Test analysis & fault simulation of microfluidic systems

This work presents a design, simulation and test methodology for microfluidic systems, with particular focus on simulation for test. A Microfluidic Fault Simulator (MFS) has been created based around COMSOL which allows a fault-free system model to undergo fault injection and provide test measurements. A post MFS test analysis procedure is also described.A range of fault-free system simulations have been cross-validated to experimental work to gauge the accuracy of the fundamental simulation approach prior to further investigation and development of the simulation and test procedure.A generic mechanism, termed a fault block, has been developed to provide fault injection and a method of describing a low abstraction behavioural fault model within the system. This technique has allowed the creation of a fault library containing a range of different microfluidic fault conditions. Each of the fault models has been cross-validated to experimental conditions or published results to determine their accuracy.Two test methods, namely, impedance spectroscopy and Levich electro-chemical sensors have been investigated as general methods of microfluidic test, each of which has been shown to be sensitive to a multitude of fault. Each method has successfully been implemented within the simulation environment and each cross-validated by first-hand experimentation or published work.A test analysis procedure based around the Neyman-Pearson criterion has been developed to allow a probabilistic metric for each test applied for a given fault condition, providing a quantitive assessment of each test. These metrics are used to analyse the sensitivity of each test method, useful when determining which tests to employ in the final system. Furthermore, these probabilistic metrics may be combined to provide a fault coverage metric for the complete system.The complete MFS method has been applied to two system cases studies; a hydrodynamic “Y” channel and a flow cytometry system for prognosing head and neck cancer.Decision trees are trained based on the test measurement data and fault conditions as a means of classifying the systems fault condition state. The classification rules created by the decision trees may be displayed graphically or as a set of rules which can be loaded into test instrumentation. During the course of this research a high voltage power supply instrument has been developed to aid electro-osmotic experimentation and an impedance spectrometer to provide embedded test.

Name dc:type.qualificationname

PhD

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Hull

Year dc:date.issued

2010

Author dc:creator

• Myers, Thomas Oliver

Advisors dc:contributor.advisor

• Bell, Ian M.
• Wilkinson, Antony

• Engineering

Language dc:language

en

Identifier

oai:hull-repository.worktribe.com:4210117

OAI identifier oai:identifier

oai:hull-repository.worktribe.com:4210117