GB2496121A

GB2496121A - Fault location in a vehicle electrical system by time domain reflectometry

Info

Publication number: GB2496121A
Application number: GB1118727.5A
Authority: GB
Inventors: Allan Tengg
Original assignee: Kompetenzzentrum das Virtuelle Fahrzeug Forchungs GmbH
Current assignee: Kompetenzzentrum das Virtuelle Fahrzeug Forchungs GmbH
Priority date: 2011-10-31
Filing date: 2011-10-31
Publication date: 2013-05-08
Anticipated expiration: 2031-10-31
Also published as: GB2496121B; GB201118727D0

Abstract

A time-domain reflectometry method is used for detecting and locating faults in an automotive cabling network. A sinc stimulus pulse is applied to the network as a calibration (step 1) and the measured return echo is stored as a reference (step 3) for future testing. Periodically, the same stimulus pulse is applied and the difference between the received echo and the reference is calculated (step 6). Peaks within the received echo are detected (step 8) and classified (step 10) to find the peak indicative of a fault (step 11). The peak selected as the fault may be on the basis of the highest scoring peak in a correlation between the peaks and the applied stimulus. The method can analyze bus and passive-star topologies. Broken cables, short circuited cables and bus transceivers, modifications to the network and degenerating plug-in connections can be detected and located. Furthermore the acquired data may be analyzed remotely.

Description

Method for Fault Detection in Automotive Networks

Background of the Invention

In the last two decades automobiles have been electrified increasingly. Today, even small sized cars are equipped with over twenty electronic control units (ECUs), distributed all over the car. Of course, all these EcUs must be interconnected and powered. Thus, several hundred meters of cables need to be placed inside a car. These copper wires do not only cost money, they increase the weight of the car and thus increase its energy consumption. Therefore a lot of effort is put in the development of new and improvement of existing bus systems to reduce the wiring costs.

State of the Art Using time-domain reflectometry for detecting and locating faults in cables belongs to the state of the art. Depending on the circumstances, two different approaches are possible for detecting broken or short-cut cables. When dealing with periodic signals the standing wave ratio can be measured or interference methods (e.g. 0. Lopez-Lapena, J. Lopez-Villegas, J. Morante, and I. Samitier. Modified interferential method for locating coaxial cable faults. In IEEE Conference Proceedings on Instrumentation and Measurement Technology conference, Volume 2, Pages 822 -825 Vol.2, Brussels, Belgium, 1996) can be applied. More meaningful and often easier to implement are methods where cables under test are stimulated with short pulses and the response is observed. A general description of this method with a prototype implementation can be found in (S c. chen and L. Roemer. The application of cepstrum technique in power cable fault detection. In Acoustics, Speech, and Signal Processing, IEEE International Conference on ICASSP 76, pages 764 -767, April 1976); a refined approach is presented in (S. Xudong, L. Dayong, W. Liwen, and C. Jianshu.

Application of wavelet in airplane cable fault location. In Control and Automation, 2007. ICCA 2007.

IEEE International conference on, Pages 127 -130, June 2007). Already back in 1976, chen and Roemer published a paper about detecting faults in power lines L41. The cable faults there are usually characterized by degradation of the dielectric, either by water inclusion or physical cracking and associated voids. Voltage reflected at the fault is modelled as distorted echo. In their technique, a voltage step test signal propagates down the line and reflects energy when a change in impedance occurs. Digitized samples are taken along the return waveform. By calculating the power cepstrum of the response, information about the fault location can be concluded. A similar goal but a different approach can be found in (E. Song, Y.-J. Shin, P. Stone, J.Wang, T.-S. choe, J.-G. Vook, and J. B. Park.

Detection and location of multiple wiring faults via time frequency-domain reflectometry.

Electromagnetic Compatibility, IEEE Transactions on, 51(1):131 -138, Eeb. 2009). They monitor power lines during operation and by detecting changes in the phase, they are able to identify faults in the line and locate them without interrupting the normal operation. Another domain for cable checking can be found in airplanes. In (S. Potivejkul, P. Kerdonfag, S. Jamnian, and V. Kinnares. Design of a low voltage cable fault detector. In Power Engineering Society Winter Meeting, 2000. IEEE, Volume 1, Pages 724 -729 Vol.1, 2000) a method is presented which uses wavelet based analysis methods for finding damaged safety-critical cables in planes. A similar domain where transmission line theory has been successfully applied to is nautics. In (i. Wang, P. Crapse, V-i. Shin, and R. Dougal.

Diagnostics and prognostics of electric cables in ship power systems via joint time-frequency domain reflectometry. In Instrumentation and Measurement Technology Conference Proceedings, 2008.

IMTC 2008. IEEE, Pages 917 -921, May 2008) the joint time-frequency domain reflectometry (JTFDR) is proposed to monitor the integrity of the wiring in the electric power system of a ship. Experimental results presented for coaxial cables, which are widely used for military applications in ship power systems, demonstrate the feasibility. JTFDR not only has the ability to detect and locate incipient defects with high accuracy but is also permits to monitor the aging process of cables to predict both future defects and the remaining service life of the cables.

