CN107963239A - A kind of booster failure detection device and detection method based on audio - Google Patents

Abstract

The invention relates to an audio-based carrier rocket failure detection device and detection method. The sound sensor array receives the sound signal of the rocket product to be tested, and performs signal acquisition and processing on the sound signal; after wavelet transform and Fourier transform, the characteristic value is carried out. Extraction; compare the eigenvalues with the eigenvalues in the sound library during normal operation, if any eigenvalue exceeds the limited threshold, it is judged that there is a fault in the product to be tested on the rocket; compare the eigenvalues with the corresponding typical faults in the sound library If it is within the envelope of the sound characteristics of a typical fault, it is judged that the fault type is the typical fault. The invention detects the fault through the sound signal, realizes the non-contact test of the rocket, reduces the test circuit inside the rocket, avoids the problems caused by the fault of the test circuit on the rocket, and reduces the weight of the rocket at the same time.

Description

An audio-based launch vehicle fault detection device and detection method

technical field

The invention relates to an audio frequency-based carrier rocket fault detection device and detection method, belonging to the field of fault detection.

Background technique

Diagnosing the system health status of the launch vehicle through testing is the basic means to ensure the successful launch of the rocket. In the past test technology, in order to diagnose the health status of the system, it is often necessary to add measurement points for signal acquisition in the system, whether it is sensitive pressure or acquisition The current, test channel pair and the system itself are a "cecal design", that is, they are in a non-working state after going to heaven. At the same time, the failure of this contact design itself will also have a certain impact on the system.

As one of the characteristics of the system when it is working, the sound is recognizable to a certain extent and can reflect the working status of the equipment to a certain extent. Since the sound must contain various dynamic information that changes with time during the operation of the equipment, and under the conditions of relatively mature digital signal technology and fault feature extraction technology, the acquisition and processing of rocket sound signals has become a reality. The process is relatively convenient and fast, and the acquisition equipment is simple and reliable, and easy to implement. These conditions have laid the foundation for the use of sound testing to diagnose rocket faults.

With the development of rocket technology and the continuous increase of equipment on the rocket, the rapid and accurate collection and capture of system failure information has become an important basis for the success of rocket testing and launching. Sound is a kind of energy release when the equipment is working, and the audio signal must imply the working status information of the system itself. How to use the audio signal sent by the product to test and determine the working state of the product is a technical problem to be solved urgently in this field.

Contents of the invention

The purpose of the present invention is to overcome the deficiencies of the prior art, to provide a carrier rocket failure detection device and detection method based on audio, in the rocket test process, by extracting features from the equipment sound signal and performing data analysis to determine whether the equipment exists Fault.

The object of the invention is achieved through the following technical solutions:

Provide an audio-based launch vehicle failure detection device, including an acoustic sensor array, a signal acquisition module and a signal processing module;

The sound sensor array receives the sound signal of the product to be tested on the rocket, and sends it to the signal processing module after being processed by the signal acquisition module; the signal processing module extracts the characteristic value of the sound signal to perform fault discrimination.

Preferably, the characteristic value includes frequency, amplitude or phase.

Preferably, the fault discrimination method of the signal processing module is: comparing the eigenvalues with the eigenvalues in the sound library during normal operation, and if any eigenvalue exceeds the limited threshold, it is determined that the product under test on the rocket is faulty.

Preferably, after the signal processing module judges that there is a fault in the rocket product to be tested, it matches the characteristic value with the sound characteristic under the corresponding typical fault in the sound library, and if it is within the envelope of the sound characteristic of a certain typical fault, then judges the fault Type is this typical fault.

Preferably, if there is no corresponding typical fault, after fault analysis and location, the sound characteristics and envelope range of the fault are added to the sound library.

Preferably, the acoustic sensor arrays are distributed in the inner, middle and outer rings of the sensor bracket along the circumferential direction, and the radial angles of the sensors between adjacent rings are inconsistent.

Preferably, the acoustic sensor arrays are evenly distributed on the spiral arms of the sensor bracket.

Preferably, the signal processing module receives the sound signal, performs wavelet transformation and Fourier transformation, and then performs feature value extraction.

At the same time, an audio-based launch vehicle failure detection method is provided, including the following steps:

(1) The sound sensor array receives the sound signal of the rocket product to be tested, and performs signal acquisition and processing on the sound signal;

(2) after wavelet transform and Fourier transform are carried out to the sound signal after collecting and processing, feature value extraction is carried out;

(3) Compare the eigenvalues with the eigenvalues in the sound library during normal operation. If any eigenvalue exceeds the limited threshold, it is judged that the product to be tested on the rocket is faulty.

