Today, equipment, sensors, and the Internet of Things have enabled remote monitoring and control of production equipment in less time. These technologies can be used for condition-based maintenance to reduce costs and increase reliability. In addition to maintenance actions, another approach to controlling system failure is adjusting the production rate, which is also known as condition-based production. In this paper, both condition-based maintenance and production decisions are integrated. We study the problem of optimizing the integrated production and maintenance of a deteriorating machine, where maintenance delays are considered. A finite planning horizon that includes several periods is considered and at the beginning of each period, the state of the system will be obtained. In each period, the production system has predetermined demand and random failures occur based on the production rate. If the machine enters a failure state, corrective maintenance must be scheduled. Otherwise, the decision-maker can choose to either schedule preventive maintenance or decide on the production rate. The problem is modeled within the framework of a Markov Decision Process, where the total production and maintenance costs over a finite planning horizon are minimized. Finally, the proposed model is solved using the value iteration algorithm to determine the optimal policies. A numerical example and case study are provided to demonstrate the applicability of the model. In this study, the proposed Condition-Based Maintenance and Production policy is compared with fixed production and maintenance policies. According to the results obtained, the proposed dynamic Condition-Based Maintenance and Production policy outperforms the static policies. The problem presented for the case where the maintenance delay is zero is also modeled, and the results for the Condition-Based Maintenance and Production dynamic policy are presented. In addition, in this case, the value of the optimal initial state for different constant policies is obtained and compared. In this case, as well, the dynamic policy performs better than other static policies.
