned existing shortcoming; The object of the present invention is to provide gearbox fault detection method, truncation error has been carried out effective inhibition, when reaching de-noising effect based on flexible time domain average; Improve the resolution of signal, improved the detection efficiency of gear case.

In order to achieve the above object, the technical scheme taked of the present invention is:

Gearbox fault detection method based on flexible time domain average may further comprise the steps:

Step 1, degree of will speed up sensor are adsorbed on the bearing (ball) cover position of gear case to be detected, through data acquisition equipment the vibration signal of gear case are gathered, and vibration signal is designated as x [n];

Step 2 adopts the chirpZ conversion shown in the formula (1), tries to achieve the frequency domain discrete sampling value of vibration signal;

$a_k=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·z_k^{-n}=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·e^{-\mathrm{j\Δ\ωkn}}=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·e^{-j2\mathrm{\πkn\Δt}/T_0}---\left(1\right)$

Wherein: x [n]---vibration signal;

a _k--the frequency domain discrete sampling value of-vibration signal;

Δ ω---normalization frequency domain sample at interval;

The data length of N---vibration signal;

T ₀--the cycle of-signal;

The Δ t---signals sampling time interval;

Step 3, a that formula (1) is calculated _kBring formula (2) into, obtain expression formula continuous time of time domain average;

$\tilde{x}\left(t\right)=\underset{k=-L}{\overset{L}{\mathrm{\Σ}}}\frac{a_k\·\mathrm{\Δt}}{{\mathrm{NT}}_0}{e}^{\mathrm{jk}{\mathrm{\ω}}_{0}t}=\underset{k=-L}{\overset{L}{\mathrm{\Σ}}}{c}_k{e}^{\mathrm{jk}{\mathrm{\ω}}_{0}t}---\left(2\right)$

Wherein:

--expression formula continuous time of-time domain average;

c _k---

The complex exponential Fourier coefficients;

ω ₀--the angular frequency of-periodic signal, ω ₀=2 π/T ₀

The L---summation upper limit;

Step 4 converts formula (2) into trigonometric function and launches form, shown in formula (3);

Wherein: a ₀--the DC component of-signal, a ₀=c ₀

a _k--the amplitude of-k subharmonic, a _k=2|c _k|;

--the phase place of-k subharmonic,

Step 5 is through to a _kCarry out weighted, realize filtering, de-noising and order information extraction to time domain average, its practical implementation method is: through choosing noise threshold, and will be less than a of threshold value _kZero setting, the de-noising function of realization time domain average; Through keeping a that specifies order _k, and with other a _kZero setting, the order abstraction function of realization time domain average; Through keeping a in the specific order scope _k, the filter function of realization time domain average;

Step 6 is carried out discrete sampling to the formula after the weighted (3), obtains the discrete time-domain mean sequence;

Step 7 through observing the periodic shock that occurs in the sequence that is obtained by step 6, is judged the type and the order of severity of gearbox fault, detects thereby accomplish gearbox fault.

The present invention has following beneficial effect than prior art:

A) the present invention can effectively overcome the influence of truncation error computational accuracy than classic method, for the identification of Weak fault provides safeguard.

B) with respect to classic method, this method can further realize functions such as the interpolation, filtering, de-noising, order extraction of time domain average, for the accurate judgement of fault type provides foundation.

C) implementation algorithm of this method is based on Fast Fourier Transform (FFT), has higher counting yield, therefore can realize the on-line analysis and the diagnosis of gear box arrangement.

Description of drawings

Fig. 1 is an embodiment experiment table structural representation.

Fig. 2 is the original vibration signal of embodiment case crush fault.

Fig. 3 is the output waveform of embodiment.

Fig. 4 is traditional time domain average method output waveform for embodiment adopts.

Embodiment

Below in conjunction with accompanying drawing and embodiment the present invention is done detailed description.

Case crush fault diagnosis with the gear case experiment table is an example; This gear case experiment table is made up of drive motor 1, first gear 2, second gear 3, the 3rd gear 4, the 4th gear 5, detent 6 parts; As shown in Figure 1, the output shaft of drive motor 1 is connected with first gear 2, first gear 2 and 3 engagements of second gear; Second gear 3 and the 3rd gear 4 are installed on the same transmission shaft; The 3rd gear 4 and the engagement of the 4th gear 5, the transmission shaft of the 4th gear 5 is connected with detent 6, and the number of teeth of first gear 2, second gear 3, the 3rd gear 4, the 4th gear 5 is followed successively by 32,80,48,64.First gear 2 is the fault tooth, and peeling off appears in one of them flank of tooth, is the fault that needs detection in this test.Vibration acceleration sensor A is adsorbed in the bearing (ball) cover position near first gear 2 in test, and is as shown in fig. 1, and the SF of vibration signal is 4000Hz, and the rotating speed of drive motor 1 is 10Hz.According to the ratio of gear relation, the first order and second level meshing frequency are respectively 320Hz and 192Hz.

The original vibration signal of gear case experiment table is as shown in Figure 2, from the waveform of signal, is difficult to find the caused impact characteristic of fault.

In order to discern fault signature, adopt the present invention that case crush is detected.