Preferably, after the product to be tested on the rocket has a fault, the characteristic value is matched with the sound characteristic under the corresponding typical fault in the sound library. If it is within the envelope of the sound characteristic of a certain typical fault, then the fault type is judged to be the Typical failure.

Preferably, the sound library includes the sound characteristics and envelope ranges of all products under test in the rocket when they are in normal operation, and the sound characteristics and envelope ranges of the products under test under typical failures.

Preferably, if there is no corresponding typical fault, after fault analysis and location, the sound characteristics and envelope range of the fault are added to the sound library.

Compared with the prior art, the present invention has the following advantages:

(1) The present invention carries out fault detection by sound signal, has realized the non-contact test to rocket, has reduced the test circuit inside rocket, has avoided the problem that test circuit fault brings on the arrow, has lightened the weight of rocket simultaneously.

(2) The present invention has good real-time performance of fault detection, and improves the efficiency of fault location. Since the sound signal is not limited by the number of measuring points on the arrow, the test coverage is improved.

(3) The fault detection method of the present invention reduces the test preparation work and can reduce the test complexity to a certain extent.

(4) The present invention establishes a relatively complete sound feature library, which can effectively ensure the comprehensiveness of product detection.

Description of drawings

Figure 1 is a flow chart of the testing system;

Fig. 2 is the distribution diagram of the sound sensor of the present invention;

Fig. 3 is the waveform diagram before and after signal denoising of the present invention;

Fig. 4 is that the present invention is short-time Fourier transform waveform;

Fig. 5 is the characteristic curve of sound of engine electromagnetic valve of the present invention;

Fig. 6 is the characteristic curve of the sound of the servomechanism of the present invention.

Detailed ways

The sound-based launch vehicle fault detection system of the present invention is used to collect sound signals from the rocket test process, process and identify the signals, and confirm the working status of the equipment on the rocket through the resolution and comparison of the sound database to improve the rocket test coverage. sex.

The launch vehicle equipment in the operating state produces a large number of sound signals. The detection system collects these acoustic signals through multiple acoustic transducers. After the sound signal is adjusted and converted and the signal is collected, it is transmitted to the host computer for processing; the host computer extracts the time-frequency domain characteristics of the system through the corresponding signal processing algorithm, and judges the working sequence of the system and whether there is a fault. , and then show the fault location through the relevant fault location algorithm. Its specific process is shown in Figure 1.

1. Selection and layout of sound sensor

The multi-channel acquisition of sound signals is to retain as much energy and information of the original signal as possible at a certain acquisition speed, that is, to select a sound sensor with good accuracy and low distortion and to a certain extent, eliminate the impact of external noise on signal acquisition. Impact. Good consistency between sensors is required.

In one embodiment, the sensor selected by the present invention is a 130F22 sound sensor produced by PCB Company of the United States.

After confirming the selection, the impact of the distribution position of the sound sensor on the sound signal acquisition should also be considered. The installation position of the acoustic sensor directly affects the original signal-to-noise ratio of its output signal, and is one of the important factors affecting system fault detection. Due to the uncertainty of the internal fault distribution of the detected device, the sound source distribution, propagation medium composition and propagation path length of the acoustic emission wave also have strong randomness. However, according to the propagation mechanism of acoustic emission waves, the shorter the propagation path, the smaller the energy attenuation of acoustic emission waves. It can be seen that an appropriate acoustic sensor array distribution position can effectively retain signal energy, thereby better retaining signal characteristics.

The distribution of sound sensors should cover all the sound signals as much as possible, so multiple sensors need to collect together. For the distribution of sensors, the distance between devices with a single sensor must first be considered, and the second is the distance distribution between multiple sensors. For the former, it is necessary to ensure that the sensor can cover as wide a sound signal as possible, and it is required that the sound signal be attenuated as little as possible.

The acoustic sensor of the present invention is arranged around the arrow body of the rocket, referring to Fig. 2, the acoustic sensors a1 to a6 are evenly arranged in the inner ring of the sensor bracket, the acoustic sensors b1 to b6 are evenly arranged in the middle ring of the sensor bracket, and the acoustic sensors c1 to c6 are evenly arranged Arranged on the outer ring of the sensor bracket, the sensors between the rings are distributed according to a certain geometric law, so that the sensor array can be sensitive to the directivity of the sound.

In one embodiment, the acoustic sensors a1 to a6 are consistent with the corresponding radial angles of the acoustic sensors c1 to c6 (inconsistent), and form an included angle of 30 degrees with the acoustic sensors b1 to b6 respectively.

