APPLICATIONS OF IN-SITU HEALTH-MONITORING AND PROGNOSTIC SENSORS Jingsong Xie and Michael Pecht CALCE Electronic Products and Systems Center University of Maryland College Park, MD 20742, USA
[email protected] [email protected] ABSTRACT This paper reviews the applications of in-situ health monitoring and prognostic sensors with the focus on those in electronic systems. It provides an insight into the issues of utilizing in-situ sensors in electronic systems for health monitoring and prognostics. Those issues are discussed in comparison to those for mechanical systems and structures. This paper also introduces a physics-of-failure based life-consumption monitoring methodology for the health monitoring and life prediction of electronic systems. Key words: in-situ sensor, health monitoring and prognostics, electronics reliability INTRODUCTION Many systems and structures, such as avionic systems, aerospace vehicles, civil infrastructures, nuclear facilities, and mining machinery, are designed for long-term operations and these operations need periodic maintenance and/or replacement for safety and reliability. Reliability is essential in these applications because failure of these systems and structures can lead to serious loss of lives and/or properties [1,2,3]. Health monitoring and prognostics (HMP) is a proactive system to identify fault and determine health conditions or likely failure occurrence [1,3,4]. It provides information for maintenance and replacement so that a failure can be prevented in advance. In addition, prognostics in a HMP system also provides prediction of life consumption and remaining life, which combined with other information on health conditions offers the possibility of maintenance and replacement on a complete per-need basis, and hence, the possibility of significant cost savings [1,5,6]. The health monitoring and maintenance of many systems and structures has so far been depending upon visual inspection and testing that are time consuming, labor intensive, and costly [3,6,7]. Using in-situ sensors in health monitoring and prognostics enables continuous monitoring of a system or structure with minimal human involvement and with no interruption of operation. It eliminates the need of disassembly or prior knowledge of likely damage locations in the health monitoring and maintenance operations.
Utilizing smart sensors developed with the latest technologies in microelectronics, micro-electro-mechanical systems (MEMS), and nano-technology has become a clear trend in health monitoring and prognostics [1,8,9]. These sensors provide extended sensing capability and performance, improved design flexibility for the in-situ monitoring applications, and reduced cost. Combined with communication and networking technologies, particular the wireless networking technologies, the latest in-situ HMP sensors offer the possibility for developing a fully autonomous, condition-based maintenance management system [3,6,10,11,12]. This paper reviews the applications of in-situ health monitoring and prognostic sensors with the focus on those in electronic systems. It provides an insight into the issues of using in-situ sensors in electronic systems for health monitoring and prognostics. Those issues are discussed in comparison to those for mechanical systems and structures. This paper also introduces a physics-of-failure based lifeconsumption monitoring methodology that is now under development in CALCE Electronic Products and Systems Center at the University of Maryland for in-situ health monitoring and life prediction of electronic systems. IN-SITU HMP SENSORS IN TRADITIONAL APPLICATIONS Health monitoring and prognostics (HMP) has traditionally been employed on safety-critical mechanical systems and structures, such as propulsion engines, aircraft structures, and railroad suspension systems [5]. Even in today, the applications of HMP systems on mechanical machineries and structures remain dominant compared with other applications, such as electronics systems. The reasons for this can be summarized from the following perspectives: First, if operations are carried out under designated conditions, majority of failures now seen on properly tested and qualified mechanical systems are of wear-out failures with typical failure mechanisms of fatigue, creep, and corrosion. People have gained sufficient knowledge and developed technologies to avoid unexpected sudden failures like those occurred in the incident of the Liberty ships during the World War II. Unlike over-stress failures,
wear-out failures are a result of damages accumulated in one or multiple processes. These failures can be prevented using periodic maintenance and replacement provided that the damage can be detected in advance. Fortunately, people have obtained enough knowledge and have mature technologies of damage detection and fault diagnose on mechanical systems. These knowledge and technologies provide a fundamental basis for developing a feasible and an effective health monitoring and diagnose system for mechanical machineries and structures. As a result, people have been able to ensure a reliable operation of a mechanical system in each mission and use maintenance to achieve the reliable operation of the system for its entire cycle of life.
