Real-time Diagnosis of Wire Degradation based on Digital Signal Analysis

Transcription

1 2017 IEEE 67th Electronic Components and Technology Conference Real-time Diagnosis of Wire Degradation based on Digital Signal Analysis Jinwoo Lee, Daeil Kwon System Design and Control Engineering UNIST (Ulsan National Institute of Science & Technology) Ulsan, 44919, Republic of Korea Abstract The wiring system of electronics is often exposed to operational and environmental stress conditions during the lifetime of use condition. As the wiring system ages, these stress conditions degrade the material properties of wires and eventually lead to wire failures that result in arcing and electromagnetic emissions. In order to prevent wire failures, various approaches to diagnosing wire degradation such as X- ray inspection, resistance analysis and reflectometry methods have been attempted mainly for routine maintenance. However, these approaches require external monitoring devices physically connected to the wiring system, which may interfere with the operation of electronics, thus limiting the possibility to diagnose the health of the wiring system in real-time. This study proposes a new approach to real-time diagnosis of wire degradation through continuous monitoring of digital signals used for data communication in electronics. Accelerated wire abrasion tests were conducted to demonstrate that the proposed method is capable of diagnosing the health of damaged wires. In the test, digital signals are transmitted through a wire subjected to abrasion. The characteristics of transmitted digital signals, such as eye diagram and eye parameters, are observed to gradually deteriorate as the wire is physically damaged. The test results indicate that changes in digital signal characteristics are closely related with the extent of wire damage. Since signal characterization can be performed within electronics where digital signals are continuously generated, the proposed method is effective in diagnosing wire degradation in real-time without external monitoring devices. Keywords-wire fault; real time diagnosis; digital signal; chafing; signal integrity; I. INTRODUCTION In commercial and industrial systems such as electronics and aircrafts, the wiring system is used for making electrical connection between electronic components. The wiring system ages through exposure to operational and environmental stress conditions such as mechanical and thermal stress conditions during the lifetime. Wires deteriorated by aging can be damaged, which may lead to wire failures resulting in arcing or electromagnetic emissions. Especially in the case of the wiring system in electronics such as an aircraft, damage of the wires is usually progressed from insulation of wire by chafing, which is typically caused by wires rubbing each other. According to the reports from U.S. Navy [1], Federal Aviation Administration (FAA) [2], and National Aeronautics and Space Administration (NASA) [3], chafing was identified as the most frequently detected faults in the wiring system of aircrafts. Furthermore, progress of chafing can cause wire failures, inducing malfunction of the wiring system such as damaged conductor. Thus, diagnostic approaches for wires are required to maintain the integrity of the wiring system. Among diagnostic approaches, time domain reflectometry (TDR) is one of the common approaches for wire. When a wire is diagnosed by TDR, a specific signal at high frequencies is transmitted through the wire and reflected at impedance discontinuities. Amplitude of reflected signal is affected by impedance variation during transmission. Distance of the impedance discontinuities can be calculated based on time delay between the injected signal and the reflected signal. Therefore, TDR can provide information about wire faults inducing impedance discontinuities and the fault location using external TDR hardware such as an oscilloscope. However, physical connection with the external hardware is a flaw that limits real-time diagnosis in electronics. In place of the conventional approaches that use external hardware, this study developed a new approach for real-time wire diagnosis using digital signal in electronics. Since the integrity of digital signal is deteriorated at impedance discontinuities on a transmission line like TDR, the proposed approach detects wire faults by monitoring characteristics of transmitted digital signal. For demonstration, an accelerated wire abrasion test was designed to induce wire faults gradually. During the test, the characteristics of the transmitted signal and TDR were monitored alternately to evaluate the proposed approach on the basis of TDR. II. RELATED WORKS Various approaches have been attempted for wire diagnosis in the wiring system. According to a study by NASA [3], the approaches can be categorized to two