Model-based Process Surveillance and Optimisation for Fault Detection and Diagnosis

Georg Ørnskov Rønsch

Research output: Book/ReportPh.D. thesis

258 Downloads (Pure)

Abstract

With the introduction of Industry 4.0, focus on digitisation and digitalisation have become very apparent. Many companies have this high on their agenda, and many resources are spent on adjusting the overall ideas and concepts to fit within the context of the given company. One of the central aspects of Industry 4.0 is the utilisation of data to achieve increased productivity, new products and business opportunities. This industrial PhD thesis has been conducted in close collaboration with an industrial partner and focuses on utilising manufacturing data to achieve increased productivity. The industrial partner is a leading producer of injection moulded construction toys. Therefore, the work is related to injection moulding; however, the findings and learnings generalise well to other industries. Initially, the formulation of the project scope was related to increasing productivity through monitoring and control of the injection moulding process. The intent was to utilise newly established machine connectivity at the industrial partner to collect readily available machine process data and explore how best to develop and implement a process monitoring solution. The expected and needed data was not collected because of the complexity encountered when initiating the data collection and changing prioritisation. This setback provoked reflections on the purpose and approaches for data utilisation in manufacturing. Eventually, it changed the project’s focus, however, still with the overall focus on improving productivity. The new focus became centred around how to become effective in scoping and exploring data utilisation1. Utilising data in a manufacturing setting is a demanding task with many obstacles
to pass. In the past and during the PhD project, this has caused many frustrations and challenges, which have been condensed and transformed into a data utilisation
framework, supporting a practical exploration and implementation approach. The framework was first presented in paper A (”An investigation of the utilisation of different data sources in manufacturing with application in Injection Moulding”) and later extended to include more learnings and aspects of data utilisation in a manufacturing context. The central part of the framework is formulating a core business objective, followed by identifying key manufacturing drivers that have to be addressed to improve the objective. For each key driver, a number of applications are identified as potential contributors for optimising the business objective. At this stage, the work moves into a practical exploration of the identified applications. This is proposed to be conducted sequentially, starting with applications that require the least new data integration and collection. As the maturity in the organisation is growing, more complex data integration and utilisation can be initiated. All applications should be evaluated against the total cost of utilising a given data source and the potential value creation. The thesis is a collection of six scientific papers where five of these include examples of data utilisation applications in a manufacturing context2. Three papers address aspects of predictive maintenance to reduce productivity loss caused by mould failure. Two of these papers utilise existing quality data and maintenance history data for mould worn-out evaluation, where the third explores the use of acoustic emissions for real-time condition-based maintenance of injection moulds. The use of acoustic emissions have proven promising for detecting the lack of lubrication and mechanical defects represented by loss latch lock. The remaining two papers explore utilising underlying equipment time-series data to capture process dynamics not detected in readily available process data. The changing process dynamic explored is caused by variations in raw materials and impacts product quality. Using designed experimentation and machine learning, a solution is presented to detect the disturbance and
identify the optimal process settings to reduce the effects on product quality. Besides the work presented in the scientific publications, the thesis includes initial exploration using a multivariate approach to monitor product quality and the injection moulding process. This work is included since it represents some of the challenges of utilising data in a manufacturing context and is important for the overall reflections, recommendations, and conclusions.
“Data is the new oil. It’s valuable, but if unrefined, it cannot really be used. It has to be changed into gas, plastic, chemicals, etc to create a valuable entity that drives profitable activity; so must data be broken down, analyzed for it to have value.”
— Clive Humby, UK Mathematician and architect of Tesco’s Clubcard.
The quote by Clive Humby is very relevant and widely used in relation to data utilisation within Industry 4.0, and often what upper management in companies refer to when initiating the journey of data utilisation. One thing that is seldom realised is the complexity related to this and the amount of resources and dedication required to make it happen. This should by no means be a reason for not starting the data utilisation journey, just a heads-up, to align expectations.
Original languageEnglish
PublisherTechnical University of Denmark
Number of pages254
Publication statusPublished - 2022

Fingerprint

Dive into the research topics of 'Model-based Process Surveillance and Optimisation for Fault Detection and Diagnosis'. Together they form a unique fingerprint.

Cite this