You are on page 1of 7

Instantaneous Angular Velocity Irregularity Evaluation

Martin Novk Department of Instrumentation and Control Engineering Faculty of Mechanical Engineering Czech Technical University in Prague Technick 4, Prague Czech Republic Email:

This paper presents some signal processing algorithms and its implementation for an engine diagnostic system based on the instantaneous angular velocity irregularity evaluation method. This method is investigated at the Department of Instrumentation and Control Engineering, Faculty of Mechanical Engineering of the Czech Technical University in Prague. This method presents a completion of the existing engine diagnostic methods based on vibration, sound or exhausts gas analysis. The presented system is based on the TMS32C6713 DSP necessary for the real time signal processing chain implementation as the system is working with the sampling frequency of 200 kHz and a very high filters order on several places of the system. This system is designated to provide a noise free signal to a engine diagnostic system eliminating the influence of sensor-engine system vibrations

1. Introduction
The combustion engine is one of the most used engines today. It is used to drive different vehicles, machines or electrical backup generators. In all these applications, there is a high need for reliability and the use of sophisticated diagnostic systems as they can provide a significant cost reduction by reducing the necessary maintenance. A diagnostic system capable of revealing a fault before a critical system failure occurs helps to plan maintenance in advance. A large number of diagnostic methods have been developed in the past. Vibration or sound analysis, exhaust gas analysis or cylinder pressure measuring to name just a few of them. One of the other methods, based on instantaneous angular velocity evaluation and analysis is being investigated at the Department of Instrumentation and Control Engineering, Faculty of Mechanical Engineering of the Czech Technical University in Prague. This method is based on the fact that the angular velocity during the engines working cycle (one or more revolutions dependent on the engine type) is not constant. Its instantaneous angular velocity changes during the engines cycle. However the changes are relatively small, having typical amplitude of 1 percent of the engines mean angular velocity which makes the evaluation difficult. One problem is how to obtain a sufficiently clear signal for this diagnostic system by maximizing the information content and by eliminating the influence of noise caused for example by the engines vibration. This paper shows some of the methods and algorithms used for the signal processing of the inductive sensors signals to evaluate the instantaneous angular velocity irregularity and the critical implementation issues of this signal processing system.

2. Overall system description

This system consists of inductive sensors (either one or two), measuring the distance between the sensor and the gear rim of the engine. The placement of the sensor and the gear rim is shown on Fig. 1.

Inductive sensor

Figure 1: Inductive sensor placement and sensors output waveforms

As the gear rim is made of ferro-magnetic material, it changes the magnetic flux inside the coil core and this changes the coils induced voltage. The sensors output voltage is therefore dependent on the derivation of relative position between the gear rim and the sensor. The output voltage has a nearly sinusoidal form, when certain conditions for the dimensions between the sensor body and the inter-tooth gap of the gear rim are met. The frequency of the output sinusoid depends on the engine mean speed and the number of teeth on the gear rim as states Eq. (1) f = nz 60

[Hz, min ,-]



Where n is the engine mean speed in RPM and z is the number of teeth on the gear rim. From Eq. (1) can be seen that the frequency of the output signal does change with the engines revolutions. By measuring the changes in the frequency of this signal it is therefore possible to determine the engines instantaneous angular velocity. This instantaneous angular velocity can in return be used by the engines diagnostic system. The information for this system has to be cleared of disturbances like engine vibrations and at the same time maximal usable information content has to be preserved. There is also a one per revolution mark, allowing to measure RPM of the engine, used later to calculate the engine mean speed. The sensors output is processed via an attenuator and variable gain wide band amplifier as it is necessary to compensate for sensors variance when used with other sensors and to have a relatively constant output voltage amplitude for the A/D converter. The block diagram of the analog preprocessing circuits is shown on Fig. 2.

u t

u t

Inductive sensor

Variable gain amplifier

to ADC

Figure 2: Block diagram of analog preprocessing circuits

The 16-bit A/D converter ADS1606, with sampling rate up to 5MSPs, samples the input signal with a sampling rate of 200 kSPs, allowing the future sampling rate increase if it will be necessary. For the initial test of this presented method, also the 96 kSPs AIC23 audio codec on the TMS320C6713 DSP starter kit as well as National Instruments PCI-6024E data acquisition card were used. The reason why to use such a high sampling rate is the necessity to measure the frequency deviations caused by the engines angular velocity irregularity, as this have the order of magnitude about 1 percent of the engines mean speed and are therefore very small. There are also some other methods in consideration, like time-frequency analysis, that need higher sampling rates.

