Thesis defense of Daniel Reguera Bakhache
Thesis defense of Daniel Reguera Bakhache
Thesis defense of Daniel Reguera Bakhache
Title of the thesis: "Metodologías Data-Driven para optimizar la interacción persona-máquina en escenarios industriales". Obtained the SOBRESALIENTE qualification.
- Title of the thesis: "Metodologías Data-Driven para optimizar la interacción persona-máquina en escenarios industriales".
- Court:
- President: Ignacio Díaz Blanco (Universidad de Oviedo)
- Vocal: Gorka Epelde Unanue (Vicomtech)
- Vocal: Asier Aztiria Goenaga (Mármoles Aztiria)
- Vocal: Urko Zurutuza Ortega (Mondragon Unibertsitatea)
- Secretary: Carlos Cernuda García (Mondragon Unibertsitatea)
Abstract
The rapid expansion of new technologies into the industrial sector has transformed traditional industrial processes into intelligent operations, in which operator-machine interaction is becoming increasingly complex. In this new scenario, artificial intelligence is gaining traction as a powerful tool to promote more efficient and effective interactions between operators and machines, by means of systems capable of adapting to the context. However, present day industrial adaptive systems do not incorporate knowledge extracted from operator-machine interaction, process status information, or HMI specifications in the adaptation process; three factors critical to achieving personalized interactions.
In this thesis we focus on improving the interaction between humans and machines in an industrial context. To this end, a series of data-driven methodologies for industrial HMIs were developed to obtain a set of intelligent temporal adaptation rules. By using machine learning techniques the methodologies: i) infer the different interaction patterns utilized in an industrial process, ii) identify the time intervals in which the proposed rules must be activated and iii) propose a set of actions to be performed in the HMI, while respecting the characteristics and integrity of the HMI.
The developed methodologies were validated in three industrial scenarios, demonstrating that by taking as input data: i) the operator-machine interaction data, and ii) industrial HMI information, it is possible to infer a set of temporal set of temporal adaptation rules. These rules, once implemented in the tested scenarios, achieved a reduction in the interaction time and a decrease in the number of events required to perform the sequence, thereby improving operator performance when carrying out supervisión and control tasks.