A systematic feature extraction and selection framework for data-driven whole-building automated fault detection and diagnostics in commercial buildings

Fault detection and diagnosis schemes based on data-driven statistical modelling are highly dependent on an accurate and exhaustive feature extraction procedure to deliver a superior performance as a monitoring strategy. Otherwise conducted, a deficient feature extraction procedure leads to a monitoring structure widely deviated from normal operating conditions. If an operating state is not identified as it, an increment in false alarm rate would be evidenced whenever the process shifts towards that condition and the monitoring scheme triggers the abnormal condition warning. On the other hand, if two similar operating conditions could not be individualized i.e. to be identified as a single operating state, a lack of sensitivity for minor — yet typical — deviations would render a monitoring strategy with prominent misdetection rates. Although Multimode Operational Mapping requires the proper identification of a finite set of normal process states, it is a challenging task especially for large-scale systems. Its complexity derives from a broad universe of monitoring variables, highly interactuating process units integrated over very dynamic network systems, among others. This is the case of natural gas transmission infrastructure, as it deals with variable upstream production rates, diverse consumption trends from customers, internal processes constrains, merged in a stringent operating scheme. This paper proposes a novel strategy to address the identification and feature extraction of normal conditions on multimode operation systems. The proposed framework uses a segmentation approach based on operator’s knowledge, the Takagi-Sugeno-Kang fuzzy engine and k-means algorithm to characterize the normal operation states of the system. The results show an improvement in the performance of Principal Component Analysis during abnormal conditions detection, in addition an increase on the sensitivity of Hotelling and Q statistics.

Download Full-text

A Data Driven Frequency Based Feature Extraction and Classification Method for EMA Fault Detection and Isolation

Volume 2: Legged Locomotion; Mechatronic Systems; Mechatronics; Mechatronics for Aquatic Environments; MEMS Control; Model Predictive Control; Modeling and Model-Based Control of Advanced IC Engines; ◽

10.1115/dscc2012-movic2012-8749 ◽

2012 ◽

Cited By ~ 2

Author(s):

Anthony J. Chirico ◽

Jason R. Kolodziej ◽

Larry Hall

Keyword(s):

Feature Extraction ◽

Fault Detection ◽

Fault Detection And Isolation ◽

Data Driven ◽

Classification Method

Download Full-text

A Benchmarking Analysis of a Data-Driven Gas Turbine Diagnostic Approach

Volume 6: Ceramics; Controls, Diagnostics, and Instrumentation; Education; Manufacturing Materials and Metallurgy ◽

10.1115/gt2018-76887 ◽

2018 ◽

Author(s):

Igor Loboda ◽

Juan Luis Pérez-Ruiz ◽

Sergiy Yepifanov

Keyword(s):

Feature Extraction ◽

Fault Detection ◽

Gas Turbine ◽

Software Tool ◽

Aircraft Engine ◽

Data Driven ◽

Fault Classification ◽

Fault Identification ◽

Diagnostic Process ◽

Promising Tool

In an effort to better compare particular gas turbine diagnostic solutions and recommend the best solution, the software tool called Propulsion Diagnostic Method Evaluation Strategy (ProDiMES) has been developed. This benchmarking platform includes a simulator of the aircraft engine fleet with healthy and faulty engines. The platform presents a public approach, at which different investigators can verify and compare their algorithms for the diagnostic stages of feature extraction, fault detection, and fault identification. Using ProDiMES, some different diagnostic solutions have been compared so far. This study presents a new attempt to enhance a gas turbine diagnostic process. A data-driven algorithm that embraces the mentioned three diagnostic stages is verified on the basis of ProDiMES. At the feature extraction stage, this algorithm uses a polynomial model of an engine baseline to compute deviations of actual gas path measurements from the corresponding values of a healthy engine. At the fault detection and fault identification stages, a common classification for fault detection and fault identification is firstly constructed using deviation vectors (patterns). One of the three chosen pattern recognition techniques then performs both fault detection and fault identification as a common process. Numerous numerical experiments have been conducted to select the best configurations of the baseline model, a pertinent structure of the fault classification, and the best recognition technique. The experiments were accompanied by a computational precision analysis for each component of the proposed algorithm. The comparison of the final diagnostic ProDiMES metrics obtained under the selected optimal conditions with the metrics of other diagnostic solutions shows that the proposed algorithm is a promising tool for gas turbine monitoring systems.