3. Signal processing chain description

The signal processing chain basically consists of several filters with frequencies adapting theirs settings in dependence of the engines mean angular velocity and a demodulator also adapting its setting the same way. The basic signal processing chain is based on the block diagram on Fig. 3. The signal processing algorithms are implemented on the TMS320C6713 DSP.
A/D Band Pass Demodulator Low Pass

Sensor Mean RPM


Diagnostic system

Figure 3: Signal processing chain block diagram

One of the critical implementation issues for this presented system was the need of very high order band and low-pass filters, the demodulator implementation, but most of all the need to calculate the filter coefficients on the fly dependent on the mean RPM value. The used filters are of order 512, sampling frequency of 200 kHz and the signal had to be processed in real time. The input band-pass filter is used to filter out the frequency deviations and disturbances caused by engines vibrations, by the gear rim magnetization etc. This is also causing amplitude modulation of the input signal, but as the amplitude is affected by the gear rim magnetization it cannot be used. This filter is a FIR type filter as these can have linear phase and has the filter order of 512. The filter coefficients are recalculated after each revolution so as to adjust the filters frequency to the engines mean speed. The filter coefficients are calculated using linear-phase least-squares error minimization FIR filter design algorithm. The same algorithm is used for the output FIR filter, in this case a low pass filter (LPF), also with linear phase.

4. Demodulator
It was initially considered and also tested to use a digital PLL to lock the local oscillator frequency to the input signal. There is however a simpler algorithm, IQ demodulation, used mainly in communication devices. This approach does not require a feedback loop from the output of the mixer to the voltage-controlled oscillator like the PLL does, but on the other hand it does require two output filters (one for the I and the second for the Q component) and it does require the computation of arctan function, but this disadvantage can be eliminated by the mathematical manipulation of the equations. The demodulator block diagram is shown on Fig. 4.

freq. adj.
calculate instantaneous frequency

cos input signal

-90 sin X Q


Figure 4: IQ demodulator block diagram

The mixer multiplies the input signal with the signal produced by the local oscillator, the I component with cosine component of the local oscillator, the Q with sine component, which can either be produced with a second look-up table, like the cosine component is or by using Hilbert transform to shift the cosine component by -90. The frequency of the local oscillator is set to the frequency based on Eq. (1) and is adjusted after each revolution, as are the lowpass IQ filters.

Consider the input signal in a sinusoidal form, based on Eq. (2) xFM (t ) = sin (w FM t + j (t ) ) Mixing the input signal with local oscillator signal yields xI (t ) = cos(wO t ) sin (w FM t + j (t ) ) xQ (t ) = sin (wO t )sin (w FM t + j (t ) ) Using trigonometric identities yields xI (t ) = 1 / 2 sin (j (t ) ) + 1 / 2 sin (2wO t + j (t ) ) xQ (t ) = 1 / 2 cos(j (t ) ) - 1 / 2 cos(2wO t + j (t ) ) (5) (6) (3) (4) (2)

From Eq. (5) and Eq. (6) it can be seen that the mixer output signal is a sum of the two components, a low frequency signal and a signal with twice the local oscillator frequency. By filtering out the high frequency component with a low pass filter, we obtain only the low frequency component we are looking for. The phase of the output signal is given by Eq. (7) x I (t ) / xQ (t ) = 1 / 2 sin (j (t ) ) / 1 / 2 cos(j (t ) ) = tan (j (t ) )

j (t ) = arctan( x I (t ) / xQ (t ))
f (t ) = dj (t ) dt


The instantaneous signal frequency can either be calculated with the arctan function or by mathematical manipulation we obtain Eq. (8) f (t ) = dj (t ) I dQ / dt - Q dI / dt = dt I 2 + Q2 (8)

This final equation is used to calculate the instantaneous angular velocity irregularity.