types: 1. injecting an electrical signal (either directly or inductively) and then measuring reflections and/or transmissions; 2. measuring externally to wires. Time domain reflectometry (TDR) belongs to the first type. In order to diagnose a wire, a diagnostic approach using TDR injects a pulse with fast rise time such as a square wave pulse or step function into the wire. When a wire fault causes an impedance discontinuity on the wire, the reflected signal partially returns back to the injected point. The fault location can be determined by time delay of the reflection based on the transmission speed and the wire length. Since TDR can detect wire faults with the fault locations, TDR has been improved as a technique for wire diagnosis. Smith et al. [4] improved the TDR technique, using spread spectrum time domain reflectometry (SSTDR). When the /17 $ IEEE DOI /ECTC

2 TDR diagnoses wires in operation, the incident pulse can interfere with other signals for operating on the wires. SSTDR utilized a sinusoidal wave modulated by pseudo noise (PN) code as an injected signal to prevent the interference. The injected signal is detectable through cross correlation due to PN code, even though the signal is buried in other signals on the wire. Some studies have improved the diagnostic approach to detect wire faults in complex wiring system. Hassen et al. [5] combined the diagnostic approach using TDR with distributed sensors in the wiring system. An orthogonal multitone time domain reflectometry (OMTDR) method was utilized to prevent signal interference like SSTDR. An arbitrary wave generator injected signals into the connected coaxial cables. The distributed sensors in the junctions of the wiring system measured the transmitted signal simultaneously. As a result, the diagnostic approach using sensor fusion can accurately detect a wire chafing based on aggregated measurement by the distributed sensors. The approach can cover a complex wiring system with only one measurement. However, the TDR approach requires the signal generator and external monitoring devices which may interfere with the operation of electronics, thus limiting the possibility of diagnosing the health of the wiring system in real-time. In the case of components in electronics, a few different approaches to real time diagnosis have been developed. Estima et al [6] developed a new algorithm for real time detection of multiple power switch open circuit faults in voltage-fed AC motor drives. Diagnostic variables were formulated based on the reference current errors calculated from input currents for motor phase currents and the measured corresponding currents. The currents can be measured in the main control system to control the AC motor. Based on the formulated diagnostic variables, the study detected multiple open circuit faults and localized the fault locations in a test AC motor drive for experiments. Kamel et al. [7] studied fault detection in power electronic converters (PEC) in the grid-tied wind system. The PEC consisted of three main circuitries: three-phase rectifier, boost chopper, and single-phase inverter. Since the circuitries were jointly related, the study classified faults by the circuitries. The faults in each circuitry were detected when the maximum output power did not achieve the desired output power calculated according to a measured input frequency of supply voltage. In order to calculate the output power of each circuitry, this study used the output voltage and the current measured by DC components in each circuitry. Therefore, the study defined normal and fault conditions based on the voltage and current measured in the PEC. Even though the diagnostic approaches in the studies [6][7] detected only limited faults, they can detect and localize the faults in real time without external monitoring devices. According to studies on real time diagnosis in electronic components, faults in the electronic components can be detected in real time when a diagnostic approach utilizes internal signals. However, the approaches such as TDR have limitations in real-time diagnosis due to additional hardware. In the wiring system, a digital signal is transmitted for data communication in electronics. Thus, this study used the digital signal to diagnose wires without employing external monitoring devices. III. METHODS A. Digital Signal Degradation During Transmission The proposed diagnostic approach in this study uses the properties of digital signals during transmission in the wiring system. When a high speed digital signal is propagated through conductor such as inner conductor of a wire, the high speed signal induced by the skin effect is concentrated on exterior site of the conductor. Digital signal concentration is induced by electrical density variation between exterior and interior site due to eddy current. The electrical