Diagnosis, fault prediction, and Remaining Useful Life (RUL) estimation are among the predictive maintenance research subjects used for maintenance cost reduction. Using the available data with different machine learning methods, especially deep learning methods, the accuracy of estimation and prediction of faults and RUL have increased dramatically. However, due to the statistical nature of the machine learning methods and the limitations of available datasets, physically interpreting this information might be impossible. On the other hand, controlling the degradation and faults in the machines as the optimum predictive maintenance solution needs the physical interpretation of the method's outcome. In order to test the new process-based methods for degradation and fault control, datasets with more information are required (compared to available datasets). In this article, we introduce an open-source degradation simulator for linear systems. This simulator can simulate the degradation in closed-loop machines whose dynamics are known. It is also possible to simulate different degradation models for different system parts simultaneously by adding different processes and output noise to the system. This simulator can generate enough data to test new machine learning-based predictive maintenance methods.