Development of deep learning model for vehicle engine health monitoring exploiting vulnerable points /
Md Abdur Rahim,
Development of deep learning model for vehicle engine health monitoring exploiting vulnerable points / Md Abdur Rahim - xiii, 164 pages : Illustration ; 30 cm.+ 1 CD ROM
College of Engineering
Thesis (Master of Science) -- Universiti Malaysia Pahang – 2022
Includes bibliographical reference
The vehicle is becoming more important in our daily lives due to its many advantages besides minimal physical effort. Nevertheless, certain unfavourable events cause them to become blurred and discourage users from using them, such as unexpected vehicle engine failure, high maintenance costs, costly spare parts, unavailability of replacement parts, excessive pollution output, and so on. These may happen due to unanticipated failures and a lack of routine engine inspections. In this case, ignoring minor irregularities of sensitive components may degrade valuable parts to significant damages. The available features for vehicular engine monitoring are mere comparisons to the contemporary artificial intelligence demands in the era of IR 4.0. Prior research suggested numerous monitoring systems; however, it neglected to consider the vehicle engine's vulnerable components and risk factors, which causes overall performance to deteriorate due to minor variations. Therefore, this research aims to develop a deep learning-based vehicular engine health monitoring model to utilise the vehicle engine's vulnerable components and related risk factors to assist stakeholders by providing real-time health information on a priority basis. Furthermore, to accomplish this objective, this research used the vulnerable components identification framework to identify the vulnerable components of the vehicular engine and then developed a decision model utilising the infrastructure vulnerability assessment model with the decision matrix. In this case, the vulnerability value of vulnerable items was used with their sensor reading to categorize the engine health condition as Good, Minor, Moderate, Critical. The model average ensemble, weighted model average ensemble, and stacked ensemble of deep learning techniques are employed to evaluate the performances of the developed vehicle engine health monitoring system decision model with the generated dataset and then validated by the machine learning approaches, i.e., support vector machine, decision tree, logistic regression, and k-nearest neighbors. After analysing the obtained performance, the Stacked Ensemble method of the deep learning algorithm reveals that the decision model outperforms with 80.9 per cent decision accuracy, among other parameters using 80 per cent training and 20 per cent testing datasets and categorise the engine health conditions perfectly. Wheres, the machine learning algorithms provided maximum decision accuracy is 78 per cent with the same data. The impact and significance of this research are indeed very substantial, i.e., people benefitted from lower maintenance costs, proactive action, improved engine longevity, reduced maintenance, and so on. Finally, this proposed model anticipates rapidly expanding the monitoring of the vehicle's overall health, railway system, alongside autonomous and electric vehicles.
THE0009421 (Local)
College of Engineering
Universities and colleges
Thesis
Development of deep learning model for vehicle engine health monitoring exploiting vulnerable points / Md Abdur Rahim - xiii, 164 pages : Illustration ; 30 cm.+ 1 CD ROM
College of Engineering
Thesis (Master of Science) -- Universiti Malaysia Pahang – 2022
Includes bibliographical reference
The vehicle is becoming more important in our daily lives due to its many advantages besides minimal physical effort. Nevertheless, certain unfavourable events cause them to become blurred and discourage users from using them, such as unexpected vehicle engine failure, high maintenance costs, costly spare parts, unavailability of replacement parts, excessive pollution output, and so on. These may happen due to unanticipated failures and a lack of routine engine inspections. In this case, ignoring minor irregularities of sensitive components may degrade valuable parts to significant damages. The available features for vehicular engine monitoring are mere comparisons to the contemporary artificial intelligence demands in the era of IR 4.0. Prior research suggested numerous monitoring systems; however, it neglected to consider the vehicle engine's vulnerable components and risk factors, which causes overall performance to deteriorate due to minor variations. Therefore, this research aims to develop a deep learning-based vehicular engine health monitoring model to utilise the vehicle engine's vulnerable components and related risk factors to assist stakeholders by providing real-time health information on a priority basis. Furthermore, to accomplish this objective, this research used the vulnerable components identification framework to identify the vulnerable components of the vehicular engine and then developed a decision model utilising the infrastructure vulnerability assessment model with the decision matrix. In this case, the vulnerability value of vulnerable items was used with their sensor reading to categorize the engine health condition as Good, Minor, Moderate, Critical. The model average ensemble, weighted model average ensemble, and stacked ensemble of deep learning techniques are employed to evaluate the performances of the developed vehicle engine health monitoring system decision model with the generated dataset and then validated by the machine learning approaches, i.e., support vector machine, decision tree, logistic regression, and k-nearest neighbors. After analysing the obtained performance, the Stacked Ensemble method of the deep learning algorithm reveals that the decision model outperforms with 80.9 per cent decision accuracy, among other parameters using 80 per cent training and 20 per cent testing datasets and categorise the engine health conditions perfectly. Wheres, the machine learning algorithms provided maximum decision accuracy is 78 per cent with the same data. The impact and significance of this research are indeed very substantial, i.e., people benefitted from lower maintenance costs, proactive action, improved engine longevity, reduced maintenance, and so on. Finally, this proposed model anticipates rapidly expanding the monitoring of the vehicle's overall health, railway system, alongside autonomous and electric vehicles.
THE0009421 (Local)
College of Engineering
Universities and colleges
Thesis