The Additive Reliability Model of Cyber Physical Systems Based on Local Polynomial Regression with Masked Failure Data

In this paper, the cyber physical system is divided into a software system and a hardware system, and both the software system and the hardware system each contain several subsystems. To solve the difficult problem of parameter estimation in parametric reliability models such as the non-homogeneous Poisson process under masked data, this paper proposed an additive reliability model of cyber physical systems based on local polynomial regression under masked data. This model uses the multinomial distribution to allocate the masked data and employs the local weighted least squares method to solve the reliability model. Finally, by using a set of open-source software failure data and a set of simulation data in the cyber physical system, this paper conducts a performance comparison and analysis between the proposed non-parametric reliability model and traditional reliability models. The empirical results show that the proposed reliability model performs better in terms of the fitting effect and demonstrates stronger applicability and superiority.