Evaluation of gas turbine diagnostic techniques under variable fault conditions

Pérez-Ruiz, Juan; Loboda, Igor; Miró-Zárate, Luis Angel; Toledo-Velázquez, Miguel; Polupan, Georgiy

doi:10.1177/1687814017727471

Cited by 7 publications

(8 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As for SVM, it has a good generalization performance with a small number of samples, it is not limited to the computational memory as in the case of PNN, and the technique has a unique and global solution in comparison with ANNs that struggle with multiple local minima. Nevertheless, one disadvantage of SVM is the elevated training time compared to ANNs when the number of samples is increased [12].…”

Section: About Some Advantages and Disadvantages Of The Compared Fault Identification Techniquesmentioning

confidence: 99%

“…Combined with machine learning and pattern-recognition techniques, the GPA approach can be an efficient tool to diagnose complex and hidden engine faults. Many machine-learning techniques have been employed for gas turbine diagnostics, for example, support vector machines (SVM) [8], genetic algorithms [9], fuzzy logic [10] and neuro-fuzzy inference systems [11], multi-layer perceptron (MLP) [12], probabilistic neural network (PNN) [13], and extreme learning machines (ELM) [14,15].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Aircraft Engine Gas-Path Monitoring and Diagnostics Framework Based on a Hybrid Fault Recognition Approach

2021

Self Cite

View full text Add to dashboard Cite

Considering the importance of continually improving the algorithms in aircraft engine diagnostic systems, the present paper proposes and benchmarks a gas-path monitoring and diagnostics framework through the Propulsion Diagnostic Methodology Evaluation Strategy (ProDiMES) software developed by NASA. The algorithm uses fleet-average and individual engine baseline models to compute feature vectors that form a fault classification with healthy and faulty engine classes. Using this classification, a hybrid fault-recognition technique based on regularized extreme learning machines and sparse representation classification was trained and validated to perform both fault detection and fault identification as a common process. The performance of the system was analyzed along with the results of other diagnostic frameworks through four stages of comparison based on different conditions, such as operating regimes, testing data, and metrics (detection, classification, and detection latency). The first three stages were devoted to the independent algorithm development and self-evaluation, while the final stage was related to a blind test case evaluated by NASA. The comparative analysis at all stages shows that the proposed algorithm outperforms all other diagnostic solutions published so far. Considering the advantages and the results obtained, the framework is a promising tool for aircraft engine monitoring and diagnostic systems.

show abstract

Section: About Some Advantages and Disadvantages Of The Compared Fault Identification Techniquesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Aircraft Engine Gas-Path Monitoring and Diagnostics Framework Based on a Hybrid Fault Recognition Approach

2021

Self Cite

View full text Add to dashboard Cite

show abstract

“…To avoid the problem of exceeding a severity limit, the present investigation applies the Archimedean spiral boundary introduced in Reference. 26 From the constructed classes, a healthy class is formed by all incipient faults, that is, patterns not exceeding a healthy engine limit. In this manner, it is possible to address anomaly detection (monitoring) and fault identification (diagnostics) stages as an integrated process.…”

Section: Procedures For Comparing Diagnostic Spacesmentioning

confidence: 99%

“…This quantity has shown good recognition performance before. 26 For the third classification, the number of patterns is variable and depends on the healthy class boundary as mentioned before.…”

Section: Basic Fault Classificationsmentioning

confidence: 99%

A comparative study of data-driven and physics-based gas turbine fault recognition approaches

Pérez-Ruiz

Loboda

González-Castillo

et al. 2021

Proceedings of the Institution of Mechanical Engineers, Part O:

