Reliability testing, analysis and prediction of balancing resistors

Abstract

This paper focuses on analytical maintenance planning using reliability engineering techniques applied to balancing resistors used in drives. Balancing resistors are one of the critical components of alternating current (AC) drives. The primary function of these resistors is to balance the voltage of the capacitor bank. Failure of these resistors can lead to catastrophic failure of the drive. A complete understanding of the failure modes and causes of these devices and their associated reliability is essential for a reliable design and safe operation of the AC drives. Design failure mode and effect analysis (DFMEA) and testing were performed to identify failure modes of balancing resistors and their associated stresses. Temperature was found to be the primary stress factor causing the failure of the resistor and the associated failure mode was found to be rupture of the resistor element due to creep at high temperature. Performing an accelerated life tests (ALT) on these devices and using the test results to predict reliability at use condition requires establishing the appropriate accelerated stress level. There are two competing factors during ALT. First, the accelerated stress level needs to be high to keep the tests duration within development time constraints. Second, the accelerated stress levels must not introduce failure modes different from those at operating conditions. Highly accelerated life tests (HALT) were performed at two stress levels to establish ALT stress level. Failure analysis of the failed devices at HALT revealed that the failure modes are identical for both levels. These tests guided us to choose the accelerated stress level for the ALT. A failure analysis of the failed devices under ALT conditions also revealed that the failure mode is similar to that of the HALT conditions. 2-parameter Weibull distribution was found to fit the ALT data. Shape and scale parameters of the Weibull distribution were evaluated using Minitab and maximum likelihood (ML) method. Since the failure mode of the HALT and ALT were identical, we deduced that the failure distribution for the tests at HALT would also follow a Weibull distribution of the same shape parameter as the ALT but with a different scale parameter. At the conclusion of HALT, we only had knowledge of the number of resistors tested, test duration and number of failed resistors. Weibayes Maximum Likelihood equation was used to determine Weibull scale parameters at 95% confidence level for the HALT tests. The scale parameters under each accelerated stress condition for both HALT and ALT followed a power function. Hence ln (power) transformation was used to establish life vs. stress relationship. The acceleration factor (AF) was then determined using the established relationship. Since lower bound estimates for scale parameters of HALT were used, the estimates on reliability under normal use are conservative. Knowing the failure distribution at ALT and AF, we used Minitab to evaluate the reliability at normal operating condition. It turned out that the reliability was not acceptable. Modification to the design such as using thermal interface grease at the mounting point of the resistor was introduced and tested. The modification brought the reliability to an acceptable level.