Gearbox fault detection method based on flexible time domain average may further comprise the steps:

Step 1, degree of will speed up sensors A are adsorbed on the bearing (ball) cover position near first gear 2, through data acquisition equipment the vibration signal of gear case are gathered, and vibration signal is designated as x [n], and its original vibration signal is as shown in Figure 2;

Step 2 is with the chirpZ conversion shown in gear case vibration signal x [n] the substitution formula (1) that records, the frequency domain discrete sampling value of vibration signal;

$a_k=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·z_k^{-n}=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·e^{-\mathrm{j\Δ\ωkn}}=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·e^{-j2\mathrm{\πkn\Δt}/T_0}---\left(1\right)$

Wherein: x [n]---vibration signal;

a _k--the frequency domain discrete sampling value of-vibration signal;

Δ ω---normalization frequency domain sample at interval;

The data length of N---vibration signal;

T ₀--the cycle of-signal;

The Δ t---signals sampling time interval;

Step 3, a that formula (1) is calculated _kBring formula (2) into, obtain expression formula continuous time of time domain average;

$\tilde{x}\left(t\right)=\underset{k=-L}{\overset{L}{\mathrm{\Σ}}}\frac{a_k\·\mathrm{\Δt}}{{\mathrm{NT}}_0}{e}^{\mathrm{jk}{\mathrm{\ω}}_{0}t}=\underset{k=-L}{\overset{L}{\mathrm{\Σ}}}{c}_k{e}^{\mathrm{jk}{\mathrm{\ω}}_{0}t}---\left(2\right)$

Wherein:

--expression formula continuous time of-time domain average;

c _k---

The complex exponential Fourier coefficients;

ω ₀--the angular frequency of-periodic signal, ω ₀=2 π/T ₀

The L---summation upper limit;

Step 4 converts formula (2) into trigonometric function and launches form, shown in formula (3);

Wherein: a ₀--the DC component of-signal, a ₀=c ₀

a _k--the amplitude of-k subharmonic, a _k=2|c _k|;

--the phase place of-k subharmonic,

Step 5 is through to a _kCarry out weighted, realize filtering, de-noising and order information extraction, in this detects,, keep time domain average result's 16-48 order, promptly only to a for the cycle of accurately extracting in the signal is impacted characteristic to time domain average _k, k=16-48 keeps, and to other a _kCarry out zero setting;

Step 6 is carried out discrete sampling to the formula after the weighted (3), obtains the discrete time-domain mean sequence, and is as shown in Figure 3;

Step 7 through observing the periodic shock that occurs in the sequence that is obtained by step 6, is judged the type and the order of severity of gearbox fault; In this test; Observation through to Fig. 3 can be found, occurs tangible impact phenomenon in the output sequence, and is spaced apart 0.1 second fault-time; Consistent with the gyration period of first gear 2, thus judge that there is fault in first gear 2.

In order to further specify validity of the present invention; Fig. 4 has provided the output sequence that adopts traditional time domain average gearbox fault detection method to obtain; As can beappreciated from fig. 4; Because truncation error exists, classic method is difficult to provide the impact characteristic that case crush causes, thereby can't effectively discern and judge the fault that first gear 2 exists.

Claims (1)

1. based on the gearbox fault detection method of flexible time domain average, it is characterized in that, may further comprise the steps:

Step 1, degree of will speed up sensor are adsorbed on the bearing (ball) cover position of gear case to be detected, through data acquisition equipment the vibration signal of gear case are gathered, and vibration signal is designated as x [n];

Step 2 adopts the chirp transform shown in the formula (1), tries to achieve the frequency domain discrete sampling value of vibration signal;

$a_k=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·z_k^{-n}=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·e^{-\mathrm{j\Δ\ωkn}}=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}x\left[n\right]\·e^{-j2\mathrm{\πkn\Δt}/T_0}---\left(1\right)$

Wherein: x [n]---vibration signal;

a _k--the frequency domain discrete sampling value of-vibration signal;

Δ ω---normalization frequency domain sample at interval;

The data length of N---vibration signal;

T ₀--the cycle of-signal;

The Δ t---signals sampling time interval;

Step 3, a that formula (1) is calculated _kBring formula (2) into, obtain expression formula continuous time of time domain average;

$\tilde{x}\left(t\right)=\underset{k=-L}{\overset{L}{\mathrm{\Σ}}}\frac{a_k\·\mathrm{\Δt}}{{\mathrm{NT}}_0}{e}^{\mathrm{jk}{\mathrm{\ω}}_{0}t}=\underset{k=-L}{\overset{L}{\mathrm{\Σ}}}{c}_k{e}^{\mathrm{jk}{\mathrm{\ω}}_{0}t}---\left(2\right)$

Wherein: --expression formula continuous time of-time domain average;

c _k---

The complex exponential Fourier coefficients;

ω ₀--the angular frequency of-periodic signal, ω ₀=2 π/T ₀

The L---summation upper limit;

Step 4 converts formula (2) into trigonometric function and launches form, shown in formula (3);

Wherein: a ₀--the DC component of-signal, a ₀=c ₀

a _k--the amplitude of-k subharmonic, a _k=2|c _k|;

--the phase place of-k subharmonic,

Step 5 is through to a _kCarry out weighted, realize filtering, de-noising and order information extraction to time domain average, its practical implementation method is: through choosing noise threshold, and will be less than a of threshold value _kZero setting, the de-noising function of realization time domain average; Through keeping a that specifies order _k, and with other a _kZero setting, the order abstraction function of realization time domain average; Through keeping a in the specific order scope _k, the filter function of realization time domain average;

Step 6 is carried out discrete sampling to the formula after the weighted (3), obtains the discrete time-domain mean sequence;

Step 7 through observing the periodic shock that occurs in the sequence that is obtained by step 6, is judged the type and the order of severity of gearbox fault, detects thereby accomplish gearbox fault.

Images