Tema da Tese: ‘An Intelligent Decision Support System for the Analytical Laboratories of a Chemistry Industry’

Autor: António João Oliveira da Silva

Orientador: Paulo Alexandre Ribeiro Cortez

Data: 15/09/2022

Programa Doutoral: Tecnologias e Sistemas de Informação

Abstract

The Industry 4.0 represents the fourth industrial revolution and involves an implementation using several Information Technologies to support production, as well as a real-time monitoring of industrial processes. The topic of Business Analytics is particularly valuable in this context, since it results from a combination of Business Intelligence with Optimization and Forecasting. The objective is to obtain data-driven knowledge that can be useful to help decision making on production processes. For example, Business Analytics can be used to analyze historical data to help detect and predict problems or failures in production. Another interesting possibility is the prediction of demand orders, which can help in the process of stock management.

This PhD work is carried out within the scope of a Research & Development (R&D) project. The main objective is the research and implementation of an Intelligent Decision Support System (IDSS) that uses Business Analytics techniques (Descriptive, Prescriptive and Predictive), integrated within the Industry 4.0 concept and applied to Analytical Laboratories of Chemical companies. Initially, the analyzed company needs were elicitated, and subsequently several IDSS modules were developed aiming to solve the Chemical company goals. The first studied module was the prediction of arrival of samples at the Analytical Laboratories by using an Auto Machine Learning (AutoML) tool. Next, a module was developed for predicting the consumption of materials in the laboratories. This module included three different forecasting approaches that were compared, one with an AutoML, another using the ARIMA methodology and the last based on a deep learning algorithm (Long Short-Term Memory). The best forecasting results were achieved by the AutoML approach. Finally, a module was developed with prescriptive methods to allocate the instruments to the analyses to be performed as well as the development of the friendly user Dashboards for the designed IDSS. The full IDSS system was evaluated by using questionnaires and open interviews with the Analytical Laboratory managers. Overall, a positive feedback was obtained.