Authors: A.A. Ilumoka and H. Tan
Affilation: UNIVERSITY OF HARTFORD, United States
Pages: 120 - 123
Keywords: MEMS reliability, neural networks, failure probability, quality enhancement
MicroElectronic and Mechanical Systems (MEMS) are a blend of microelectronic and mechanical devices closely coupled and simultaneously manufactured to monitor, analyze, and control everyday processes such as automobile function, biological systems operation and aeronautical flight. The work reported here establishes a methodology for failure probability prediction for microengines – a type of MEMS - using neural networks. The time-to-failure for each manufactured microengine is a random variable with a known probability density function (pdf). The approach here is to use a neural network to predict the pdf of failure time rather than a specific random failure time. A database of microengine attribute data (fabrication process, physical specifications, operating environment) and microengine performance data (cycles-to-failure) derived from actual measurements on fabricated microengines is employed. First, the database is partitioned into training (80%) and validation data (20%), then a backpropagation neural network is applied to the training data to produce an accurate mapping between microengine attributes and microengine cycles-to-failure. For failure probability prediction, microengine attributes constituted the inputs to the neural network while time-to-failure statistics (mean, median and shape parameters) constituted neural network outputs. Once neural network training was complete, validation data was used to verify predicted microengine failure statistics. Further, by reversing the inputs and outputs to the neural network, optimal MEMS attributes for robust and reliable performance (i.e. lower failure probability) can be determined after training a second neural network.