A Probabilistic Approach for Reliability and Life Prediction of Electronics in Drilling and Evaluation Tools

Abstract


 
 
 The capability to predict performance and lifetime of drilling electronics is the key to preventing costly downhole tool failures and ensuring success of any drilling operation. Drilling electronics operate under extremely harsh downhole environments with temperatures beyond 150C and vibration levels exceeding 15g. In addition to temperature and vibration, there are several factors affecting electronic reliability that have high uncertainty and cannot be accurately measured. There is a growing trend in the oil and gas industry to drill faster and operate at higher temperatures and pressures, forcing tools to operate beyond design specifications. This has resulted in increased failure rate leading to higher maintenance costs and system downtime for drilling operators as well as service providers. This paper develops a methodology to estimate the life of drilling electronics by using operational data, drilling dynamics and historical maintenance information. The methodology combines parameter estimation techniques, statistical reliability analysis and Bayesian math in a probabilistic framework. Parameter estimation is used to calibrate statistical equations to field data and probabilistic analysis is used to obtain the likelihood of failure. In the paper, the model parameters are represented as random variables, each with a probability distribution. Drilling electronics under downhole conditions can have several failure modes and each failure mode can be caused by the interaction of several variables. When information on each failure mechanism is not readily available, the failure is expressed in terms of several candidate models. Bayesian updating is used to incorporate real time operational history for a specific part and select the most accurate failure model for that part. Tis is for the first time, a systematic approach is developed for predicting the life of electronics in downhole drilling environments using statistical modeling and probabilistic methods on life cycle history and operational data from the field.