density variation can be quantified by the skin depth(), which means the depth of a conductor layer where 63% of currents are concentrated. The skin depth is represented in the following equation with the frequency of signal () where and denote the conductor resistivity and the material s permeability, respectively. Thus, the skin depth becomes shallow when the frequency of signal is increased. Therefore, the denser digital signal is concentrated in the exterior site of the conductor at a high frequency. When digital signal is transmitted for data transmission in electronics, the characteristic impedance of transmission line should be controlled for signal quality. However, external damage on insulation such as chafing can deteriorate the characteristic impedance of the transmission line because the insulation material affects the characteristic impedance of the wire. The digital signal is deteriorated when the signal passes impedance discontinuities. Especially, the digital signal is affected by external damage since the denser signal is transmitted on the exterior of the conductor rather than the center due to the skin effect. Based on the skin effect of the digital signal, the proposed approach can detect wire faults by analyzing the digital signal which is transmitted through wires. The approach based on monitoring the transmitted digital signal was studied to diagnose interconnects in electronics. In the case of solder joints, which is one of the interconnects, we demonstrated in our previous study that the approach using digital signal can detect physical degradation of the solder joints under mechanical stress conditions [8], thermomechanical stress conditions [9], and chemical stress conditions [10][11]. Our previous studies verified that the digital signal is affected adversely by impedance discontinuity due to deteriorated solder joints on impedance controlled test boards. 1937

3 Figure 1. Schematic of Eye diagram and Eye parameters B. Eye diagram and Eye parameter Because of the skin effect, digital signal is deteriorated when it passes the impedance discontinuities of the externally damaged wire. Digital signal can be converted by measurable characteristics to detect its degradation. In this study, an eye diagram was monitored to represent the characteristics of continuously transmitted digital signal in one diagram. Digital signal is continuously generated for data communication in electronics. Digital signal is a waveform with continuous transition between two states, logical 1 and 0, to contain information. The waveform of the digital signal can be sampled at a regular time unit interval as regular bits of the digital signal. An eye diagram is created when the bits of signal are superimposed based on transition edges. Consistency of the eye diagram shape indicates the state of the wire because the eye diagram shrinks when the transmitted digital signal is deteriorated. Eye parameters in Fig. 1 are quantitative indices of the eye diagram. Eye parameters are determined by the distribution of superimposed signals. For example, the mean voltage of signals for logical 1 or logical 0 is called 1 level or 0 level. Rise time and fall time means the transition time of 0 level to 1 level and 1 level to 0 level, respectively. When the crossing points are determined by each mean value of two horizontal histograms across the narrow strip, jitter is determined based on the deviation of each histogram. Eye width is determined based on the two horizontal histograms on crossing points, and eye height is determined based on the two vertical histograms on 1 level and 0 level. In this study, all of the eye parameters are monitored during the test to represent the deterioration of the digital signal. C. Sequential Probabilistic Ratio Test In order to detect faults in the monitored signal without a standard fault detection threshold, the sequential probabilistic ratio (SPRT) test was performed to determine fault detection thresholds. SPRT can provide the statistical thresholds by the binary hypothesis test [12][13]. When a null hypothesis and an alternative hypothesis of the binary test, respectively, indicate a health state and an anomaly state with manually fixed ratio of false and missed alarm, SPRT can help to determine whether or not the new measurement data fall in the null hypothesis. In order to apply the digital signal, a probability distribution of the null hypothesis can be assured as a Gaussian distribution with mean(0) and variance( ) based on the measured data from the healthy wire. The alternative hypothesis for the mean test assures that the probability distribution of anomaly state is a Gaussian distribution with a mean of M and a variance of, where M is a defect level. A probability distribution of the alternative hypothesis for variance test is a Gaussian distribution with mean