Search

Machine learning (ML) techniques are widely adopted in manufacturing systems for discovering valuable patterns in shopfloor data. In this regard, machine learning models learn patterns in data to optimize process parameters, forecast equipment deterioration, and plan maintenance strategies among other uses. Thus, this article presents the dataset collected from an assembly line known as the FASTory assembly line. This data contains more than 4,000 data samples of conveyor belt motor driver's power consumption. The FASTory assembly line is equipped with web-based industrial controllers and smart 3-phase energy monitoring modules as an expansion to these controllers. For data collection, an application was developed in a timely manner. The application receives a new data sample as JavaScript Object Notation (JSON) every second. Afterwards, the application extracts the energy data for the relevant phase and persists it in a MySQL database for the purpose of processing at a later stage. This data is collected for two separate cases: static case and dynamic case. In the static case, the power consumption data is collected under different loads and belt tension values. This data is used by a prognostic model (Artificial Neural Network (ANN)) to learn the conveyor belt motor driver's power consumption pattern under different belt tension values and load conditions. The data collected during the dynamic case is used to investigate how the belt tension affects the movement of the pallet between conveyor zones. The knowledge obtained from the power consumption data of the conveyor belt motor driver is used to forecast the gradual behavioural deterioration of the conveyor belts used for the transportation of pallets between processing workstations of discrete manufacturing systems.

The vast adoption of machine learning techniques in developing smart solutions increases the need of training and testing data. This data can be either collected from physical systems or created using simulation tools. In this regard, this paper presents a set of data collected using a digital twin known as the FASTory Simulator. The data contains more than 100 K events which are collected during a simulated assembly process. The FASTory simulator is a replica of a real assembly line with web-based industrial controllers. The data have been collected using specific-developed orchestrator. During the simulated process, the orchestrator was able to record all the events that occurred in the system. The provided data contains raw JavaScript Object Notation (JSON) formatted data and filtered Comma Separated Values (CSV) formatted data. This data can be exploited in machine learning for modelling the behaviour of the production systems or as testing data for optimization solution for the production system. Finally, this data has been utilized in a research for comparing different data analysis approaches including Knowledge-based systems and data-based systems.