Rotor blade vibration, blade tip clearance, fatigue accumulation, foreign object damage (FOD) detection, and stall [5,13,14,15,17] Static and dynamic structural deformation in high-temperature environment [16] Crack propagation and remaining life prediction [1] Rotor head fault detection and fatigue life prediction [1]
It is well known in the reliability engineering community that redundancy is an effective approach to improve the overall operational reliability of a system without improving technology itself. Although redundancy can be integrated into mechanical systems in some cases, such as the use of multiple instead of one engine on an aircraft, designing redundant mechanical systems is feasible in engineering only for very limited cases. Therefore, the health monitoring and prognostic approach, combined with periodic maintenance, needs to play a major role in highreliability operation of mechanical systems and structures. In addition, health monitoring and prognostics of mechanical systems and structure has a much longer history than electronic systems. An enhanced HMP system utilizing the latest in-situ sensing technologies sees a broader need and economic benefit on the mechanical systems and structures than on the electronic systems. People have had a much more ready and mature sensing, fault diagnose, and reliability prognostic technologies for mechanical systems than for electronic systems. Table 1 provides examples of in-situ health monitoring and prognostics applications on mechanical systems and structures. Other applications include motors and hydraulic oils for mining and other machineries [21,22,23,24].
Table 1. Examples of in-situ HMP sensor applications on mechanical systems and structures Application Category Jet engine and other Turbomachinery
Application Crack propagation in rotating disks and blades [6]
Sensing Mechanism Crackwire
Aircraft Structures
Corrosion monitoring [1,2]
Structural damage [18]
Large infrastructure
Composite structure damage, fatigue [19,20] Deformation monitoring [3,10]
Capacitive sensing, eddy current, infrared, optical/laser, micro/millimeter wave
Ceramic strain gage
Acoustic emission and fatigue cycle measurement Acoustic emission and fatigue cycle measurement Electrochemical sensing, optical fiber Fiber-optic accelerometer and displacement, ultrasonic, acoustic emission Piezoelectric/ piezoceramic sensing Inertia sensing
IN-SITU HMP APPLICATION IN MICROELECTRONIC DEVICES AND SYSTEMS Compared to mechanical systems, using in-situ sensors in electronic devices for the health monitoring and prognostics purpose is an emerging application. Some build-in IC sensors and cells that are designed for the reliable operation of safety-critical ICs have been developed and introduced [25,26,27,28]. However, in contrast to a long application history of in-situ HMP sensing for mechanical systems, those sensors and cells for electronic system remain in a development stage. Many issues need to be addressed before any actual applications become possible. So far, no reports have yet shown
services of these HMP sensors and cells in any actual applications. • Unlike mechanical systems and structures, electronic devices lack technologies in general for fault detection and diagnose. Fault of electronic devices are not limited to physical damages or defects. They also vary significantly in nature. Even for the physical damages, many of them are in a micron, sub-micron or even nano-scale and do not necessarily link to failures or a loss of designated electrical performance or function. Therefore, health monitoring and prognostics of electronic systems lacks straightforward approaches to proceed, compared to that of mechanical systems and structures. It can help to improve the mission reliability of an electronic system by integrating an in-situ HMP system. However, considering the complexity of an electronic system, it can also be difficult to quantify the improvement on a solid theoretical or experimental basis. The proper functioning of most electronic systems depends on both the hardware and the software (also including firmware) reliability of the system. Hence, an electronic system failure can be a result of hardware problems, software problems, or both problems. However, in-situ sensors and HMP systems are effective primarily on hardware failures. Therefore, using the HMP systems alone is not able to ensure highly reliable operation of complex electronic systems. In contrast to mechanical systems and structures, which encounter inherent difficulty in implementing redundancies, design of redundancy is almost always feasible to an electronic system from the engineering standpoint. A simple theory on how to quantify the reliability improvement with redundancy is also available. To be able to evaluate reliability in a quantitative manner is essential to many safety-critical applications, such as avionics and nuclear operations, due to the specified limits in reliability that the federal government requires on the operation of certain electronic devices in these applications. Because of these issues, in-situ health monitoring and prognostics is unlikely to replace the role of redundancy in the high-reliability operations of electronic systems. The potential application of in-situ HMP sensors in electronic devices will be likely used more for the purpose of cost reduction in maintenance and replacement than for the purpose of reliability improvement. For an in-situ HMP system to be implemented in actual electronic systems, other issues that need to be addressed include: • •
Dependability of data interpretation and end-oflife prediction Additional design, manufacturing and operating cost for integrating an in-situ HMP system vs. the
