Nonparametric modeling and optimal maintenance using online monitoring in environments with seasonal variation

Preventive maintenance intervals for a product or system are generally specified by the manufacturer without consideration of the usage environment or usage frequency. This means that the intervals may not be optimal for each product or system instance if degradation and/or failure depend not only on time but also certain other factors. For example, some products tend to fail if the temperature is lower than a certain value. Product lifetime is usually in terms of calendar time or usage time. Online monitoring system enables us to identify such features. We investigated the failures of a product for which lifetime is in terms of usage time. A set of real was analyzed using a logistic regression model on the usage time scale with seasonal effects. On the basis of our results, we propose an optimal maintenance policy with lower expected costs than the existing one.