In another embodiment, the sensor bracket includes helical arms uniformly distributed along the axial direction, each arm is placed with multiple sensors, and the sensors on different arms are distributed according to a circle with a certain radius.

The sensor bracket is placed within the sound receiving range of the product under test of the rocket, and receives the sound signal emitted by the product during operation.

2. Signal Conditioning and Acquisition

After the signal is collected by the sound sensor, it has a large background noise and environmental noise, and the output voltage amplitude of the sound sensor may not be well matched with the subsequent signal acquisition card. The purpose of signal preprocessing is to Before the signal is transmitted to the host computer, the signal is processed to a certain extent, so that it can further filter out the influence of noise on the basis of the previous link of acquisition, and provide support for the completion of the acquisition process of quickly converting the acquisition signal into a digital signal, and then It meets the requirements of the host computer for further calculation, analysis and extraction.

The lower computer receives sound signals sent by multiple sound sensors, adjusts the voltage range to the collection range of the acquisition board, completes high-speed acquisition of multi-channel sound data, and sends it to the upper computer.

3. Signal processing algorithm

The upper computer performs signal processing after receiving the sound signal. The target data collected from the acoustic sensor, after signal preprocessing, greatly reduces the noise in the signal and improves the signal-to-noise ratio. However, the amount of signal data obtained at this time is still very large, and it is not feasible to directly classify and identify these data. In order to effectively classify and identify, it is necessary to transform the data to obtain the target features that can reflect the essence of classification.

1) Wavelet transform

Because the working environment of rocket testing is complex, it is difficult for the detection system to ensure that no noise signals are mixed in when collecting sound signals. If these signals cannot remove these noises, the quality of signal feature extraction information will be affected to a large extent, thereby affecting to the fault judgment result. Therefore, it is very necessary to perform denoising processing on the collected sound signal.

For the collected signals, fast time domain and frequency domain analysis should be carried out to provide analysis basis for the next step of working status detection and fault location detection. Because of this, first of all, it is necessary to denoise the signal by means of wavelet transform.

The denoising methods can be classified into three categories: using wavelet filter characteristics to denoise, using wavelet transform modulus maxima to denoise, and threshold value processing to denoise.

2) Short-time Fourier analysis

The short-time Fourier transform is widely used in the analysis of non-stationary signals. On the basis of Fourier transform, the non-stationary signal is regarded as composed of a series of short-term stationary signals, and the short-term nature is realized by adding a window. , and cover the entire time domain by the translation parameter. That is, the product of the window function and the non-stationary signal to be analyzed is used to realize window opening and translation near the window, and then perform Fourier transform.

The use of short-time Fourier transform can make the signal show the time-domain characteristics and frequency-domain characteristics at the same time, which is conducive to further processing to obtain effective time-frequency comprehensive identification characteristics. After denoising the detection signal, the obtained signal can be divided into frames according to the timing command signal of the system, and short-time Fourier analysis can be performed on it, and different features can be extracted in the time domain and frequency domain. , including frequency, amplitude, phase, etc.

3) Fault location

Establish a sound library by collecting the sound characteristics and envelope ranges of all tested products during normal operation, and the sound characteristics and envelope ranges of products under typical faults.

Compare the measured eigenvalues with the corresponding eigenvalues in the sound library during normal operation. If any eigenvalue exceeds the limited threshold, it is judged that the product under test is faulty.

If it is judged that there is a fault in the product, the measured characteristic value is matched with the sound characteristic under the corresponding typical fault in the sound library. If the sound characteristic of a typical fault is within the range, it is judged that the fault type is the typical fault.

If there is no corresponding typical fault, it indicates that a new fault type has appeared. After fault analysis and location, the sound characteristics and envelope range of the fault are added to the sound library.

Detection methods of several typical products:

1. Engine solenoid valve

As shown in Figure 5, by processing the working sound of the solenoid valve, its frequency characteristic value is obtained, and the occurrence time of the frequency spectrum line (corresponding to the exact time of the solenoid valve action) can be obtained through the change time of the frequency spectrum line, which is consistent with the sound library Compare the frequency characteristic spectral lines to determine whether the occurrence time of the spectral lines corresponds. If it is not within the threshold range, it indicates that the solenoid valve is faulty.

Then compare it with the typical fault frequency characteristics to judge whether it is excess material stuck, insufficient power supply current or insufficient air supply pressure.

2. Servo mechanism

As shown in Figure 6, by processing the working sound of the servo machine, its frequency characteristic value is obtained, and compared with the frequency characteristic spectrum line in the sound library, it is judged whether the spectrum line is correct. If it is not correct, it indicates that the servo mechanism is faulty; at the same time, Judging whether the time of spectral line switching corresponds to the theoretical working time of the product, if it is inconsistent with the theoretical value, it indicates that the servo mechanism is faulty.