cost saving for using the system in maintenance and replacement Operating reliability of the in-situ sensing and data processing system itself
IN-SITU HMP CELLS There are two types of in-situ cells or built-in sensors developed for the health monitoring and/or life prediction of ICs. One type is a circuit to monitor the parameters that are considered as indicators of reliability and can be used to define failures. The other type is simply one or multiple independent cells or devices whose failure correlates the health condition of the actual circuit to be monitored. Both types are additional testing circuitry parasitic to actual functional cores. An exa mple of the first-type cells is built-in-current sensors (BICS) to detect common defects of ICs due to normal wear-out or environmental stresses, such as bridging, open, and parasitic transistor defects by measuring the quiescent power supply current (IDDQ ) of an IC [25]. People has developed a quiescent current monitor (QCM) system that consists of multiple BICS and digital control logic to detect elevated IDDQ current in real-time during system operation. The QCM performs leakage current measurements on every transition of a system clock to get maximum coverage of the IC in real-time. BICS are placed on the end of every standard cell row of the monitored system core logic and each one is used to monitor the quiescent current for its row and contributes to a single state-of-health output. The quiescent current above a certain level indicates an unacceptable condition or failure of an IC circuit. The InstaCellTM prognostic cell developed by the U.S. Air Force Ridgetop Group is a good example of the secondtype cells [26]. Since the cells are designed and manufactured with the same technology and in the same process with the actual circuit, the health monitoring and prognostic methodology assumes that the functional performance and the reliability condition of the monitor cells can represent that of the actual circuit to be monitored. If there are any factors that affect the reliability of the actual circuit in its life cycle environment, they will also affect the monitor cells in the same way. By building the monitor cells in defect-seeded, but calibrated structures and/or being operated under elevated stress conditions, the monitoring cells can have higher failure rate and shorter operating life than the actual circuit. They are calibrated and designed to fail at different time prior to the failure of the actual circuit, working like a “life-consumption scale”. The life condition of the actual circuit to be monitored is then determined by the scale of the monitor cells. PHYSICS-OF-FAILURE-BASED LIFE CONSUMPTION MONITORING The parameters that are monitored in an in-situ HMP system can generally categorized as those that can be used
to define the failures of the systems to be monitored and those that cannot. The length of crack in aircraft structures [6], IDDQ in the health monitoring of ICs [25] and oil consumption in Boeing’s extended-range twin-engine operations (ETOPS) program [29] are examples of the first category of parameters. The measurement results of the parameters of this category in an in-situ monitoring environment provide a direct indicator of health condition and reliability. The parameters of the second category are primarily those parameters that are measures of environmental and stress conditions, such as temperature, humidity, vibration and radiation. In general, the technologies that are required to monitor the parameters of the first category are application-specific, while those technologies required for monitoring the second categories are usually standardized and may have multiple options in selection of monitoring devices. Based on parameters that need to be monitored, an in-situ HMP regime can also be in two categories: a test-based system and a physics -of-failure (PoF) based system. A testbased in-situ HMP system requires direct measurement or monitoring of the parameters that can be used to define failures. With this approach, the health condition of an electronic system to be monitored can be determined by benchmarking the test results against the failure criteria defined in the reliability requirements. The life consumption and remaining life prediction can be achieved by extrapolating the test data to the defined failure point with or without a degradation model available on the parameters. However, in actual application of complex electronic systems, defining and monitoring failure-indicating parameters can be difficult and sometimes it may not be feasible from the perspective of engineering and business. Another danger with this HMP system is that it does not require a thorough understanding of failure mechanisms that govern the failure processes. As a result, the parameters that are monitored in the process may not be sensitive enough for effective monitoring until the ultimate failure occurs. The alternative approach is to adopt a PoFbased in-situ HMP system. This system considers actual environmental and loading conditions, and requires the understanding of the dependence of failure on those external environmental and loading conditions. In return, the health monitoring and prognostics of an electronic system can be completed basically by monitoring the system’s life-cycle environment [30,31,32,33]. Table 2 provides an example to explain the process of using the PoF-based approach in health monitoring of electronic systems. Based on the actual environmental conditions recorded by the in-situ sensors and used as inputs to the failure models of identified major failure mechanisms, an estimated total time-to-failure can be obtained for this conditions in a given monitoring interval.
