2010 International Conference on Computational Intelligence and Communication Networks 2010 International Conference on Computational Intelligence and Communication

Systems

A mobile based Application for Detecting fault by sound analysis in Car Engines using Triangular window & Wavelet transform
Yogesh Karunakar, Alhad Kuwadekar, Khalid Al Begain University of Mumbai askyogi@gmail.com University of Glamorgan akuwadek@glam.ac.uk University of Glamorgan kbegain@glam.ac.uk
Abstract: Sound energy belies an enormous amount of information which can provide assistance in various applications like navigation, communication, recognition, medical diagnosis, detection and therapy, analysis & design of structures and many more. This paper presents a method for detecting faults in car engines just by recording their sounds through window based applications in mobile phones and processing it in the mobile phone using our algorithm. The algorithm preprocesses the sound to remove the noise if present, the noise free car engine sound is passed through a triangular window to acquire the Multiresolution coefficients for discrete wavelet transform and compare resulting plots. The Wavelet transform analysis [1, 2] is done and the plots are scrupulously analyzed and the faults can now be detected by referring to plots. With a little know-how non technical persons can now detect the abnormalities in car engines just be using this automation. Keywords Car engine sounds, Multiresolution, Discrete Wavelet transform, Triangular Window.

1. Introduction Even the common man in todays world would eye to own a car for his luxury. De facto car now is no longer a luxury it has become a necessity and a commodity in most of the bourgeois family too. It is the availability of proper assistance for technical faults in the mechanisms that is of acute shortage. If only a non technical person can detect a fault in the engine just by inspection using a simple hand held mobile phones which will always be available. In this paper though, we have not made provision for diagnostics but the person who owns the car can at least detect the fault prior to the trip he makes so that he is not stranded somewhere there is no help available. With the rapid growth of signal processing technology, the vibration and sound emission signals can be used to monitor the condition of a mechanical system. An effective condition monitoring can prevent serious damage. There exist a number of fault diagnosis techniques in the field, and most of the conventional techniques are used to observe the amplitude difference in time or frequency domain for damage diagnosis. Meanwhile, order-tracking analysis is also used to avoid the smearing problem in frequencyvarying signals. Unfortunately, most of the conventional approaches are difficult to deal with in the case where the machinery operates under non-stationary rotational speed, is based on the assumption of stationary signals, and is
978-0-7695-4254-6/10 $26.00 2010 IEEE DOI 10.1109/CICN.2010.103 523

inherently unsuited for non-stationary. Time-frequency analysis can be used to improve the drawback of the Fourier transform. The primary advantage of time-frequency analysis is the representation of signals in both the time and frequency domains. The short time Fourier transform (STFT) has been applied to analyze the signals of the fault in both the time and frequency domains. However, STFT has a limitation of time resolution because of using fixed time windows. On the other hand, the continuous wavelet transform with an adjustable window size has been proven to have higher efficiency in accurate information of analysis signals. The continuous wavelet transform (CWT) with more precise time resolution can improve the performance of STFT. The faults in the vehicle engine may be due to various reasons. The faults may be termed as intermittent failure or permanent Both the failures can further be classified according to the suddenness with which failure occurs. A fault diagnosis system based on Discrete Wavelet Transform (DWT) technique for internal combustion engine using sound emission signals is proposed by Juan Da Wu and Chiu Hong Liu [6]. In this paper, the noise signals emanated from the Hyundai i20 engine are decomposed into five level using wavelet transform. Energy levels from the 5 level decomposed signals are obtained using filter banks and these data are used as a feature vector to classify the fault present in the Car engine. In this paper, the discrimination of energy levels in different sub-band frequencies can used as a training feature and to classify the faults in the Car Engine as a further work. 2. Continuous wavelet transform For analysis of signals FFT was the best known option but FFT had a draw back it cannot be used for signals which are time varying it worked fine for stationary signals only. For sound signals, which are time varying a modification of FFT must be done. This was called the STFT or the Short Time Fourier Transform, but this also came with some shortcomings. The continuous wavelet transform was developed as an alternative approach to the short time Fourier transforms to overcome the resolution problem. The wavelet analysis is done in a similar way to the STFT analysis, in the sense that the signal is multiplied with a function, similar to the window function in the STFT, and the transform is computed separately for different segments

of the time-domain signal. However, there are two main differences between the STFT and the CWT: 1. The Fourier transforms of the windowed signals are not taken, and therefore single peak will be seen corresponding to a sinusoid, i.e., negative frequencies are not computed. 2. The width of the window is changed as the transform is computed for every single spectral component, which is probably the most significant characteristic of the wavelet transform. The continuous wavelet transform is defined as follows :

Filtering a signal corresponds to the mathematical operation of convolution of the signal with the impulse response of the filter. The convolution operation in discrete time is defined as follows:

