In this study, we propose a methodology for the identification of potential fault occurrences of railway point-operating machines, using unlabeled signal sensor data. Data supplied by Network Rail, UK, is processed using a fast Fourier transform signal processing approach, coupled with the mean and max current levels to identify potential faults in point-operating machines. The method developed can dynamically adapt to the behavioral characteristics of individual point-operating machines, thereby providing bespoke condition monitoring capabilities in situ and in real time. The work described in this paper is not unique to railway point-operating machines, rather the data pre-processing and methodology is readily applicable to any motorized device fitted with current sensing capabilities. The novelty of our approach is that it does not require pre-labelled data with historical fault occurrences and therefore closely resembles problems of the real world, with application for smart city infrastructure. Lastly, we demonstrate the problems faced with handling such data and the capability of our methodology to dynamically adapt to diverse data presentations.

Diesel engine failures are one reason for delays and breakdowns on the UK rail network, resulting in significant fines and related financial penalties for a train operating company. Preventing such failures is the ultimate goal, but forecasting or predicting future failures before they occur would be highly desirable. In this study, we take real world Diesel Multiple Unit sensor data, recorded in the form of event data, and repurpose it for the remote condition monitoring of critical diesel engine operations. A methodology based on windowing of data is proposed that demonstrates the effective processing of event data for predictive modelling. This study specifically looks at predicting engine failures, and through this methodology, models trained on the processed data resulted in accuracies of 88%. Explainable AI methods are then utilised to provide feature importance explanations for the model’s performance. This information helps the end user understand specifically which sensor data from the larger dataset is most relevant for predicting engine failures. The work presented is useful to the railway industry, but more specifically to train operator companies who ideally want to foresee failures before they occur to avoid significant financial costs. The methodology proposed is applicable for the predictive maintenance of many systems, not just railway diesel engines.

Background: Increasing diabetes prevalence is a major public health concern. In this study we ask whether linked open data can be used to predict prescription volumes of drugs used in the treatment of diabetes across small geographies of England. Methods: We propose and demonstrate a methodology of utilising publicly available open data to infer the geo-spatial distribution of prescribed drugs for diabetes, at the lower layer super output area level. Multiple datasets are acquired, processed, and linked together, enabling a more in-depth analysis. Combining these linked datasets with published deprivation factors of geographies across England, we build highly predictive regression models. Results: Regression models were trained and are capable of accurately predicting diabetes prescribing volumes based on deprivation indicators of various geographies across England. Models built with data covering the city of Bradford, England, produced a predicted against actual correlation value of R² = 0.672 using multiple linear regression and 0.775 using Least Absolute Shrinkage and Selection Operator (LASSO). Median age and air quality factors proved to be significant markers for diabetes prescribing. Conclusions: The results of this study suggest our methodology is robust and accurate. Such predictive models are useful to health authorities in light of increasing costs and increasing prevalence of diabetes. While using publicly available open data negates any issues of data privacy.