The percentage life consumed in this interval is hence determined by Consumed percentage life = time interval / estimated total time-to-failure × 100 Adding up the consumed percentage life for all the monitoring intervals gives the total life consumption so far Total life consumed so far = Σ percentage life consumed at each interval Considering the variations in operating condition and percentage life consumption at different intervals, an updated estimation on the total time-to-failure so far can be obtained by Updated total time-to-failure = total time monitored / total life consumed so far Total time monitored = Σ all time intervals Finally, estimation on the remaining life up to the current monitoring interval is determined by Remaining life = updated total time-to-failure total time monitored A physics -of-failure based life-consumption monitoring methodology is being developed in CALCE Electronic Products and Systems Center at the University of Maryland for in-situ health monitoring and prognostics of electronic systems [30,31,32,33]. This is an on-going effort made on failure modeling and health monitoring process development. Studies have also been conducted on experimental verification of this approach. Figure 1 shows an example of the results obtained these studies completed at CALCE Electronic Products and Systems Center [33]. In this case, solder joint fatigue was identified as the dominant failure mechanism and the estimated life of 61 days was obtained using the in-situ life consumption monitoring. The result correlates well to the measured result of 66 days in end-of-life.
Table 2. An example of using the PoF-based approach in in-situ electronics HMP Monitoring schedule in days
1
2
3
4
5
6
7
Total
Estimated timeto-failure (days)
8
5
12
7
18
7
4
N/A
Consumed percentage life (%)
12.5 20.0 8.3 14.3 5.6 14.3 25.0 100.0
Estimated total life (days)
8.0
6.2
7.3
7.3
8.2
8.0
7.0
N/A
Estimated remaining life (days)
7.0
4.2
4.3
3.3
3.2
2.0
0.0
N/A
70
Remaining Life (days)
60 50 40 30 Estimated life based on SAE
20 environmental handbook data- 34 days 10
Actual life from resistance monitoring - 66 days
Estimated life based on LCM - 61 days
0 0
5
10 15 20 25 30 35 40 45 50 55 60 65 70 Time in Use (days)
Figure 1. A demonstration of the PoF-based in-situ HMP methodology by comparing the actual end-of-life measurement results and the HMP results using the PoFbased methodology [33]
SUMMARY With potential applications in the health monitoring and prognostics of electronic systems, two types of in-situ cells and sensors have been developed. However, for these emerging technologies to be used in actual applications of electronic systems, many issues need to be addressed. Considering the engineering feasibility in sensor implementation, in-situ HMP sensors are unlikely to play a major role in applications, such as avionics, that require stringent reliability operation with specified quantitative limits imposed by the federal government. They will be likely used more for the purpose of cost reduction in condition-based maintenance and replacement than for the purpose of reliability and safety. A physics -of-failure-based in-situ HMP methodology for electronic systems is under development in CALCE Electronic Products and Systems Center at the University