As seen in the above equation, the transformed signal is a function of two variables, and s, the translation and scale parameters, respectively. (t) is the transforming function, and it is called the mother wavelet. The term mother wavelet gets its name due to two important properties of the wavelet analysis as explained below: The term wavelet means a small wave. The smallness refers to the condition that this (window) function is of finite length (compactly supported). The wave refers to the condition that this function is oscillatory. The term mother implies that the functions with different region of support that are used in the transformation process are derived from one main function, or the mother wavelet. In other words, the mother wavelet is a prototype for generating the other window functions (with varying geometrical shapes) occurring in clusters commonly found in images with high cell density. The proposed system is computationally efficient and efficient as it makes use of Fast Fourier Transform. 3. Multiresolution Analysis The discretized continuous wavelet transform enables the computation of the continuous wavelet transform by computers, but it is not a true discrete transform. As a matter of fact, the wavelet series is simply a sampled version of the CWT, and the information it provides is highly redundant as far as the reconstruction of the signal is concerned. This redundancy, on the other hand, requires a significant amount of computation time and resources. The discrete wavelet transform (DWT), on the other hand, provides sufficient information both for analysis and synthesis of the original signal, with a significant reduction in the computation time. DWT coefficients are usually sampled from the CWT. The procedure starts with passing this signal through a half band digital low pass filter with impulse response h[n].

A half band low pass filter removes all frequencies that are above half of the highest frequency in the signal. After passing the signal through a half band low pass filter, half of the samples can be eliminated according to the Nyquists rule, since the signal now has a highest frequency of p/2 radians instead of p radians. Simply discarding every other sample will sub-sample the signal by two, and the signal will then have half the number of points. The scale of the signal is now doubled. Note that the lowpass filtering removes the high frequency information, but leaves the scale unchanged. Only the sub-sampling process changes the scale. Resolution, on the other hand, is related to the amount of information in the signal, and therefore, it is affected by the filtering operations. Half band lowpass filtering removes half of the frequencies, which can be interpreted as losing half of the information. Therefore, the resolution is halved after the filtering operation. The subsampling operation after filtering does not affect the resolution, since removing half of the spectral components from the signal makes half the number of samples redundant anyway. Half the samples can be discarded without any loss of information. In summary, the lowpass filtering halves the resolution, but leaves the scale unchanged. The signal is then sub-sampled by 2 since half of the number of samples is redundant. This doubles the scale. This procedure can mathematically be expressed as

The DWT analyzes the signal at different frequency bands with different resolutions by decomposing the signal into a coarse approximation and detail information. DWT employs two sets of functions, called scaling functions and wavelet functions, which are associated with low pass and highpass filters, respectively. The decomposition of the signal into different frequency bands is simply obtained by successive highpass and lowpass filtering of the time domain signal. The original signal x[n] is first passed through a halfband highpass filter g[n] and a lowpass filter h[n]. After the filtering, half of the samples can be eliminated according to the Nyquists rule, since the signal now has a highest frequency of p /2 radians instead of p. The signal can therefore be sub-sampled by 2, simply by discarding every

524

other sample. This constitutes one level of decomposition and can mathematically be expressed as follows:

of the exact frequency), but their widths are longer (which correspond to poor time resolution, since there is more ambiguity regarding the value of the exact time). At higher frequencies the width of the boxes decreases, i.e., the time resolution gets better, and the heights of the boxes increase, i.e., the frequency resolution gets poorer. 4. Implementation and methodology The vehicles with faults and no faults at the garage was examined by experts and faults like damaged car engine, flooded engine, timing belt problem, drive-belt slip problem, battery discharge problem, starter problem and fuel problems are obtained and recorded for each car. A windows based mobile phone was used to record the sound signal from the car engine. The ISO standards [9] were followed for the measurement of sound. The ISO3744 and ISO3745 specifications are part of the ISO 3740 series of specifications that define methods for determining sound power levels of various types of equipment from measurement of sound pressure. ISO3745 is a precision method and ISO3744 is an engineering grade method. Both require free field or a free field over a reflecting plane environment, which means that the test has to be conducted in a fully anechoic or a hemi-anechoic chamber. The measured sound pressure drop with distance from the source is then compared to the theoretical drop (6dB per doubling of distance) from a point source in a free field and the difference has to be less than certain values. In case this difference exceeds the maximum values, other standards have to be used, that is either ISO 9614-1 or ISO 9614-2 (sound intensity based). The position of recorder is kept 1meter above the ground level and 1.3 meter from the center of the car engine to measure the noise signal based on the Acoustical International Standard (ISO 9614 & ISO 3745).

where yhigh[k] and ylow[k] are the outputs of the highpass and lowpass filters, respectively, after sub-sampling by 2. This decomposition halves the time resolution since only half the number of samples now characterizes the entire signal. However, this operation doubles the frequency resolution, since the frequency band of the signal now spans only half the previous frequency band, effectively reducing the uncertainty in the frequency by half. The above procedure, which is also known as the sub-band coding, can be repeated for further decomposition. TIME AND FREQUENCY RESOLUTIONS The illustration in Figure 1. is commonly used to explain how time and frequency resolutions should be interpreted. Every box in Figure 1. corresponds to a value of the wavelet transform in the time-frequency plane. Note that boxes have a certain non-zero area, which implies that the value of a particular point in the time-frequency plane cannot be known. All the points in the time-frequency plane that falls into a box are represented by one value of the WT.

