UC Berkeley

Industrial systems enable modern life. They benefit tremendously by adapting digi-

tal communication technologies and leveraging automation algorithms and data availability.

Their importance to basic human needs such as electricity, heating, food, transportation,

clothing, and more also means that their constant availability and reliability is impera-

tive in modern societies. Plant monitoring strategies that can collect information and use

it to analyze and understand plant behavior is a key technology for optimizing industrial

systems. Enabling plant monitoring insights to be communicated to human operators is

essential to ensure the information can be used. While data-based methods continue to

find new applications, model-based methods that incorporate unique plant characteristics

and industry-specific considerations alone have the dual benefits of explainability and ex-

trapolability. By developing plant monitoring systems that enable operators to understand

plant state, quickly identify developing faults, and mitigate issues before they cause harm,

designers can radically improve industrial system operations and management. Through

thoughtful human-centered design of the interfaces between human and machine, they can

elevate the role of industrial operators to orchestrate the plant monitoring system’s set of

autonomous routines.

This dissertation presents a methodology for the systematic design and implementation

of a plant monitoring and operator support system running a fault diagnostic and decision

support engine that can be adapted for a variety of industrial monitoring applications. It then

demonstrates, by proof-of-concept application to an experimental thermal-hydraulic facility

- the Compact Integral Effects Test (CIET) - and advanced control room testbed - the

Advanced Reactor Control and Operations (ARCO) facility - the iterative plant monitoring

system development process. The focus of this dissertation is the advanced nuclear power

industry and the Fluoride salt-cooled High-temperature Reactor (FHR).

This dissertation is organized into eight sections. The first section introduces the back-

ground and motivation for model-based industrial monitoring systems before the second

section provides an overview of the state-of-the-art for nuclear and other industry plant

2

monitoring systems before focusing on nuclear industry challenges and opportunities. The

third section details the iterative fault diagnostic system development methodology and the

fourth section describes one approach to decision support and fault mitigation algorithm

design. These sections also walk the reader through an example application. The fifth sec-

tion then introduces the ARCO-CIET facility used in the case study and the sixth section

describes the operator support and human-machine interface design for ARCO. Finally, the

seventh section presents the case study plant monitoring system design and results before

the eighth section discusses promising applications of the overall design methodology.

This dissertation presents a methodology with the potential to guide the plant moni-

toring system development process across a variety of industries with the following original

contributions: a methodology for iterative fault diagnostic system development using in-

terdisciplinary information, recommendations for choosing plant models to build context

between different monitoring objectives, a methodology for developing decision support rou-

tines, and guiding principles for plant monitoring system human-machine interface design

and implementation in modern industrial control rooms.