A Novel Inverse Filtering Method for Systems with Multiple Input Signals (ICCAS 2018)

The initial motivation of this paper is the multiple cylinder pressure (CP) signals estimation in modern engines using inverse filtering. Inverse filtering is a common method for input signal estimation of single-input (SI) systems while there exist some problems estimating inputs of multiple-input (MI) systems if the output number is less than the input number, such as ill-conditioned inversion of the transfer function matrix. To deal with these problems, in this paper we introduce a novel inverse filtering approach for MI systems by converting a MI system into a SI system using delay systems and afterwards using the Kalman filter to estimate the inputs. However, it requires that the input signals have to be periodic in angle domain with same shapes, and there has a fixed and known phase difference between every two adjacent signals. These requirements are reasonable and practical for many kinds of signals in rotating machineries, e.g., the CP signals of different cylinders in an engine under stationary conditions. Finally, numerical simulations were carried out to demonstrate the performance of the proposed inverse filtering algorithm.