of 0 and variance of or, where V is a sensitivity constant determined manually. Since the mean values of both monitored eye parameters and TDR were shifted during the test, the mean test was used to determine whether the monitored signals are shifted over M or not. The state of measured data was indicated by an SPRT index calculated by the following equation where represents the sequentially measured data for the test. An upper limit and a lower limit were determined by the fixed ratio of false and missed alarm to indicate the state. When the SPRT index was under the lower limit, SPRT judged that the measured wire is in the health state. On the contrary, when the SPRT index was over the upper limit, SPRT indicated that the wire is in the anomaly state. In this study, signals from undamaged wires before experiments were used in training the health state. Based on the training data, SPRT indicated continuously whether the measured data were statistically different from the probability distribution of the training data. Thus, SPRT can identify significant changes in the signals monitored from the eye parameters. IV. EXPERIMENT In order to demonstrate that the approach using digital signal can diagnose wire chafing, the digital signal was monitored during an accelerated life test (ALT) under wire abrasion. Diagnosis capability of the digital signal was identified by fault detection time depending on the extent of chafing. Since the diagnostic approach using TDR has also detected wire faults based on impedance discontinuities, the diagnosis results by the digital signal and the TDR were compared to identify the difference in diagnosis capabilities. The TDR signal and the digital signal were monitored alternately in one sampling interval. The fault detection by the digital signal was compared with the fault detection by the TDR signal. Furthermore, in order to verify that digital signal has a degradation trend analogous to TDR signal, the correlation between the two signals was analyzed based on the Pearson correlation coefficient (r) calculated by the following equation (3) where means the covariance, means the standard deviation. The device under control (DUT) for ALT was determined as a coaxial cable at a controlled characteristic impedance of 50 Ohms named RG 174. The cable is insulated by Polyvinyl Chloride (PVC), which is one of the common materials for (2) 1938

4 insulation. Mechanical specification of RG 174 cable is shown in Fig. 2. An RG 174 cable was connected to the test circuit with SMA connectors. In order to damage the wire gradually by chafing, a wire abrasion test was designed on the basis of scrape abrasion in ISO Taber 5750 Linear abraser with a wire abrasion kit in Fig. 3 was used to chafe the specimen gradually. The wire was installed on the wire abrasion kit. An abrasion needle on the test zig in Fig. 4 travelled horizontally to chafe the wire with a normal force from the insulation to the conductor. Travel length and speed of the test zig were determined to be 15.49mm and 2 cycles/min respectively. A 150g weight and 0.25mm diameter needles were used. During the gradual chafing, an Altera Stratix V GX transceiver generated PRBS7 (Pseudo-Random Binary Sequence 7) patterned digital signal with a speed of 1.25Gbps. The conditions of the digital signal were determined to observe significant degradation of the signal integrity. A digital communication analyzer (DCA) Kesight 86100D monitored the transmitted signals both the digital signal characteristics and the TDR signal using two separated channels. First, a channel on the DCA was used to generate a step function signal with 0.2V-35ps rise time and analyze the reflected signal. Next, another channel on the DCA was used to monitor the transmitted digital signal and analyze the signal integrity. The digital signal was accumulated for 30 seconds to create an eye diagram. An external clock set at MHz by the signal generator helped to conduct sampling for the transmitted digital signal. The sequential monitoring process was conducted every minute. A switch driver Keysight 11713C and Keysight L7106C switches helped alternated the monitoring sequence by switching conductive paths between the channels on the DCA as shown in Fig. 4. The characteristic impedance of the test circuit shown in the figure was controlled at 50Ohms. An instrumental control software controlled all the devices in the test circuit for the automated monitoring process. The test was manually stopped when the conductor of the wire was exposed before the conductor was damaged. Additionally, two tests were stopped after fault detection by digital signal to estimate the extent of chafing after fault detection. The monitored signal, both the eye parameters and the TDR signal, were analyzed by SPRT to detect wire faults. For the training state of SPRT, both signals were measured during 24 hours before the test. Based on the training data, the defect level of the monitored signal was determined to be depending on the degradation trend of the signal. For example, (a) Figure 2. Schematic of cross sectional RG 174 Figure 3. Taber 5750 Linear abraser Figure 4. Wire abrasion (a) schematic picture 1939