5. Signal processing results

The presented results were obtained with a two cylinder combustion engine. The instantaneous angular velocity irregularity for this case is shown on Fig. 5. This signal is the output to the engine diagnostic system. Now consider this signal as a reference signal, describing the instantaneous angular velocity irregularity for an engine with no known faults. By subtracting in real time the corresponding sections of the incoming signal and this reference signal, deviations from the no-fault state can be found. An example is shown on Fig. 6. where the result of this subtraction between the reference signal and a signal with fuel supply reduction is shown. This reference signal is valid only for the RPM it was measured and has to be recalculated for other RPM or another approach based on statistical theory, knowledge based system or a neural network has to be used.

0.8 0.6 deviation from mean speed (percent) 0.4 0.2 0 -0.2 -0.4 -0.6 -0.8



0.15 time (s)




Figure 5: Two cylinder engine instantaneous angular velocity irregularity

deviation from mean (percent) signal with engine fault 1 0 -1 0 0.05 0.1 0.15 time (s) error signal 0.2 0.25 0.3

cylinder fault detected here

error (percent)

0.5 0 -0.5 0 0.05 0.1 0.15 time (s) 0.2 0.25 0.3

Figure 6: Two cylinder engine instantaneous angular velocity irregularity

6. Performance requirements
This system is based on the TMS 320C6713 DSP. During the development of this system also a 54xx DSP was considered to lower system costs and power consumption but due to the high sampling rate of the input signal and high filters order the 6713 DSP was chosen. There is also the fact that this is a system under development and it has to be possible to use higher sampling rates for joint time-frequency analysis considered in the future as well other methods for signal processing. At the time being, the CPU load is low; it takes approximately 5000 cycles to calculate the result for one incoming sample.

7. Conclusions
The presented system has been realized on the TMS 320C6713 DSP fulfilling the high speed requirements of the proposed engine diagnostic method providing a signal with sufficient resolution and by eliminating the noise signal caused by the sensor-engine system vibrations. The signal processing algorithms were realized partially in C and in assembly. It was initially considered to use a fixed-point processor but there were large errors caused by rounding that masked the properties of the signal and increasing the overall error of signal processing. The presented results were therefore obtained by using floating-point double precision arithmetic on the TMS320C6713 DSP.

8. References
[1] Novk, L. No dismantling engine diagnostic by evaluation the instantaneous angular velocity PhD. thesis, Czech Technical University in Prague, Faculty of Mechanical Engineering, 2003 [2] Brown, T.; Neill, W. : Determination of Engine Cylinder Pressures from Crankshaft Speed Fluctuations, International Off-Highway and Powerplant Congress and Exposition 1992, Milwaukee, WI, USA, ISSN 0099-5908 [3] Rizzoni, G.: Engine fault diagnosis by signature analysis of crankshaft speed fluctuations, 1988, 18th International Symposium on Automotive Technology and Automation, Florence, Italy [4] Rizzoni, G.;Ribbens, W.B.: Onboard detection of internal combustion engine misfires, 1990, IEEE Workshop on Electronic Applications in Transportation, Dearborn, MI, USA [5] Rizzoni, G.: A passenger vehicle onboard computer system for engine fault diagnosis, performance measurement and control, 1987, 37th IEEE Vehicular Technology Conference, Tampa, FL, USA [6] Rizzoni, G.;Ribbens, W.B.: Crankshaft Position Measurement for Engine Testing, Control and Diagnosis, 1989, 39th IEEE Vehicular Technology Conference, San Francisco, CA, USA [7] Azzoni, P.; Marseguerra, M. : Assessment of the potential of a Wiener-Hilbert filter for automatic diagnosis of spark ignition engine faults, Mechanical Systems and Signal Processing, March 1995, vol. 9(2), page 119 - 128 [8] Hsi-Yung Feng : A rugged and reliable system for measuring engine crankshaft speed fluctuations, Journal of Dynamic Systems, Measurement and Control, December 1998, volume 120, p. 533, ISSN 0022-0434 [9] Mauer, G.: On-Line Cylinder Fault Diagnostics for Internal Combustion Engines, IEEE Transactions on Industrial Electronics, 1990, vol.37, no.3, ISSN 0278-0046