Concurrent estimation of time-to-failure and effective wear

We propose a novel algorithm for the estimation of remaining lifetime of a generic device based on an intuitive and widely-applicable model of failure due to accumulation of effective wear. The simultaneous quantification of an abstract, multivariate wear function is achieved by a new learning algorithm which we term “cumulative effect regression”. We develop the theory of the algorithm and compare its performance to traditional anomaly and pattern detection tools. Experimental results from X-ray tubes strongly validate the algorithm and demonstrate the utility of Operating Charactersitics (OC) curves as a powerful evaluation tool.