5 location was arranged as time series data. According to the plot in Fig. 6 (a), the eye height was deteriorated gradually as the TDR signal was during the tests. The degradation trend of the eye height had a strong linear relationship with TDR based on the correlation coefficient. The tests 2 and 3 in Fig. 7 were stopped after the SPRT alarm for eye height was observed to estimate the extent of chafing. The damaged wires after the tests 1, 2, and 3 are shown in Fig. 8 with chafed depths. As shown in Fig. 8, the outer shields of the wires were chafed. In the case of the tests 2 and 3, the wire abrasion test chafed mm and 0.025mm of the wire dielectric based on the schematic of RG 174 in Fig. 9. Therefore, the jacket and the outer shield abrasion was enough to cause both the eye height and the TDR deterioration. A summary of each test result are shown in Table 1. The differences in the fault detection times among the tests are considered to result from the difference of wire tolerance. According to the correlation coefficients, r in Table 1, the eye height had a strong relationship with TDR. Therefore, similarly to TDR, eye height was deteriorated during the tests. The diagnosis capability of each approach can be compared (a) (a) Figure 5. Schematic (a) and picture of test circuit since the eye parameters such as eye height was decreased during the test, the defect level for eye height was determined to be. V. RESULTS The signals monitored during the first abrasion tests are shown in Fig. 6. Among the monitored eye parameters, the eye height was selected to represent the digital signal characteristics because the eye height was the most sensitive parameter [10][11]. The TDR measurement at the fault Figure 6. Plot for eye height and TDR signal during ALT (a), and scatter plot with correlation coefficient, r for test

6 (a) Figure 8. Schematic of upper half region of cross sectional RG RG-174 (unit:mm) by the difference in fault detection. The results indicate that the wire faults were detected in the roughly same time, within 30 minutes in the tests 1, 2, and 5. In the other tests, the eye height can also detect wire faults before the exposure of the conductor. Thus, the proposed approach using digital signal characteristics detected wire faults as TDR did, although TDR always detected faults faster than the eye height. Test # TABLE I. r TEST RESULTS WITH EYE HEIGHT AND TDR Fault Detection (min) Eye height Figure 7. Plot for eye height and TDR signal for test 2 (a), and test 3 TDR Difference (min) Figure 9. Schematic of upper half region of cross sectional RG 174 VI. DISCUSSION As the test results indicate, the approach using digital signal could detect a wire fault after the jacket and the outer shield were chafed. Since the eye height detected the wire fault before the conductor was exposed, the proposed approach is capable of preventing malfunctions of the wiring system by wire failures such as conductor damages. While a digital signal generator and a signal analyzer were used for the proposed approach in this study, a field-programmable gate array (FPGA) can be implemented in order to generate digital signal and analyze the characteristics of the digital signal in the field. Then, a real-time wire diagnosis can be performed without using additional analyzers when a software module for analyzing digital signal is installed in the FPGA. In the field, when an FPGA is connected to the wiring system, the proposed approach can continuously diagnose wires based on the transmitted digital signal. This advantage enables the approach using digital signal to diagnose wires in real time without using external monitoring devices or circuitries. Thus, a real-time diagnosing with digital signal can detect the onset of wire faults. 1941

7 VII. CONCLUSION This study developed a diagnostic approach to detecting wire faults in the wiring system using digital signals. In order to demonstrate the capability to diagnose wire faults using the digital signal, an accelerated wire abrasion test was designed to chafe wires gradually. During the test, the characteristics of the digital signal transmitted through the wires were monitored. TDR was monitored along with the digital signal characteristics to compare diagnosis capabilities. Wire fault was detected based on SPRT. The proposed approach detected wire faults before the conductor was damaged. In the field, wires are usually installed in commercial and industrial systems as a complex wiring system with complicated transmission lines