Modeling and Intelligent Control for Spatial Processes and Spatially Distributed Systems

Abstract

Dynamical systems are often characterized by their time-dependent evolution, named temporal dynamics. The space-dependent evolution of dynamical systems, named spatial dynamics, is another important domain of interest for many engineering applications. By studying both the spatial and temporal evolution, novel modeling and control applications may be developed for many industrial processes. One process of special interest is additive manufacturing, where a three-dimensional object is manufactured in a layer-wise fashion via a numerically controlled process. The material is printed over a spatial domain in each layer and subsequent layers are printed on top of each other. The spatial dynamics of the printing process over the layers is named the layer-to-layer spatial dynamics.

Additive manufacturing provides great flexibility in terms of material selection and design geometry for modern manufacturing applications, and has been hailed as a cornerstone technology for smart manufacturing, or Industry 4.0, applications in industry. However, due to the issues in reliability and repeatability, the applicability of additive manufacturing in industry has been limited. Layer-to-layer spatial dynamics represent the dynamics of the printed part. Through the layer-to-layer spatial dynamics, it is possible to represent the physical properties of the part such as dimensional properties of each layer in the form of a heightmap over a spatial domain. Thus, by considering the spatial dynamics, it is possible to develop models and controllers for the physical properties of a printed part. This dissertation develops control-oriented models to characterize the spatial dynamics and layer-to-layer closed-loop controllers to improve the performance of the printed parts in the layer-to-layer spatial domain.

In practice, additive manufacturing resources are often utilized as a fleet to improve the throughput and yield of a manufacturing system. An additive manufacturing fleet poses additional challenges in modeling, analysis, and control at a system-level. An additive manufacturing fleet is an instance of the more general class of spatially distributed systems, where the resources in the system (e.g., additive manufacturing machines, robots) are spatially distributed within the system. The goal is to efficiently model, analyze, and control spatially distributed systems by considering the system-level interactions of the resources. This dissertation develops a centralized system-level modeling and control framework for additive manufacturing fleets.

Many monitoring and control applications rely on the availability of run-time, up-to-date representations of the physical resources (e.g., the spatial state of a process, connectivity and availability of resources in a fleet). Purpose-driven digital representations of the physical resources, known as digital twins, provide up-to-date digital representations of resources in run-time for analysis and control. This dissertation develops an extensible digital twin framework for cyber-physical manufacturing systems. The proposed digital twin framework is demonstrated through experimental case studies on abnormality detection, cyber-security, and spatial monitoring for additive manufacturing processes. The results and the contributions presented in this dissertation improve the performance and reliability of additive manufacturing processes and fleets for industrial applications, which in turn enables next-generation manufacturing systems with enhanced control and analysis capabilities through intelligent controllers and digital twins.