PhD Thesis

Advanced Energy Management/Control Strategies for Smart Manufacturing Systems

Work default illustration

Information

  • Started: 01/04/2017
  • Finished: 27/03/2020

Description

This thesis is devoted to the study of optimisation-based control techniques for the design of management strategies that contribute to improving the energy efficiency of smart manufactur- ing systems. Currently, the manufacturing industry is suffering a transformation towards smart, flexible, and energy-efficient manufacturing systems promoted by the advances in sensing tech- nologies, data management techniques, and communication and connectivity tools. This trans- formation requires the manufacturing systems will be modularised and reconfigurable to be able of adapting to changes in productions programs, piece designs, and to the time-varying piece demands, while keeping an energy-efficient and sustainable operation. Therefore, to achieve a smarter manufacturing industry, suitable control systems should be designed to satisfy the re- quirements of this transformation, as well as to contribute to minimise the energy consumption and maximise the plant profit. In this regard, optimisation-based controllers and non-centralised control architectures could be suitable for the design of control systems that allow minimising the total energy consumption of such systems while remaining their productivity and taking into account the operational conditions and the factors that affect such systems. Thus, using these advanced control techniques, the control systems can be suitably updated to include the new information about the changes in the operation of manufacturing systems as well as the energy-market information to minimise the total energy cost during the plant operation.

First, this dissertation presents and discusses the strategies currently implemented by the manufacturing industry to improve its energy efficiency. Based on this review, the research gaps in this field are identified and it is discussed how optimisation-based control techniques can contribute to face the challenges of the new era of the manufacturing industry (Industry 4.0). Thus, according to the literature review, the manufacturing industry is classified by levels, i.e., machine, process line, and plant levels, for the design of optimisation-based controllers. Moreover, with the aim to design control strategies that do not affect plant productivity, i.e., the number of processed pieces in a fixed period, the compositional elements of manufacturing systems are also classified according to the operations performed. Then, based on the latter classification, control strategies are proposed to minimise either the total energy consumption of manufacturing systems or the energy costs related to the operation of such systems.
At both machine and process line levels, control strategies are designed based on model predictive control approach to minimise the total energy consumption of manufacturing systems. The underlying idea behind the proposed control strategies consists of managing independently those devices (or systems) that are not directly involved in the machining operations. In this regard, energy consumption models are required to predict the total energy consumption profile of manufacturing systems and, based on this prediction, to select the activation/deactivation instants for the manipulated devices that minimise the energy consumption and guarantee the proper operation of such systems. On the other hand, due to at the process line level the size and complexity of manufacturing systems increases, a control strategy based on two control modes is proposed to reduce the computational burden of such controllers and, they can be implemented in real time. Thus, due to the manufacturing systems exhibit periodic behaviour, an algorithm to detect the periodicity of such systems is proposed and, then, to switch from a control mode based on online optimisation to an autonomous control mode without solving in real time an optimisation problem.

On the other hand, due to the need for flexible and reconfigurable manufacturing systems, non-centralised control strategies are proposed at higher industrial levels to minimise their en- ergy consumption. In this regard, both cooperative and non-cooperative local controllers are designed considering a fixed system partitioning and using alternative direction methods of mul- tipliers to solve the optimisations problems in a distributed fashion. Besides, due to the nature of the proposed control objectives, which are focused on minimising the energy consumption of manufacturing systems, a way to define the consensus stage among the local controllers with coupled dynamics is proposed. Then, the proposed algorithms are extended to the plant level using economic cost functions, and the closed-loop performance and the computational burden for both centralised and non-centralised control architectures are compared.

Finally, at the plant level, control strategies are designed based on the economic model predictive control approach and oriented to maximise the plant profit and minimise the opera- tional costs related to the plant operation. Thus, at this level, control objectives are focused on determining the economic-optimal production programming of the plant that the control strate- gies at lower levels should follow. In this regard, the production programming of the plant is determined taking into account the pieces demand, the energy consumption of manufacturing systems, and the current energy market and their fluctuations. All control strategies proposed in this thesis are tested in simulation and considering different scenarios, which were designed based on the real operation of an automotive part manufacturing plant.

The work is under the scope of the following projects:

  • IKERCON: Control avanzado de procesos complejos de manufactura (web)
  • DEOCS: Monitorización, diagnostico y control tolerante a fallos de sistemas ciberfísicos con métodos basados en datos (web)