of Maryland. This methodology offers a feasible and a cost-effective alternative to the directly test-based healthmonitoring and life-prediction approach. REFERENCES [1] Lewis, S.A.; Edwards, T.G.; “Smart sensors and system health management tools for avionics and mechanical systems,” Digital Avionics Systems Conference, 1997. 16th DASC, AIAA/IEEE, Volume: 2, 26-30 Oct. 1997, Page(s): 8.5 -1-8.5-7 vol.2 [2] Cooper, K.R.; Elster, J.; Jones, M.; Kelly, R.G.; “Optical fiber-based corrosion sensor systems for health monitoring of aging aircraft,” AUTOTESTCON Proceedings, 2001. IEEE Systems Readiness Technology Conference, 20-23 Aug. 2001, Page(s): 847 -856 [3] Wang, D.H.; Liao, W.H.; “Instrumentation of a wireless transmission system for health monitoring of large infrastructures,” Instrumentation and Measurement Technology Conference, 2001. IMTC 2001. Proceedings of the 18th IEEE, Volume: 1, 2123 May 2001, Page(s): 634 -639 vol.1 [4] N. Kelkar, D. Dasgupta, M. Pecht, I. Knowles, M. Hawley, and D. Jennings, “Smart Electronic Systems for Condition-Based Health Management,” Quality and Reliability Engineering International, Vol. 13, pp. 3-7, 1997. [5] von Flotow, A.; Mercadal, M.; Tappert, P.; “Health monitoring and prognostics of blades and disks with blade tip sensors,” Aerospace Conference Proceedings, 2000 IEEE, Volume: 6, 18-25 March 2000, Page(s): 433 -440 vol.6 [6] Gerardi, J.J.; Archambault, J.; Brooks, W.G.; “Wireless crackwire sensor for in-situ engine monitoring,” Aerospace Conference, 2003. Proceedings. 2003 IEEE, Volume: 7, March 8-15, 2003, Page(s): 3045 -3052 [7] Blemel, K.; Furse, C.; “Applications of microsystems and signal processing for wiring integrity monitoring, Aerospace Conference, 2001, IEEE Proceedings, Volume: 7, 10-17 March 2001, Page(s): 7 -3253 vol.7 [8] Iturralde, A.; “A review of sensing technologies for semiconductor process applications,” Semiconductor Manufacturing, 1995, IEEE/UCS/SEMI International Symposium on 17-19 Sept. 1995, Page(s): 120 -126 [9] Krabach, T.; “Breakthrough sensor technology for space exploration in the 21st century,” Aerospace Conference Proceedings, 2000 IEEE, Volume: 6, 1825 March 2000, Page(s): 565 -569 vol.6 [10] Kinawi, H.; Taha, M.M.R.; El-Sheimy, N.; “Structural health monitoring using the semantic wireless: a novel application for wireless networking,” Local Computer Networks, 2002. Proceedings. LCN 2002. 27th Annual IEEE Conference on, 6-8 Nov. 2002, Page(s): 770 –778 [11] Harman, R.M.; “Wireless solutions for aircraft condition based maintenance systems,” Aerospace
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
Conference Proceedings, 2002. IEEE, Volume: 6, 916 March 2002, Page(s): 6-2877 -6-2886 vol.6 Beach, R.D.; Moussy, F.; “Totally implantable realtime in vivo video telemetry monitoring system for implant biocompatibility studies,” Instrumentation and Measurement, IEEE Transactions on, Volume: 50 Issue: 3, June 2001, Page(s): 716 -723 Tappert, P.; von Flotow, A.; Mercadal, M.; “Autonomous PHM with blade-tip-sensors: algorithms and seeded fault experience,” Aerospace Conference, 2001, IEEE Proceedings, Volume: 7, 1017 March 2001, Page(s): 7 -3295 vol.7 Dowell, M.; Sylvester, G.; “Turbomachinery prognostics and health management via eddy current sensing: current developments,” Aerospace Conference, 1999. Proceedings. 