IoT Based Real-Time predictive maintenance system for medical equipment using an integrated advanced analytics (IAA) model

Abstract

Today medical technologies are improving, chronic diseases are also increasing day to day. The healthcare sector, like other industries, is having a high demand to cope with those influencing factors to satisfy the expectations of customers. Again, the success of all industries relates to attaining satisfaction with clients with a high level of services and productivity. Among the success main factors, the maintenance of equipment plays a significant impact to the overall production and effective service delivery. To date, the Rwandan hospitals that always have a long queue of patients waiting for service, perform a repair after failure as common maintenance practice that may involve unplanned resources, cost, time, and completely or partially interrupt the remaining hospital activities. Hence, hospitals need to be well equipped with equipment in good conditions ensured through the maintenance performed using different methods and technologies. Yet, regardless of effort put into ordinary maintenances, unplanned downtime due to discontinued monitoring of health status, deterioration or misuse of equipment may also happen and may take long to be corrected. Aiming to reduce unplanned crucial equipment downtime associated time and cost, this research brings up an IoT based real-time predictive maintenance (IoT based PdM) system using an Integrated Advanced Analytic (IAA) model to save maintenance time, improve maintenance accuracy and reduce the related cost in referral hospitals of Rwanda by performing planned necessary preventive maintenance and working out from unplanned downtime. The proposed IoT based PdM comprises of three main parts. The first part proposes the Predictive Maintenance (PdM) structure powered by Internet of Things (IoT) to be adopted by hospitals to predict early failure before it happens for mechanical equipment used in hospitals. Because prediction relies on data, the structure design consists of a simplest developed IoT device prototype with the purpose of collecting real-time data for predictive model construction, equipment health status classification and later to host adopted predictive model. The real-time sequential data in the form of time series have been collected from selected equipment’s components in King Faisal Hospital Rwanda and then used to build a proposed predictive time series model to be employed in proposed structure. Since, the data from different components are independent from each other at some instant and that each part may push down the whole equipment independently, the Long Short-Term Memory (LSTM) Neural Network model was used to learn univariate data from different components and performed with an accuracy of 90% and 96% to different two selected components. Considering that LSTM did not perform well on independent multi-variates time series data from different components of the equipment, and that on side of maintenance activities priority for complex equipment: maintainers are manually deciding on crucial actions to be performed prior to others. We claim that the integration of knowledge based expert system might combine different independent condition parameters to come up with abridged univariate data to be fed to the sequential model for an effective predictive maintenance analytics through accurate maintenance priorities. As results, the second part of this research proposed the maintenance activities xx prioritization using Fuzzy expert system for small and medium sized hospitals. It considers the expertise of maintainers in faults detection and classification through the various monitoring of the physical condition parameters from equipment's components. Parameters' condition severity in respect of the total equipment downtime are considered to predict maintenance activities’ priority. Reflecting to the need of precise and quick prediction as well as real-time monitoring on the Edge, the on device TinyModels to provide the real-time sights on fault roots and remaining useful life of the equipment was proposed in third part of this research work. Considering the labeled data as maintenance priorities by fuzzy logic based on the maintainer’s expertise, this part used the labeled data set to compute the actual remaining useful life and then presents the ability of the two realtime tiny predictive analytics models: tiny long short-term memory (TinyLSTM) and sequential dense neural network (DNN) from Edge Impulse models. Both models (TinyModels) are used to predict the remaining useful life of the equipment by considering the status of its different components. The equipment degradation insights were assessed through the real-time data gathered from operating equipment. The predictive analytic models were developed and performed well, with an evaluation loss of 0.01 and 0.11, respectively, for the LSTM and DNN model from Edge Impulse. Both models were converted into TinyModels for on-device deployment. Unseen data were used to simulate the deployment of both TinyModels. Conferring to the evaluation and deployment results, both TinyLSTM and TinyModel from Edge Impulse are powerful in real-time predictive maintenance, but the model from Edge Impulse is much easier in terms of development, conversion to Tiny-version, and deployment. To conclude this work, adding to the effectiveness of this IoT based Real-Time PdM with an Integrated Advanced Analytic (IAA) model into hospitals, it may be adopted by any industry interested in real-time monitoring based on performance and conditional data from their equipment. And, since the data affect the performance of the model, maintenance decision making and assumption of Remaining Useful Life (RUL), the TinyModel shall be updated and customized for project implementation.