search icon
email icon
rrss gif icons
twitter icon
linkedin icon youtube icon

Enhanced Model Reference Adaptive Controllers for Fault-Tolerant Controls in Industrial Applications

Jorge Rodríguez Guerra


  • DIRECTORES: Carlos Calleja Elcoro y Ana María Macarulla Arenaza
  • UNIVERSIDAD: Universidad de Deusto


Industrial systems, similarly to natural ecosystems, are continually fluctuating. Both environments are molded by internal and external factors, conditioning the inhabiting species lifestyle. The manufacturing machines and the living beings populating factories and landscapes, respectively, suffer continuous transformations to accommodate their features to the locations they occupy. This constant adaptation process has promoted the emergence of characteristics in the individuals that strengthen their survivability.

These advantages increase populations lifespan, especially the mechanisms diagnosing dangerous situations to draw up a contingency plan. Biological species present several techniques to ensure that their individuals survive despite the emergence of a dire circumstance. Manufacturing machines have imitated these organic procedures to maintain the production stable despite the emerge of malfunctioning sources on their components.

Fault-Tolerant Control techniques introduce a revolutionary approach to detect, measure and isolate faults on a first instance during a phase called Fault Detection and Isolation, and, to reconfigure the control algorithm in a second instance during a phase called Control Redesign. Both phases represent the mechanism previously mentioned, as this technique analyses the drawback source and updates the system performance accordingly to bypass the injury. This reconfiguration procedure encapsulates an Adaptive Control algorithm that modifies the controller response attending to the new dynamics or faults in the manufacturing machine.

Between the multiple configurations for these algorithms, Model Reference Adaptive Controllers offers the most suitable platform. However, despite their benefits, industrial systems require higher levels of flexibility, robustness and upgradability hardly achievable by conventional MRAC architectures. This research project has been focused on improving the current structure, enhancing its three major components:

  1. Model Based Technology: Biological species learn from their parents the abilities to survive in the environment. Similarly, manufacturing machines follow the signals from the reference model to reduce the tracking error. This accommodation process substitutes the transfer function describing the closed-loop system response with a Digital-Twin connected to a replica of the original control algorithm. This enhanced model increases MRACs flexibility, as they adapt to a more extensive number of fault scenarios.
  2. Adaptive Control Algorithms: Biological species present healing capabilities, recovering their bodies from injuries and wounds after short-term periods. Similarly, manufacturing machines controllers adjust their performance to the fault dynamics in the plant through the adaptive gains. Across each iteration in the production cycle, these gains accommodate the controller responses to minimise the tracking error, recovering the system optimal performance. Besides, an additional loop ensures the reference model is adapted to the faulty mechanical limits. This enhanced algorithm increases MRACs robustness, as it maintains the system stable without interrupting the production independently of the fault harm grade.
  3. Advance Manufacturing Techniques: Biological species inhabit ecosystems in communities, distributing their roles accordingly to the individuals' particular features. Similarly, manufacturing machines are affected by multiple faults with heterogeneous sources. Instead of designing a unique adaptive gain to surpass each incident, a Bank of Controllers structure switches between these parameters to the optimal situation. This enhanced platform increases MRACs upgradability, as the system increases the number of fault scenarios surpassed optimally introducing additional adaptive gains into the Bank of Controllers.

Through these enhances, MRACs present an architecture compatible with industrial systems. This novel methodology has been validated against a Hydraulic-Press explicitly designed for this research project. The experiments carried out to corroborate the investigation confirm the initial statement, in the presence of faults, the control algorithm detects, measure and isolate its source, reconfiguring the controller afterwards to surpass it without spreading its damage.

close overlay