This article considers a condition‐based maintenance for a system subject to deterioration. The deterioration is modeled by a non‐homogeneous gamma process, more precisely the gamma process and the preventive maintenance are imperfect or worse than old. The corrective maintenance actions are as good as new. The maintenance efficiency or non‐efficiency parameters as well as the deterioration parameters are considered to be unknown. The monitoring data under consideration give indirect information on the maintenance parameters. Therefore, an expected maximum algorithm is applied for parameter estimation.