Prioritisation assessment and robust predictive model for a comprehensive medical equipment maintenance using machine learning techniques / Aizat Hilmi Zamzam

Aizat Hilmi, Zamzam (2022) Prioritisation assessment and robust predictive model for a comprehensive medical equipment maintenance using machine learning techniques / Aizat Hilmi Zamzam. PhD thesis, Universiti Malaya.

	PDF (The Candidate's Agreement) Restricted to Repository staff only Download (217Kb)
	PDF (Thesis PhD) Download (3279Kb)

Abstract

Medical equipment reliability is critical to the quality of healthcare services. Nevertheless, maintaining the reliability of medical equipment in terms of availability, durability, safety, and economical is a challenging mission. A comprehensive and cost-effective medical equipment maintenance management covering preventive maintenance, corrective maintenance, and replacement plan is needed to achieve the three goals. The study aims to develop a comprehensive strategic maintenance management for sustaining the medical equipment reliability in a cost-effective way. Data such as maintenance history and inventory information on 13,350 units of medical equipment located in health clinics in fourteen states throughout Malaysia were used as samples. The datasets are established according to nineteen features and criteria for this study. The development of predictive models for objectives 1 and 2 of this study involves the application of seven supervised machine learning algorithms. The effectiveness of these models is assessed through eleven performance evaluation parameters. Classifiers that produce the best models are selected for the optimisation process. The optimal models are produced by adjusting the selected classifiers' hyperparameters to reduce the misclassification rate during the prediction process. The achievement of objective 1 demonstrates that the SVM, DT, and NN classifiers have developed optimised predictive models for first failure, failure to year ratio, and failure rectification action, respectively. Meanwhile, the development of predictive models to achieve research objective 2 involves two techniques for assessment of maintenance priorities, namely k-means and classification. The results of these prioritisation assessment techniques are then applied in the development of predictive models. A comparison of the effectiveness demonstrates that he combination of k-means and the NN classifier has developed optimised predictive models for all maintenance management activities at an accuracy rate of over 99.5%. The development of a comprehensive strategic maintenance management includes the elements of maintenance prioritisation and failure analysis to achieve objective 3. It involves the rationalisation of priorities for preventive maintenance, corrective maintenance, and replacement plan predictive models. Moreover, this rationalisation is combined with a first failure analysis prediction, which involves the adjustment of the frequency of planned preventive maintenance and maintenance costs. Integration between rationalisation and a combination of elements shows a reduction in preventive and corrective maintenance costs through the implementation of cost analysis. The results of the analysis found that a 61.4% cost-saving was obtained from the current maintenance costs. This cost-saving can cover 10% of the total estimated cost of procurement of obsolete equipment. This percentage is equivalent to 1,982, which is 40% of the total obsolete equipment proposed for replacement. The establishment of a comprehensive maintenance management through a combination of failure analysis and maintenance prioritisation predictive models can be a mechanism for the implementation of predictive maintenance. Furthermore, it also serves as a tool for clinical engineers in implementing more effective and efficient medical equipment maintenance management.

Actions (For repository staff only : Login required)