Methodology to screen vendors for predictive maintenance in the chemical industry. Issue 1 (3rd December 2021)

• Record Type: Journal Article
• Authors: Braun, Birgit; Dessauer, Michael; Henderson, Kaytlin; Peng, You; Seasholtz, Mary Beth
• Is Part Of: Journal of advanced manufacturing and processing. Volume 4:Issue 1(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12-03
• Subjects: chemical manufacturing -- equipment maintenance -- predictive maintenance -- vendor evaluation; Chemical engineering -- Periodicals; Manufacturing processes -- Technological innovations -- Periodicals; Manufacturing processes; Electronic journals; Periodicals; 660
• DOI: 10.1002/amp2.10109
• Languages: English
• ISSNs: 2637-403X

Abstract

Abstract: As an industry leader in digitalization and implementation of value‐added data‐driven methodologies, Dow is executing a structured evaluation of predictive maintenance (PdM) vendor offerings. PdM offers a tailored alternative to scheduled maintenance or run‐to‐failure operations, but the identification of suitable solutions offered by third parties is not trivial given the large number of offerings. This paper describes a methodology developed by Dow to deal with the challenge of efficiently screening many vendors with relevant PdM offerings. Prior to the evaluation process, scoring criteria for vendor performance must be identified. For Dow, these included the requirements (1) models can be created and deployed easily, (2) modeled asset health provides information for root causes, (3) the software operates in our preferred IT architecture, (4) confidential data cannot leave the premises, and (5) models have some transparency. The process involves four steps beginning with vendor identification, which explored existing relationships and landscape surveys. Following was the completion of a questionnaire by vendors about the offering. Upon positive completion, a dataset for two reflux pumps was provided for a first demonstration of the tool. The model performance was compared to internal modeling efforts, of which key results are shared in this paper. The last step involved an in‐depth evaluation including on‐site installation and online deployment of the PdM models, allowing scoring of all categories. What is presented herein is a framework that can be utilized for screening predictive maintenance modeling tools as well as many analytics applications arising in the age of Industry 4.0.