Neil McDowell

Nonlinear PCA With the Local Approach for Diesel Engine Fault Detection and Diagnosis

This brief examines the application of nonlinear statistical process control to the detection and diagnosis of faults in automotive engines. In this statistical framework, the computed score variables may have a complicated nonparametric distribution function, which hampers statistical inference, notably for fault detection and diagnosis. This brief shows that introducing the statistical local approach into nonlinear statistical process control produces statistics that follow a normal distribution, thereby enabling a simple statistical inference for fault detection. Further, for fault diagnosis, this brief introduces a compensation scheme that approximates the fault condition signature. Experimental results from a Volkswagen 1.9-L turbo-charged diesel engine are included.

Semi-physical Neural Network Model in Detecting Engine Transient Faults using the Local Approach

Abstract This paper investigates detection of an air leak fault in the intake manifold subsystem of an automotive engine during transient operation. Previously, it was shown that integrating the local approach with an auto-associative neural network model of the engine, significantly increased the sensitivity of fault detection. However, the drawback then is that the computational load is naturally dependent on the network complexity. This paper proposes the use of the available physical models to pre-process the original signals prior to model building for fault detection. This not only extracts existing relationships among the variables, but also helps in reducing the number of variables to be modelled and the related model complexity. The benefits of this improvement are demonstrated by practical application to a modern spark ignition 1.8 litre Nissan petrol engine.

Deactivation of Oxidation Catalysts by Oil-Derived Sulphur

The most common mode of deactivation suffered by catalysts fitted to two-stroke engines has traditionally been thermal degradation, or even meltdown, of the washcoat and substrate. The high temperatures experienced by these catalysts are caused by excessively high concentrations of HC and CO in the exhaust gas which are, in turn, caused by a rich AFR and the loss of neat fuel to the exhaust during the scavenging period. The effects of catalyst poisoning due to additives in the oil is often regarded as a secondary, or even negligible, deactivating mechanism in two-stroke catalysts and has therefore received little attention. However, with the introduction of direct in-cylinder fuel injection to some larger versions of this engine, the quantities of HC escaping to the exhaust can be reduced to levels similar to those found on four-stroke gasoline engines. Under these conditions, the effects of poisoning are much more significant to catalyst durability, particularly for crankcase scavenged derivatives which allow considerable quantities of oil to escape into the exhaust in a neat, or partially burned form. In this paper the effects of oil-derived sulphur on catalyst performance are examined using specialised test apparatus. The oil used throughout the study was formulated specifically for a two-stroke engine fitted with direct in-cylinder fuel injection. The sulphur content of this oil was 0.21% by mass and particular attention was paid to the role of this element in the resulting deactivation. The catalyst was also designed for two-stroke applications and contained a high palladium loading of 300g/ft 3 (28g/l) to prolong the life of the catalyst. It was found that the sulphur caused permanent deactivation of the CO reaction and increased the light-off temperature by around 40°C after oiling for 60 hours. This deactivation was progressive and led to a reduction in surface area of the washcoat, particularly in the micropores of around 5A diameter. By using a validated catalyst model the change in surface area of the precious metal was estimated. It was found that the simulated palladium surface area had to be reduced by a factor of around 7.5 to produce the light-off temperature of the deactivated catalyst. Conversely, the light-off temperature of the C 3 H 6 reaction was barely affected by the deactivation.

Nonlinear PCA for Process Monitoring using the Local Approach

Abstract Process monitoring using nonlinear principal component analysis (NLPCA) is revisited, in particular that the score variables produced by the NLPCA model may not be statistically independent, nor follow a normal distribution. The Hotellings T 2 statistic is therefore unavailable for monitoring. This is addressed by introducing the statistical local approach into NLPCA based monitoring. The statistics from the local approach follow a normal distribution irrespective of the distribution of the score variables. This produces a Hotellings T 2 statistic with an underlying central χ 2 distribution as in linear PCA case. The associated benefits are exemplified using some examples.

Nonlinear PCA for Process Monitoring Using the Local Approach

Process monitoring using nonlinear principal component analysis (NLPCA) is revisited, in particular that the score variables produced by the NLPCA model may not be statistically independent, nor follow a normal distribution. The Hotelling’s T2 statistic is therefore unavailable for monitoring. This is addressed by introducing the statistical local approach into NLPCA based monitoring. The statistics from the local approach follow a normal distribution irrespective of the distribution of the score variables. This produces a Hotelling’s T2 statistic with an underlying central XX2 distribution as in linear PCA case. The associated benefits are exemplified using some examples.

Application of Auto-associative Neural Networks to Transient Fault Detection in an IC Engine

The tailpipe emissions from automotive engines have been subject to steadily reducing legislative limits. This reduction has been achieved through the addition of sub-systems to the basic four-stroke engine which thereby increases its complexity. To ensure the entire system functions correctly, each system and / or sub-systems needs to be continuously monitored for the presence of any faults or malfunctions. This is a requirement detailed within the On-Board Diagnostic (OBD) legislation. To date, a physical model approach has been adopted by the automotive industry for the monitoring requirement of OBD legislation. However, this approach has restrictions from the available knowledge base and computational load required. A neural network technique incorporating Multivariant Statistical Process Control (MSPC) has been proposed as an alternative method of building interrelationships between the measured variables and monitoring the correct operation of the engine. Building upon earlier work for steady state fault detection, this paper details the use of non-linear models based on an Auto-associate Neural Network (ANN) for fault detection under transient engine operation. The theory and use of the technique is shown in this paper with the application to the detection of air leaks within the inlet manifold system of a modern gasoline engine whilst operated on a pseudo-drive cycle.Copyright