Refine
Document Type
- Article (1)
- Conference Proceeding (2)
Language
- English (3)
Is part of the Bibliography
- no (3) (remove)
Keywords
- AI (3) (remove)
Institute
Methods of machine learning (ML) are difficult for manufacturing companies to employ productively. Data science is not their core skill, and acquiring talent is expensive. Automated machine learning (Auto-ML) aims to alleviate this, democratizing machine learning by introducing elements such as low-code or no-code functionalities into its model creation process. Due to the dynamic vendor market of Auto-ML, it is difficult for manufacturing companies to successfully implement this technology. Different solutions as well as constantly changing requirements and functional scopes make a correct software selection difficult. This paper aims to alleviate said challenge by providing a longlist of requirements that companies should pay attention to when selecting a solution for their use case. The paper is part of a larger research effort, in which a structured selection process for Auto-ML solutions in manufacturing companies is designed. The longlist itself is the result of six case studies of different manufacturing companies, following the method of case study research by Eisenhardt. A total of 75 distinct requirements were identified, spanning the entire machine learning and modeling pipeline.
Methods of machine learning (ML) are notoriously difficult for enterprises to employ productively. Data science is not a core skill of most companies, and acquiring external talent is expensive. Automated machine learning (Auto-ML) aims to alleviate this, democratising machine learning by introducing elements such as low-code / no-code functionalities into its model creation process. Multiple applications are possible for Auto-ML, such as Natural Language Processing (NLP), predictive modelling and optimization. However, employing Auto-ML still proves difficult for companies due to the dynamic vendor market: The solutions vary in scope and functionality while providers do little to delineate their offerings from related solutions like industrial IoT-Platforms. Additionally, the current research on Auto-ML focuses on mathematical optimization of the underlying algorithms, with diminishing returns for end users. The aim of this paper is to provide an overview over available, user-friendly ML technology through a descriptive model of the functions of current Auto-ML solutions. The model was created based on case studies of available solutions and an analysis of relevant literature. This method yielded a comprehensive function tree for Auto-ML solutions along with a methodology to update the descriptive model in case the dynamic provider market changes. Thus, the paper catalyses the use of ML in companies by providing companies and stakeholders with a framework to assess the functional scope of Auto-ML solutions.
In recent times, both geopolitical challenges and the need to counteract climate change have led to an increase in generated renewable energy as well as an increased demand for clean electrical energy. The resulting variability of electricity production and demand as well as an overall demand increase, put additional stress on the existing grid infrastructure. This leads to strongly increased maintenance demands for distribution system operators (DSOs). Today, condition monitoring is used to address these challenges. Researchers have already explored solutions for monitoring critical assets like switchgear and circuit breakers. However, with a shrinking knowledgeable technical workforce and increasing maintenance requirements, mere monitoring is insufficient. Already today, DSOs ask for actionable recommendations, optimization strategies, and prioritization methods to manage the growing task backlog effectively. In this paper we propose a vision of a grid-level cognitive assistance system that translates the outcome of diagnosis and prognosis systems into actionable work tasks for the grid operator. The solution is highly interdisciplinary and based on empirical studies of real-world requirements. We also describe the related work relevant to the multi-disciplinary aspects and summarize the research gaps that need to be closed over the next years.