Content provided in partnership with

Manufacturing Industry

Model-based tampering for improved process performance-an application to grinding of shafts

Journal of Manufacturing Processes, 2003 by Palanna, Rajkumar, Bukkapatnam, Satish, Settles, F Stan

Abstract

This paper presents the model-based tampering (MBT) methodology to address performance degradation in complex nonlinear processes. MBT involves characterizing the dynamics of a process through experiments to derive nonlinear stochastic differential equation (n-SDE) models that can accurately and compactly represent a process, and using these models to dynamically compensate for the effects of process degradation on the performance. This research has yielded a new method to derive n-SDE models of complex processes through a signal-based reconstruction of Fokker-Planck equations.

Most Popular Articles in Business: Research and Markets : Tesco Plc - SWOT Framework Analysis; Do Us a Flavor - Ben & Jerry's Issues a Call for Euphoric New Flavors; eBay made easy: ready to start an eBay business? These 5 simple steps will ...; Katrina's lawsuit surge: a legal battle to force insurers to pay for flood ...; Wal-Mart's newest distribution center opened last month near the southwest ...; More »

MBT was applied to control surface finish during cylindrical grinding of shafts of air bearings used in aircraft environmental control systems (ECS)-a core competency of Honeywell. It has been shown that n-SDE models help capture patterns in process outputs that cannot be captured using any simpler models, and that MBT reduces the surface finish variation by 62% over the current industry practice.

Keywords: Shaft Grinding, Sensor-Based Modeling, Nonlinear Stochastic Differential Equations, Surface Finish Control

(Proquest Information and Learning: ... denotes non-USASCII text omitted.)

Introduction

Nomenclature is given in the Appendix.

In recent years, industries have been striving to achieve increasingly high levels of quality and performance from manufacturing processes (Evans and Lindsay 1989; Deming 2000). Characterizing a process so that the real signal that conveys the process state is separated from noise is key to assuring that a process is not tampered based on noise. This idea has been thoroughly expounded by W. Edwards Deming, who illustrated this concept using red bead, funnel, and other experiments (Latzko and Saunders 1995).

Generally, the word "tamper" has a negative connotation, but the authors chose this word to reinforce the message of how the modeling technique presented in this research can compensate for the natural (harmful) instinct of machinists and process engineers to tamper with the process without proper information, which often leads to a deterioration of process quality. Model-based tampering (MET) provides a tool to channel these natural instincts into a positive outcome for the process.

Significance

Grinding has historically been considered a complex process to control because the dynamics involved are highly complicated. However, with surface finish specifications shrinking during the past 30 years, industries are experiencing an ever-growing need to effectively control grinding processes (Palanna and Bukkapatnam 2002). In particular, cylindrical grinding of shafts has numerous critical applications in the aerospace industry. Even though a shaft might cost only a small fraction of the entire product cost, it is a major determinant of the performance and durability of complex rotary systems. For example, air bearings are a key technology that gives a competitive edge to aerospace companies such as Honeywell. Air bearings allow machines to reach speeds of more than 100,000 rpm. Higher speeds of rotating assemblies provide a technological edge because they can make aircraft lighter for the same performance output.

A cylindrical shaft is the primary rotary component of an air bearing assembly. If the shaft surface is too smooth (R^sub a^ 10 [mu]in.*), then the load-carrying capacity of the air bearing is reduced. Hence, industry is striving to control surface finish to within a fine acceptable range using the current infrastructure formed by several conventional grinding machines. Industry also desires having consistently low variations of surface finish so that engineers can push the design envelope. There are, broadly, two options to achieve such a fine surface finish control: One approach is to develop designs that are robust to manufacturing process variations. This usually is a very expensive proposition, but many tools such as design for six sigma (DFSS), design for manufacture and assembly (DFMA), design for reliability (DFR), concurrent engineering, and so on, can help achieve this objective. The second option is to improve the output performance by addressing specific manufacturing issues. This research is an effort to push the envelope of this second option.

Research Summary and Objectives

This research builds on the growing evidence that suggests that the dynamics underlying many machining processes are nonlinear and stochastic. Also, a lot of work has been done in the statistical and quality engineering body of knowledge about the signal-to-noise relationship and ways to deal with it. Many quality gurus have preached that one has to understand a process using control charts and only adjust the process when one notices signals and not tamper for noise (Deming 2000; Wheeler 1995). However, theoretical knowledge for tampering is mostly restricted to linear systems with simple statistics. Developing tamper strategies in a nonlinear stochastic environment can benefit many real-world processes, such as aerospace shaft grinding. In MET, control inputs are synthesized based on nonlinear stochastic differential equation (n-SDE) modeling so one can tamper a process based on a signal and not on noise. An advantage of using n-SDE models is that even when the noise is simplified to be a Gaussian white noise process, the state variables may not be Gaussian but can exhibit high levels of bursts and intermittencies resembling signals from a real grinding process. These characteristics of n-SDEs make models compact and, yet, closer to reality. The key idea here is to develop real-time n-SDE models of the following form,

Manufacturing Industry

Model-based tampering for improved process performance-an application to grinding of shafts

Journal of Manufacturing Processes, 2003 by Palanna, Rajkumar, Bukkapatnam, Satish, Settles, F Stan

Find Research Guides for:

Find Featured Titles for: Health

Most Popular Articles in Business

Most Popular Publications in Business