Title: Multi-scenario and multi-criteria approach to evaluate prediction techniques used to recognise failure patterns

Authors: Mahdi Mohammadi Tehrani; Mohammadsadegh Mobin; Yvan Beauregard; Michel Rioux; Jean Pierre Kenne

Addresses: Mechanical Engineering Department, ETS University (École de Technologie Supérieure), Montreal, Canada ' Industrial Engineering and System Engineering Department, Okland University, USA ' Mechanical Engineering Department, ETS University (École de Technologie Supérieure), Montreal, Canada ' Systems Engineering Department, ETS University (École de Technologie Supérieure), Montreal, Canada ' Mechanical Engineering Department, ETS University (École de Technologie Supérieure), Montreal, Canada

Abstract: Predicting the next failure of a specific component in a machine results in higher performance of the equipment in terms of its reliability. This study examines the five most applicable prediction methods used to predict HVAC filter blockage. Considering different scenarios that use design of experiment (DOE), a multiple criteria decision making (MCDM) tool, known as TOPSIS, is utilised to compare the performance of each predictive technique according to their performance in various contexts. The results show that the performance of the system is improved when the best predictive method has been selected not only on the premises of a purely mathematical approach, but also when the solid mathematical prediction is supplemented by the judgments of experts and edited data. In addition, the empirical data is retrieved and a new method to evaluate the prediction techniques based on the applications of DOE and MCDM tools is illustrated.

Keywords: prediction techniques; predictive maintenance; multiple criteria decision making; MCDM; design of experiment; DOE; heating ventilation; air conditioning system.

DOI: 10.1504/IJADS.2021.115994

International Journal of Applied Decision Sciences, 2021 Vol.14 No.4, pp.361 - 386

Received: 05 Oct 2019
Accepted: 27 Dec 2019
Published online: 06 Jul 2021 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article