Download Full-text

Systematic Feature Selection Process Applied in Short-Term Data-Driven Building Energy Forecasting Models: A Case Study of a Campus Building

Volume 3: Vibration in Mechanical Systems; Modeling and Validation; Dynamic Systems and Control Education; Vibrations and Control of Systems; Modeling and Estimation for Vehicle Safety and Integrity; Modeling and Control of IC Engines and Aftertreatment Systems; Unmanned Aerial Vehicles (UAVs) and Their Applications; Dynamics and Control of Renewable Energy Systems; Energy Harvesting; Control of Smart Buildings and Microgrids; Energy Systems ◽

10.1115/dscc2017-5073 ◽

2017 ◽

Author(s):

Liang Zhang ◽

Jin Wen ◽

Yimin Chen

Keyword(s):

Feature Selection ◽

Selection Process ◽

Building Energy ◽

Data Driven ◽

Forecasting Model ◽

Energy Forecasting ◽

Systematic Feature ◽

Selection Framework ◽

Term Data

An accurate building energy forecasting model is a key component for real-time and advanced control of building energy system and building-to-grid integration. With the fast deployment and advancement of building automation systems, data are collected by hundreds and sometimes thousands of sensors every few minutes in buildings, which provide great potential for data-driven building energy forecasting. To develop building energy forecasting models from a large number of potential inputs, feature selection is a critical procedure to ensure model accuracy and computation efficiency. Though the theory of feature selection is well developed in statistics and machine learning fields, it is not well studied in the application of building energy modeling. In this paper, a feature selection framework proposed in an earlier study is examined using a real campus building in Philadelphia. This feature selection framework combines domain knowledge and statistical methods and is developed for short-term data-driven building energy forecasting. In this case study, the feasibilities of using this feature selection framework in developing whole building energy forecasting model and chiller energy forecasting model are studied. Results show that, for both whole building and chiller energy forecasting applications, the model with systematic feature selection process presents better performance (in terms of cross validation error of forecasted output) than other models including that with conventional inputs and that uses only single feature selection technique.

Download Full-text

Gradient Ascent Optimization for Fault Detection in Electrical Power System Based on Wavelet Transformation

Current Signal Transduction Therapy ◽

10.2174/1574362414666190619092910 ◽

2019 ◽

Vol 14 ◽

Author(s):

Iyappan Murugesan ◽

Karpagam Sathish

Keyword(s):

Feature Extraction ◽

Fault Detection ◽

Power System ◽

Electrical Power ◽

Daubechies Wavelet ◽

Electrical Power System ◽

Gradient Ascent ◽

Neural Learning ◽

Weight Value ◽

Transmission And Distribution

: This paper presents electrical power system comprises many complex and interrelating elements that are susceptible to the disturbance or electrical fault. The faults in electrical power system transmission line (TL) are detected and classified. But, the existing techniques like artificial neural network (ANN) failed to improve the Fault Detection (FD) performance during transmission and distribution. In order to reduce the power loss rate (PLR), Daubechies Wavelet Transform based Gradient Ascent Deep Neural Learning (DWT-GADNL) Technique is introduced for FDin electrical power sub-station. DWT-GADNL Technique comprises three step, normalization, feature extraction and FD through optimization. Initially sample power TL signal is taken. After that in first step, min-max normalization process is carried out to estimate the various rated values of transmission lines. Then in second step, Daubechies Wavelet Transform (DWT) is employed for decomposition of normalized TLsignal to different components for feature extraction with higher accuracy. Finally in third step, Gradient Ascent Deep Neural Learning is an optimization process for detecting the local maximum (i.e., fault) from the extracted values with help of error function and weight value. When maximum error with low weight value is identified, the fault is detected with lesser time consumption. DWT-GADNL Technique is measured with PLR, feature extraction accuracy (FEA), and fault detection time (FDT). The simulation result shows that DWT-GADNL Technique is able to improve the performance of FEA and reduces FDT and PLR during the transmission and distribution when compared to state-of-the-art works.

Download Full-text