Prognostics Implementation Methods for Electronics Jie Gu, University of Maryland Nikhil Vichare, PhD, University of Maryland Terry Tracy, Raytheon Company Michael Pecht, PhD, University of Maryland Key Words: electronics, failure mode, failure mechanism, prognostics, reliability SUMMARY & CONCLUSIONS Prognostics is a method that enables monitoring the state of reliability of a product in real time, and therefore can be used to provide advance warning of a failure, to minimize unscheduled maintenance, to provide condition-based maintenance, and to help in product design and development. This paper identifies six levels of prognostics implementation for electronics, from on–chip packaging to complete systems of systems. An approach is then presented for selecting the implementation levels to cost–effectively optimize coverage. The process of selecting the prognostic approach and its implementation at various levels in electronics is enabled using failure modes, mechanisms, and effects analysis. 1 INTRODUCTION The purpose of prognostics is to identify potential failures in advance and to provide the information necessary for risk mitigation and management. For example, this information can be used to minimize unscheduled maintenance or to extend maintenance cycles by the use of condition-based maintenance management. Prognostics has been used for many applications, including aerospace vehicles, civil infrastructures, nuclear facilities, and mining machinery [1,3]. The need for prognostics of electronic systems is growing since most control functions are being conducted by electronics. Several studies have proposed techniques for applying prognostics to electronic products and systems, includes field effect transistors, power converters [9], printed circuit boards (PCB) and interconnects [2], global positioning systems, and enterprise servers [17,18]. These efforts have predominantly focused on prognostics implementation for custom–specific configurations. At this time, there has been no literature to address the question of how to begin the prognostics implementation process for a new or legacy system, and at what level(s) the implementation makes sense. This extremely important question directly influences the implementation costs and also the costs incurred due to failures that could have been prevented by prognostics implementation. This paper addresses this question, to enable asset managers to provide maximum failure/fault coverage using minimum resources (sensors, implementation costs). 2 PROGNOSTIC APPROACH FOR ELECTRONICS There are three broad categories of prognostic implementation: (1) using expendable prognostic cells, such as “canaries” and fuses, that fail earlier than the host product to provide advance warning of failure; (2) monitoring and reasoning of parameters, such as shifts in performance parameters, progression of defects, that are precursors to impending failure; and (3) modeling stress and damage in electronics utilizing exposure conditions (e.g., usage, temperature, vibration, radiation) coupled with physics–of– failure (PoF) models to compute accumulated damage and assess remaining life[3]. Prognostic cells, such as fuses and canary devices, are mounted on or incorporated into product to provide advance warning of failure for specific wear–out failure mechanisms. The time to failure of these prognostic cells can be pre– calibrated with respect to the time to failure of the actual product. Because of their location, these cells experience substantially similar dependencies as does the actual product. Stresses that contribute to degradation of the circuit include voltage, current, temperature, humidity, and radiation. Since the operational stresses are the same, the damage mechanism is expected to be the same for both the prognostic cell and the actual circuit. However, the prognostic cell is designed to fail faster due to increased stress on the cell structure by means of scaling. For example, scaling can be achieved by a controlled increase of the current density inside the cell. With the same amount of current passing through both circuits, a higher current density is achieved if the cross–sectional area of the current–carrying paths in the cell is decreased. A failure precursor is an event that signifies impending failure. A precursor indication is usually a change in a measurable variable that can be associated with subsequent failure. For example, a shift in the output voltage of a power supply might suggest impending failure due to a degrading feedback regulator and opto–isolator circuitry. Failures can then be predicted by using a causal relationship between a measured variable that can be correlated with subsequent failure. The life–cycle loads of a product can be generated from manufacturing, shipment, storage, handling, operating and