To overcome the problem of inaccurate localisation of faults, a method for detecting faults in the physical layer of an automotive network is described. It has the ability to detect four classes of failures: Broken cables: In the hostile environment of a car, cables and plug-in connections can be damaged in several ways (e.g. physical force) vibration, corrosion, ...). Apart from the obvious loss of several nodes, the signal reflections at the breaking point may render even the remaining network inoperative.

Short-circuits in the network: Even worse than broken cables are short-circuits on the network; caused either by damaged transceiver ICs or abrasion of cable isolation.

Increased contact resistance: Automotive connectors are known to be very error-prone components. Moisture, salt and oxygen cause connectors to degenerate over time. This leads to increased contact resistance. Subsequently reflections occur on the bus and may cause sporadic transmission errors. Since this degeneration progresses slowly, it is possible to detect it with the methodology presented in the following even before network failures can be noticed.

Modifications of the network: When dealing with safety critical systems, supplementary extensions of network may lead to unpredictable results. Especially CAN networks can be extended very easily by everyone. However such unprofessional modifications can cause a bus overload and violations of real-time constraints -a potentially dangerous situation which can be detected as well.

Description of the Invention

For this methodology it is necessary to characterize the cables that are used in the network. For better understanding, figures are provided.

Fig. 1 shows a flow chart of the invention for fault localisation Fig. 2 shows an example of the stimulus, two echoes and their squared difference Fig. 3 shows the visualisation of the parameters w1, w2 and h The most important properties to be determined are the propagation speed (vp) and the damping factor (24. For this purpose a piece of cable with precisely known length must be connected to the tester device and the calibration must be triggered manually.

Then a sinc shaped pulse (e.g. with a duration of 3 ns) is applied to the network to be monitored and its impulse response is measured 2. Assuming that the network is initially fault-free the measured impulse response of 2 is stored permanently in the tester device as reference r 3. Now the tester device has gathered all information it needs to monitor the network over time.

Depending on the application and the safety-criticality of the monitored network, a periodic fully automatic inspection of the network is done by the tester device 4. Therefore the tester device again sends the same sinc shaped pulse during a network idle time and records the echo e 5. Due to damping effects caused by the cable for the frequency spectrum of the stimulus signal, the amplitudes of echoes decreases exponentially with the distance from the tester device. When using high-speed low-resolution analogue-to-digital converters (ADCs) a technique called cornpanding is used very often to maintain a decent signal-to-noise ratio (SNR). This means that the conversion from analogue to digital is not linear. Smaller signal amplitudes have a higher resolution than large amplitudes. When using a simple subtraction (e -r) to calculate the difference between reference-and measured echo 6, small differences far away from the tester node are lost in the noise caused by the presence of strong peaks. To cope with this damping-and companding-issues the difference echo d at the time t is calculated according to following equation:

-__________

n -A(t t.) 2. represents the damping of the used cable. Hence the effect of damping is compensated and the compansion-effect reduced. Fig. 2 depicts a situation where a network node has a short-circuited input stage. In the plot on the top the reference echo r and the measured echo e are depicted; the plot below visualizes the difference echo d calculated by the equation above.

Depending on the difference echo d it must be decided if there exists a problem in the network or not 7. Three properties of the difference are relevant in our method: the peak value, the effective value of the pulse-response and an amplitude-histogram. A threshold value must be defined for the peak value and the effective value. If there is no fault present on the network, the envelope of the amplitude histogram is gauss-like-shaped, caused by the noise of the ADC, otherwise it indicates a problem in the physical layer.

If step 7 indicates a problem in the network, it can be located by finding the first relevant' peak in the difference signal 8. To filter additional peaks, a correlation of the stimulus pulse and the difference echo d is performed, resulting in df. All remaining peaks in the filtered difference echo d1 are assigned a score according to the following equation: P = Pan? Pnuln p. p Pamp denotes to the amplitude of the peak. Higher peaks are more preferred. Pnum is set to max for the first peak and reduced by a constant factor for every following peak. This favours the first peaks.

p and Pw2 account for the peak width -peak and stimulus should have similar width corresponding to a higher score. PwIw2 takes the pulse shape (ratio of WI and w2) into account. The energy of the pulse (its area A) in relation to its amplitude is incorporated by the last term PAh. A detailed description of w1, w2, and h can be seen in Fehler! Verweisquelle konnte nicht gefunden werden..

This step is summarized in the block feature extraction 9 and peak classification 10 respectively.

The peak selection 11 is done by taking the peak with the highest score from step 10. Together with the propagation speed of the cable (vp) the time delay (tn) of the selected peak 11 can be calculated into a distance 1 = t, Iv,. Based on a 3D wiring plan of the network, the precise location of the fault can be located 12. In case of networks with passive stars, the localization can be ambiguous. In this case the branch of the network with highest failure probability must be taken. Therefore each component in the netlist has assigned a failure probability value. In general the failure probability of plugs is assumed to be high, the failure probability of transceiver ICs is assumed to be medium and the probability of simple cables is assumed to be low, depending on its length and location. In the last step 13 the error is reported to the user of this system by using the data of a 3D wiring plan. An unique feature of this method are the links between 11-12 and 12-13: It is not required for the tester device to know the exact topology nor the 3D wiring plan of the network. The localization and visualization can be done at a remote processing unit (i.e. at the manufacturer of the network).