Identifiability of a Series of Discrete Process Cycles Based on Multirate Data

Abstract

Listen

Translate article

On-line identification of the manufacturing process based on process data is a crucial step for model-based control and diagnostics. A typical discrete manufacturing process generates multirate data streams. Whereas various sensors provide in-process information about the process, many important process outcomes such as product qualities are usually measured via postprocess inspection. This paper proposes a method for studying the identifiability of model parameters of the manufacturing process using both in-process and postprocess data. The identification of the model parameters based on multirate output is formulated using the maximumlikelihood method. The Fisher information matrix for a multirate-sampled discrete manufacturing system is derived to study identifiability of model parameters. A method to calculate the sensitivity matrices in the Fisher information matrix is also proposed. A case study is conducted using a model of metal removal in the cylindrical grinding process to demonstrate the efficacy of the proposed method for assessing the identifiability. It is observed using both sensor signal and postprocess output for identification effectively improves the identifiability.