If there is a fault, it is further compared with the typical fault frequency characteristics to determine the fault location.

By collecting the sound emitted by the equipment during the launch vehicle test, the equipment health status can be diagnosed and analyzed in real time or afterwards, and it also has a certain fault location function. It can collect and record the sound signal during the test; it can locate the sound signal through different measuring point positions of the sound sensor array; it can establish a test sound database; it can identify the collected sound and compare it with the typical sound.

The waveform comparison before and after denoising is shown in Figure 3. The picture on the left is before denoising, and the picture on the right is after denoising. It can be seen that the amplitude of the corresponding noise signal is significantly reduced, while the characteristic signal becomes more obvious.

It is used for time-frequency domain feature extraction between related signals, such as the feature difference between signals in normal and faulty states of equipment, as shown in Figure 4. It can be seen from Figure 4 that the sound characteristics (amplitude, frequency, phase) of the fault state are quite different from those of the normal working state. By setting the corresponding threshold, it can be judged whether there is a fault.

The above description is only the best specific implementation mode of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of changes or modifications within the technical scope disclosed in the present invention. Replacement should be covered within the protection scope of the present invention.

The content that is not described in detail in the specification of the present invention belongs to the well-known technology of those skilled in the art.

Claims (12)

1. a kind of booster failure detection device based on audio, it is characterised in that including sound transducer array, signal is adopted Collect module and signal processing module；

The voice signal of product to be measured on the sound transducer array received rocket, signal acquisition is carried out through signal acquisition module Signal processing module is sent to after processing；Signal processing module extracts the characteristic value of voice signal, carries out fault distinguishing.

2. the booster failure detection device based on audio as claimed in claim 1, it is characterised in that the characteristic value bag Include frequency, amplitude or phase.

3. the booster failure detection device based on audio as claimed in claim 2, it is characterised in that signal processing module Carry out fault distinguishing method be：Characteristic value during by characteristic value with being worked normally in voice bank is contrasted, if any spy Value indicative then judges that there are failure for product to be measured on rocket beyond threshold value is limited.

4. the booster failure detection device based on audio as claimed in claim 3, it is characterised in that signal processing module Sentence rocket product to be measured to deposit after a failure, by the sound property progress under characteristic value typical fault corresponding with voice bank Match somebody with somebody, if in the range of the sound property envelope in a certain typical fault, failure judgement type is the typical fault.

5. the booster failure detection device based on audio as claimed in claim 3, it is characterised in that if do not corresponded to Typical fault, then after carrying out accident analysis positioning, the knocking noise characteristic and envelope scope are added into voice bank.

6. the booster failure detection device based on audio as claimed in claim 3, it is characterised in that sound transducer battle array Row be circumferentially distributed in sensor stand it is interior, in, outer shroud, the sensor radial angle between adjacent ring is inconsistent.

7. the booster failure detection device based on audio as claimed in claim 3, it is characterised in that sound transducer battle array Row are evenly distributed on the spiral support arm of sensor stand.

8. the booster failure detection device based on audio as claimed in claim 1 or 2, it is characterised in that signal processing Module receives voice signal, carries out wavelet transformation and Fourier transformation, then carry out characteristics extraction.

9. a kind of booster failure detection method based on audio, it is characterised in that include the following steps：

(1) voice signal of sound transducer array received rocket product to be measured, signal acquisition process is carried out to voice signal；

(2) after carrying out wavelet transformation, Fourier transformation to the voice signal after acquisition process, characteristics extraction is carried out；

(3) characteristic value during by characteristic value with being worked normally in voice bank is contrasted, if any feature value is beyond restriction threshold Value, then judge that there are failure for product to be measured on rocket.

10. the booster failure detection method based on audio as claimed in claim 9, it is characterised in that：Production to be measured on rocket Product are deposited after a failure, the sound property under characteristic value typical fault corresponding with voice bank are matched, if being in certain In the range of the sound property envelope of one typical fault, then failure judgement type is the typical fault.

11. the booster failure detection method based on audio as claimed in claim 10, it is characterised in that：The voice bank bag Include sound property when all products to be measured work normally in rocket and the sound under envelope scope, and product typical fault to be measured Characteristic and envelope scope.

12. the booster failure detection method based on audio as claimed in claim 10, it is characterised in that：If do not correspond to Typical fault, then after carrying out accident analysis positioning, the knocking noise characteristic and envelope scope are added into voice bank.