1999 IEEE, Volume: 3, 6-13 March 1999, Page(s): 1 -9 vol.3 Dowell, M.; Sylvester, G.; Krupp, R.; Zipfel, G.; “Progress in turbomachinery prognostics and health management via eddy-current sensing,” Aerospace Conference Proceedings, 2000 IEEE, Volume: 6, 1825 March 2000, Page(s): 133 -143 vol.6 Gregory, O.J.; Qing Luo; “Ceramic strain gages for propulsion health monitoring,” Digital Avionics Systems Conferences, 2000. Proceedings. DASC. The 19th, Volume: 2, 7-13 Oct. 2000, Page(s): 6E2/1 -6E2/8 vol.2 Jones, R.; Hazell, M.S.; “The in-situ measurement of aero-engine blade flap using an optical fibre sensor,” Advanced Vibration Measurements, Techniques and Instrumentation for the Early Prediction of Failure, IEE Colloquium on, 8 May 1992, Page(s): 6/1 -6/5 Larson, E.C.; Parker, B.E., Jr.; “A subspace-based approach to structural health monitoring” Digital Avionics Systems Conferences, 2000. Proceedings. DASC. The 19th, Volume: 2, 7-13 Oct. 2000, Page(s): 6C5/1 -6C5/8 vol.2 Griffin, S.F.; Bronowicki, A.J.; Betros, R.S.; “Piezoceramic sensors and actuators for smart composite structures,” Decision and Control, 1991, Proceedings of the 30th IEEE Conference on, 11-13 Dec. 1991, Page(s): 2543 -2545 vol.3 Edmonds, J.; Hickman, G.; “Damage detection and identification in composite aircraft components,” Aerospace Conference Proceedings, 2000 IEEE, Volume: 6, 18-25 March 2000, Page(s): 263 -269 vol.6 Trutt, F.C.; Sottile, J.; Kohler, J.L.; “On-line condition monitoring of induction motors”, Industry Applications Conference, 2001. Thirty-Sixth IAS Annual Meeting. Conference Record of the 2001 IEEE, Volume: 2, 30 Sept.-4 Oct. 2001, Page(s): 1369 -1375 vol.2 Trutt, F.C.; Sottile, J.; Kohler, J.L.; “Online condition monitoring of induction motors,” Industry Applications, IEEE Transactions on, Volume: 38 Issue: 6, Nov.-Dec. 2002, Page(s): 1627 -1632
[23] Rae Duk Lee; Han Jun Kim; Semenov, Yu.P.; “A coil-type capacitive sensor for measurement of the deterioration of engine oil,” Precision Electromagnetic Measurements, 2002. Conference Digest 2002 Conference on, 16-21 June 2002, Page(s): 184 -185 [24] Ozanyan, K.B.; Carey, S.J.; Hindle, F.P.; McCann, H.; Winterbone, D.E.; Young, S.W.; Black, J.; “Alloptoelectronic sensors for combustion-related processes,” Advanced Sensors and Instrumentation Systems for Combustion Processes (Ref. No. 2000/080) IEE Seminar, 27 June 2000, Page(s): 8/1 8/4 [25] Smith, P.A.; Campbell, D.V.; “A practical implementation of BICS for safety-critical applications,” Defect Based Testing, 2000. Proceedings. 2000 IEEE International Workshop on, 30 April 2000, Page(s): 51 -56 [26] InstaCell® Diagnostic/Prognostic IP Library, Ridgetop Group, Inc., January 30, 2002. [27] M. Pecht, M. Dube, M. Natishan, and I. Knowles, “An Evaluation of Built-In Test,” IEEE Transactions on Aerospace and Electronic Systems, Vol. 37, No. 1, pp. 266-272, January 2001. [28] P. Lall, M. Pecht, and E. Hakim, Influence of Temperature on Microelectronics and System Reliability, CRC Press, New York, NY, 1997. [29] The Boeing Company, “ETOPS Maintenance,” Aero Magazine, No. 7, July 1999. [30] Mishra, S., Pecht, M., and D. Goodman, “In-situ Sensors for Product Reliability Monitoring”, IEEE/SPIE Proceedings of Design, Test, Integration and Packaging of MEMS/ MOEMS, pp. 10-19, Cannes, France, May 2002. [31] Ramakrishnan and M. Pecht, “Implementing a Life Consumption Monitoring Process for Electronic Product, Accepted for publication in IEEE Components Packaging and Manufacturing Technology. [32] Mishra, S., Pecht, M., Smith, T., McNee, I., and R. Harris, “Life Consumption Monitoring Approach for Remaining Life Estimation”, European Microelectronics Packaging and Interconnection Symposium, IMAPS, pp. 136-142, Cracow, Poland June 2002. [33] Mishra, S., and M. Pecht, “Remaining Life Estimation for Electronic Products Using Environmental Monitoring” – submitted to IEEE Components and Packaging Technologies
