Algebraic and FFT-Based Methods for Discrete-Time Matrix Convolutions with Applications to Semi-Markov Models
This work provides more efficient computational tools for researchers and practitioners working with semi-Markov models, particularly for evaluating reliability and availability functions, representing an incremental improvement in computational efficiency.
This paper investigates matrix-valued sequence convolution products for Markov renewal equations, developing FFT-accelerated convolution and Newton-type inversion schemes. The framework is applied to discrete approximations of matrix Stieltjes convolutions in semi-Markov models, demonstrating substantial runtime reductions while maintaining accuracy.
We study the convolution product of matrix-valued sequences and its role in the computation of Markov renewal equations. Explicit representations and recursive formulae for the convolutional inverse are derived and used to construct FFT-accelerated convolution and Newton-type inversion schemes, together with a Gauss--Jordan alternative in truncated power-series rings. The proposed framework is also applied to discrete approximations of matrix Stieltjes convolutions, which arise in continuous-time semi-Markov models. These tools are then used for the numerical evaluation of semi-Markov reliability and availability functions. The numerical results show substantial reductions in runtime, while preserving close agreement with exact benchmark solutions, direct computations, and Monte Carlo simulations.