Figure 1: Block Diagram of the system Looking at Figure 1, first thing to notice is that although the widths and heights of the boxes change, the area is constant, each box represents an equal portion of the time-frequency plane, but giving different proportions to time and frequency. Note that at low frequencies, the height of the boxes are shorter (which corresponds to better frequency resolutions, since there is less ambiguity regarding the value

Figure 2: Block diagram for implementation of the system. The recorded signal is saved as .wav format and this sound. The above experiment is conducted on ten different cars of

525

same model. In this experiment, the sound emanated from the car engine is recorded at four different gear conditions speed is maintained constant at 1500 rpm and kept at stationary condition. The engine noise signal are digitized and decomposed into 5 levels of decomposition using triangular wavelet. Harr wavelets have good energy compaction and easy to implement but they produce jagged results in sound files. The result is plotted after logarithmic compression of frequency scale so that the vertical blocks yield equal size thereby reducing the time and frequency resolution problems mentioned in section 4 of this paper. The five level decompositions are obtained for various engine fault signals; damaged car engine, flooded engine, timing belt problem, drive-belt slip problem, battery discharge problem, flat battery problem, starter problems and fuel problems are shown in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 Figure 8 , Figure 9 and Figure 10 respectively. The abnormalities can be seen at lowest mid and the highest frequency slots in the plots shown.

(a)

(b)

Figure 5: (a) without (b) with timing belt problem

(a)

(b)

Figure 6: (a) without (b) with drive belt slip problem (a) (b)

Figure 3: (a) non damaged (b) Damaged car engine

(a)

(b)

Figure 7: (a) without (b)with discharged battery problem (a) (b)

Figure 4: (a) without (b) with flooded engine problem

526

(a)

(b)

Figure 8: (a) without (b) with Flat battery problem

5. Conclusion and Future Work The current work describes an algorithm which is detects faults in car engines. The current algorithm is developed using Visual C++ and based on the Windows Mobile platform. The current system takes about 1 minutes 35 seconds to detect if there is fault in car engine. However this time period is much longer than the time it takes to obtain results on a PC. The same algorithm on a PC takes about 45 seconds to provide the appropriate results. However in the future it is important to integrate the neural network within the existing mobile phone for diagnosis and classification of faults. This will help the user to detect and understand the type of fault and also get the diagnostic advice required from the software. The next step is to reduce the time it takes for the algorithm to execute on the mobile phone. An enhancement in the technology would make it provide quick results which can be sent via email, SMS or MMS to the garage and mechanics in the nearby location. 6. References [1] Jakub Krl, Petr Jezdik, Analza chovn zapalovachostroj v automobilu koda (in Czech), study independent thesis, November 2006. [2] Jedk, P.:Centralised diagnostic of electronic and electric equipment in vehicles, proceeding of IMAPS 2006, conference Brno 14.-15.9.2006, ISBN 80-214-3246-2 [3] Jedk, P.-Novk, J.: Centralised diagnostic of electronic and electric equipment in vehicles - engine lighting equipment testing, International Scientific Journal of Computing. 2008. [4] Petr Jedk , Jir Novk: Cars ignition system diagnostics using continuous wavelet transforms XIX IMEKO World Congress Fundamental and Applied Metrology September 611, 2009, Lisbon, Portugal. [5] Wang, W. J., & McFadden, P. D. (1996). Application of the wavelet transform to gearbox vibration signals for fault detection. Journal of Sound and Vibration, 192, 927939. [6] Wang, W. J., & McFadden, P. D. (1995). Application of orthogonal wavelets to early gear damage detection. Mechanical Systems and Signal Processing, 9, 497507. [7] Newland, D. E. (1995). Progress in the application of wavelet theory to vibration analysis. Proceedings of ASME 15th Biennial Conference on Mechanical Vibration and Noise, Boston, 3, 13131322. [8] Newland, D. E. (1999). Ridge and phase identification in the frequency of transient signals by harmonic wavelets.

(a)

(b)

Figure 9: (a) without (b) with Starter problem

(a)

(b)

Figure 10: (a) without (b) with fuel pump problem

527

American of Mechanical Engineers, Journal of Vibration and Acoustics, 121, 149155. [7] Lin, J. (2001). Feature extraction of machine sound using wavelet and its application in fault diagnosis. NDT & E International, 34(1), 2530. [8] Meltzera, G., & Dien, N. P. (2004). Fault diagnosis in gears operating under non-stationary rotational speed using polar wavelet amplitude maps. Mechanical Systems and Signal Processing, 18, 985992. [9] Glenn Pietila Gabriella Cerrato , Keith Davis: Practical comparisons of ISO 3744 and ISO 3745 Sound Power standards for automotive compressor testing , NOISE-CON 2008 Dearborn, Michigan, July 28-30, 2008.

528