between terminations. Diagnosis of the complex wiring system are often limited by external monitoring devices due to accessibility issues. However, the proposed method can be applied without requiring external monitoring devices. A complex wiring system can be diagnosed in-situ by appropriately deploying the signal generating and analyzing modules between both ends of a transmission line of interest within the network. Conventional approach that uses external monitoring devices usually diagnoses wiring systems only after the system operations halt for system maintenance. One advantage of the proposed approach is the real-time diagnosis capability, which allows system diagnosis while the system is in operation. Thus the time required for maintenance can be reduced. For instance, the proposed approach can perform online wiring system health monitoring of an aircraft in flight. After the aircraft lands, the time for management on the ground can be significantly reduced owing to the real-time diagnosis results. Shorter downtime yields shorter mean time to repair (MTTR), and thus resulting in higher system availability. In addition, the proposed approach can diagnose wires based on the integrity of transmitted digital signal. It can detect in real time any faults that induce degradation of the digital signal integrity during transmission. Other approaches for real-time diagnosis [6][7] can only detect faults which are analyzed to determine the fault states for internal signals. However, the proposed approach provides the extent of faults based on impedance variation on transmission lines. Thus, it has a potential to be used for real-time diagnosis of all the commercial and industrial systems where digital signal is transmitted for data communication. Our future study includes methods of improving the accuracy of the proposed approach, sensing faults without additional analyzing devices, and developing prognostic approaches to predict the remaining life time of wires in the wiring system based on the digital signal characteristics. REFERENCE [1] D. Lee and P. Arnason, U.S. navy wiring systems lessons learned, Joint Conference on Aging Aircraft, 2000 [2] C. Smith, Transport aircraft intrusive inspection project (an analysis of the wire installations of six decommissioned aircraft) final report prepared by the intrusive inspection working group, ATSRAC, Dec [3] K. Wheeler. and D. Timucin, Aging aircraft wiring fault detection survey, NASA Ames Research Center, June 2007 [4] P. Smith, C. Furse, and J. Gunther, Analysis of spread spectrum time domain reflectometry for wire fault location, IEEE Sensors Journal, vol. 5, no. 6, Dec. 2005, pp , doi: /JSEN [5] W. Hassen, F. Auzanneau, L. Incarbone, F. Pérès, and A. Tchangani, Distributed sensor fusion for wire fault location using sensor clustering strategy, Int. J. of Distributed Sensor Networks, vol.11, no.5, Apr. 2015, pp doi: [6] J. Estima and A. Cardoso, A new algorithm for real-time multiple open-circuit fault diagnosis in voltage-fed PWM motor drives by the reference current errors, IEEE Trans. Ind. Electron., vol. 60, no. 8, pp , Aug. 2013, doi: /TIE [7] T. Kamel, Y. Biletskiy, and L. Chang, "Real-time diagnosis for opencircuited and unbalance faults in electronic converters connected to residential wind systems", IEEE Trans. Ind. Electron., vol. 63, no. 3, pp , Mar. 2016, doi: /TIE [8] D. Kwon, M. Azarian, and M. Pecht, Degradation of digital signal characteristics due to intermediate stages of interconnect failure, Proc. IEEE Signal Propagation on Interconnects (SPI 10), IEEE Press, May 2010, pp.55-58, doi: /SPI [9] J. Yoon, I. Shin, J. Park, and D. Kwon, A prognostic method of assessing solder joint reliability based on digital signal characterization, Proc. IEEE Electron. Components and Technology Conf. (ECTC 15), IEEE Press, May 2015, pp doi: /ECTC [10] J. Lee, and D. Kwon, Use of Digital Signal Characteristics for Solder Joint Failure Precursors, Proc. IEEE Electron. Components and Technology Conf. (ECTC 16), IEEE Press, May 2016, pp doi: /ECTC [11] J. Lee, and D. Kwon, A digital technique for diagnosing interconnect degradation by using digital signal characteristics, Microelectronics J., vol. 60, Feb. 2017, pp , doi: [12] K. Gross, and W. Lu, Early Detection of Signal and Process Anomalies in Enterprise Computing Systems, Proc. IEEE Int. Conf. Mach. Learning Appl. (ICMLA 02), Jun. 2002, pp [13] L. Lopez, Advanced electronic prognostics through system telemetry and pattern recognition methods, Microelectronics Rel., vol 47, no. 12, Jan. 2007, pp , doi: /j.microrel [14] ISO, (2011), ISO :2011 Road vehicles V and 600 V singlecore cables -- Part 1: Dimensions, test methods and requirements for copper conductor cables, Geneva, Switzerland: ISO ACKNOWLEDGEMENT This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning (NRF-2014R1A1A ) 1942