Anomaly detection refers to detecting data points, events, or behaviour that do not comply with expected or normal behaviour. For example, a typical problem related to anomaly detection on an industrial level is having little labelled data and a few run-to-failure examples, making it challenging to develop reliable and accurate prognostics and health management systems for fault detection and identification. Certain machine learning approaches for anomaly detection require normal data to train, which reduces the need for historical data with fault labels, where the main task is to differentiate between normal and anomalous behaviour. Several reconstruction-based deep learning approaches are explored in this work and compared towards detecting anomalies in air compressors. Anomalies in such systems are not point-anomalies, but instead, an increasing deviation from the normal condition as the system components start to degrade. In this paper, a descriptive range of the deviation based on the reconstruction-based techniques is proposed. Most anomaly detection approaches are considered black box models, predicting whether an event should be considered an anomaly or not. This paper proposes a method for increasing the transparency and explainability of reconstruction-based anomaly detection to indicate which parts of a system contribute to the deviation from expected behaviour. The results show that the proposed methods detect abnormal behaviour in air compressors accurately and reliably and indicate why it deviates. The proposed approach is capable of detecting faults without the need for historical examples of similar faults. The proposed method for explainable anomaly detection is crucial to any prognostics and health management (PHM) system due to its purpose of detecting deviations and identifying causes.
Prognostics is an engineering discipline focused on predicting the time at which a system or a component will no longer perform its intended function. Due to the requirements of system safety and reliability, the correct diagnosis or prognosis of abnormal condition plays a vital role in the maintenance of industrial systems. It is expected that new requirements in regard to autonomous ships will push suppliers of maritime equipment to provide more insight into the conditions of their systems. One of the stated challenges with these systems is having enough run-to-failure examples to build accurate-enough prognostic models. Due to the scarcity of enough reliable data, transfer learning is established as a successful approach to improve and reduce the need to labelled examples. Transfer learning has shown excellent capabilities in image classification problems. Little work has been done to explore and exploit the use of transfer learning in prognostics. In this paper, various deep learning models are used to predict the remaining useful life (RUL) of air compressors. Here, transfer learning is applied by building a separate prognostics model trained on turbofan engines. It has been found that several of the explored transfer learning architectures were able to improve the predictions on air compressors. The research results suggest transfer learning as a promising research field towards more accurate and reliable prognostics.
Computer science graduates face a massive gap between industry-relevant skills and those learned at school. Industry practitioners often counter a huge challenge when moving from academics to industry, requiring a completely different set of skills and knowledge. It is essential to fill the gap between the industry's required skills and those taught at varsities. In this study, we leverage deep learning and big data to propose a framework that maps the required skills with those acquired by computing graduates. Based on the mapping, we recommend enhancing the computing curriculum to match the industry-relevant skills. Our proposed framework consists of four layers: data, embedding, mapping, and a curriculum enhancement layer. Based on the recommendations from the mapping module, we made revisions and modifications to the computing curricula. Finally, we perform a case study of the Norwegian IT jobs market, where we make recommendations for data science and software engineering-related jobs. We argue that by using our proposed methodology and analysis, a significant enhancement in the computing curriculum is possible to help increase employability, student satisfaction, and smart decision-making.
Industry 4.0 has revolutionized the use of physical and digital systems while playing a vital role in the digitalization of maintenance plans for physical assets in an optimal way. Road network conditions and timely maintenance plans are essential in the predictive maintenance (PdM) of a road. We developed a PdM-based approach that uses pre-trained deep learning models to recognize and detect the road crack types effectively and efficiently. We, in this work, explore the use of deep neural networks to classify roads based on the amount of deterioration. This is done by training the network to identify various types of cracks, corrugation, upheaval, potholes, and other types of road damage. Based on the amount and severity of the damage, we can determine the degradation percentage and have a PdM framework where we can identify the intensity of damage occurrence and, thus, prioritize the maintenance decisions. The inspection authorities and stakeholders can make maintenance decisions for certain types of damages using our deep learning-based road predictive maintenance framework. We evaluated our approach using precision, recall, F1-score, intersection-over-union, structural similarity index, and mean average precision measures, and found that our proposed framework achieved significant performance.
10 April 2024 in CAAI Transactions on Intelligence Technology
Transportation systems primarily depend on vehicular flow on roads. Developed countries have shifted towards automated signal control, which manages and updates signal synchronisation automatically. In contrast, traffic in underdeveloped countries is mainly governed by manual traffic light systems. These existing manual systems lead to numerous issues, wasting substantial resources such as time, energy, and fuel, as they cannot make real‐time decisions. In this work, we propose an algorithm to determine traffic signal durations based on real‐time vehicle density, obtained from live closed circuit television camera feeds adjacent to traffic signals. The algorithm automates the traffic light system, making decisions based on vehicle density and employing Faster R‐CNN for vehicle detection. Additionally, we have created a local dataset from live streams of Punjab Safe City cameras in collaboration with the local police authority. The proposed algorithm achieves a class accuracy of 96.6% and a vehicle detection accuracy of 95.7%. Across both day and night modes, our proposed method maintains an average precision, recall, F 1 score, and vehicle detection accuracy of 0.94, 0.98, 0.96 and 0.95, respectively. Our proposed work surpasses all evaluation metrics compared to state‐of‐the‐art methodologies.