Self Cite

View full text Add to dashboard Cite

The present paper compares the fault recognition capabilities of two gas turbine diagnostic approaches: data-driven and physics-based (a.k.a. gas path analysis, GPA). The comparison takes into consideration two differences between the approaches, the type of diagnostic space and diagnostic decision rule. To that end, two stages are proposed. In the first one, a data-driven approach with an artificial neural network (ANN) that recognizes faults in the space of measurement deviations is compared with a hybrid GPA approach that employs the same type of ANN to recognize faults in the space of estimated fault parameter. Different case studies for both anomaly detection and fault identification are proposed to evaluate the diagnostic spaces. They are formed by varying the classification, type of diagnostic analysis, and deviation noise scheme. In the second stage, the original GPA is reconstructed replacing the ANN with a tolerance-based rule to make diagnostic decisions. Here, two aspects are under analysis: the comparison of GPA classification rules and whole approaches. The results reveal that for simple classifications both spaces are equally accurate for anomaly detection and fault identification. However, for complex scenarios, the data-driven approach provides on average slightly better results for fault identification. The use of a hybrid GPA with ANN for a full classification instead of an original GPA with tolerance-based rule causes an increase of 12.49% in recognition accuracy for fault identification and up to 54.39% for anomaly detection. As for the whole approach comparison, the application of a data-driven approach instead of the original GPA can lead to an improvement of 12.14% and 53.26% in recognition accuracy for fault identification and anomaly detection, respectively.

show abstract

“…Machine learning regression approaches are widely applied to provide reliable performance predictions of compressor components and anomaly detection for gas turbine combustors [8,[18][19][20]. Whereas artificial neural networks remain the most commonly used for diagnosing fault conditions [21], gas turbine performance is now being more rigorously monitored by simulation, optimization [22] and hybrid networks such as a combination between ANN and high dimensional model representation [23]. However, all of these methods involve correlations regressions, correlations and statistical relationships between input parameters and dependent variables to underpin their machine learning capabilities.…”

Section: Introductionmentioning

confidence: 99%

Combined cycle gas turbine power output prediction and data mining with optimized data matching algorithm

Wood

2020

SN Appl. Sci.

View full text Add to dashboard Cite

Electrical power output (PE) for a combined cycle gas turbine (CCGT) consisting of 9568 data records collected over a 6-year period is evaluated by the transparent open box (TOB) machine-learning method to provide accurate PE predictions and insight to prediction errors. The PE predictions derived by applying the TOB optimized data matching technique are more accurate than published predictions for the dataset from fifteen correlation-based, machine-learning algorithms. TOB achieves this high-accuracy using a tuning subset of < 150 (~ 1.5%) data records. Its accuracy is confirmed by testing the optimized solutions against all dataset records in 15 runs spread across five shuffled datasets. The dataset has a few extreme outliers associated with its four independent variables and these negatively impact the prediction accuracy of machine-learning methods. Through its transparency and forensic-like auditability of its calculations for individual data records, the TOB algorithm is able to mine the dataset to provide useful insight to the interactions of the outliers with other data records. This enables a filtered dataset (9533 records), excluding 35 carefully selected data records, to be customized to deliver much improved prediction accuracy (RMSE = 2.89 MW). Mining the dataset also reveals significant differences in prediction accuracy achieved for different sectors of the PE distribution. This insight identifies that prediction accuracy could be further improved by dividing the dataset into separately optimized subsets, three along its main PE trend plus a fourth, small subset consisting of the outliers. The TOB algorithm demonstrates its value as a machine-learning tool capable of generating accurate predictions and easily auditable data mining. It is well suited for CCGT efficiency and performance optimization.

show abstract

Evaluation of gas turbine diagnostic techniques under variable fault conditions

Cited by 7 publications

References 15 publications

Aircraft Engine Gas-Path Monitoring and Diagnostics Framework Based on a Hybrid Fault Recognition Approach

Aircraft Engine Gas-Path Monitoring and Diagnostics Framework Based on a Hybrid Fault Recognition Approach

A comparative study of data-driven and physics-based gas turbine fault recognition approaches

Combined cycle gas turbine power output prediction and data mining with optimized data matching algorithm

Contact Info

Product

Resources

About