Lean burn combustion monitoring strategy based on data modelling

Fu, Ruowei; Harrison, Robert F.; King, Stephen P.; Mills, Andrew R.

doi:10.1109/aero.2016.7500876

Cited by 4 publications

(2 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This method does not require historical fault data and is more interpretable. In contrast to Tarassenko et al who proposed training a neural network (NN) predictor of fault-free temperature reading from a set of measured variables (Tarassenko et al , 2000), we avoid the complexity of dealing with the convergence issues of such nonlinear models with no loss of performance (Fu et al , 2016). This is applicable to an in-service environment where the trends in behaviour are key but may be less suited to a test environment – as discussed in Section 4.…”

Section: Overview Of Solution Spacementioning

confidence: 99%

See 1 more Smart Citation

Sensing for aerospace combustor health monitoring

Mills

Kadirkamanathan

2020

AEAT

View full text Add to dashboard Cite

Purpose This paper proposes new methods of fault detection for fuel systems in order to improve system availability. Novel fault systems are required for environmentally friendly lean burn combustion, but can carry high risk failure modes particularly through their control valves. The purpose of the developed technology is the rapid detection of these failure modes, such as valve sticking or impending sticking, and therefore to reduce this risk. However, sensing valve state is challenging due to hot environmental temperatures, which results in a low reliability for conventional position sensing. Design/methodology/approach Starting with the business needs elicited from stakeholders, a quality functional deployment process is performed to derive sensing system requirements. The process acknowledges the difference between test-bed and in-service aerospace needs through weightings on requirements and maps these customer requirements to systems performance metrics. The design of the system must therefore optimise the sensor suite, on- and off-board signal processing and acquisition strategy. Findings Against this systems engineering process, two sensing strategies are outlined which illustrate the span of solutions, from conventional gas path sensing with advanced signal processing to novel non-invasive sensing concepts. While conventional sensing may be feasible within a test cell, the constraints of aerospace in-service operation may necessitate more novel alternatives. Acoustic emission (detecting very high frequency surface vibration waves) sensing technology is evaluated to provide a non-invasive, remote and high temperature tolerant solution. Through this comparison, the considerations for the end-to-end system design are highlighted to be critical to sensor deployment success in-service. Practical implications The paper provides insight into different means of addressing the important problem of monitoring faults in combustor systems in gas turbines. By casting of the complex design problem within a systems engineering framework, the outline of a toolset for solution evaluation is provided. Originality/value The paper provides three areas of significant contributions: a diversity of methods to diagnosing fuel system malfunctions by measuring changes fuel flow distributions, through novel means, and the combustor exit temperature profiles (cause and effect); the use of analytical methods to support the selection (types and quantities) and placement of sensors to ensure adequate state awareness while minimising their impact on the engine system cost and weight; and an end-to-end data processing approach to provide optimised information for the engine maintainers allowing informed decision-making.

show abstract

Section: Overview Of Solution Spacementioning

confidence: 99%

“…In comparison to the raw sensor output(Figure 8), considerable robustness has been added through analysisas shown by the margin of separation. Prediction errors in normal and faulty condition(Fu et al 2016)…”

mentioning

confidence: 99%