Despite the growing popularity of the Industry 4.0 paradigm, many companies still face some hurdles that hinder the adoption of a data-driven strategy to leverage the potential value of non-personal data generated by industrial processes and equipment. In particular, some of the professional roles required to deal with a Machine Learning (ML) project, e.g. Data Scientists and ML Engineers, are still difficult to find for manufacturing companies, especially in SMEs. Therefore, an automatic ML tool specialized in data-driven solutions for manufacturing would be very valuable.
The tool should be easy to use also for people with limited or no background in ML, and it should allow companies to quickly deploy smart monitoring solutions based on data, such as Anomaly Detection and Predictive Maintenance. This tool is the final goal of the proposed experiment. It is based on background knowledge the applicant has about industrial Machine Learning (ML) solutions and Decision Support System software design. Indeed, Statwolf was founded in 2014 by three Ph.D. holders in Machine Learning applied to Industry 4.0, to fill the gap between state-of-the-art academic methodologies and productive solutions.
The company’s core competences are Machine Learning and the design of Decision Support Systems. In particular, the company has designed a proprietary platform for end-to-end data analytics, the Statwolf Platform. Besides standard Business Intelligence features, the Statwolf Platform also features Machine Learning capabilities. The proposed tool for automatic ML in manufacturing would be based on Statwolf platform which provides Extraction, Transformation & Loading (ETL) capabilities, an analytical engine, ML modules and MLOps tools. Since all these components are already deployed in the market and currently used in many productive solutions, the experiment is expected to be very close to market [TRL8 system complete and qualified].
Implementing an automatic Machine Learning (ML) tool for the design of data-driven solutions for manufacturing companies. The tool should be easy to use also for people with limited or no background in ML.
- [Improved ML accessibility] AutoML4.0 allows users to deploy smart monitoring solutions without the need for advanced Data Science or ML Engineering skills.
- [Easier Data & ML assessment] The tool makes feasibility assessment simpler and can motivate the adoption of a data-driven strategy.
- [Production-readiness by design] The proposed software is based on the Statwolf platform that already offers state-of-the-art MLOps tools for the release of ML solutions.
- Impacts (social, environmental, …) Broader adoption of Machine Learning technologies can lead to huge impact in terms of energy and waste efficiency
- Thanks to the adoption of ML, manufacturing companies can be more effective in their operations and production, leading them to be more competitive in the global market.
- The diffusion of ML-based solutions will lead to an increment of higher qualified jobs in manufacturing
- Enriched product portfolio. AutoML4.0 will enrich the product portfolio and allow for a sales strategy where lower customization (and costs) is required.
- New market opportunities. The proposed tool will produce new market opportunities by addressing companies interested in undertaking a data-driven strategy but lacking Data Science and ML Engineering skills.
- Higher efficiency. AutoML4.0 will help Data Scientists at Statwolf in quickening Machine Learning feasibility analysis.
- Availability of implemented ML Algorithms for diagnostics and prognostics
- Availability of a Proof-of-Concept of AutoML4.0 Release of open datasets
- Letters of interest from manufacturing companies
- Availability of a Business Plan for the exploitation of the experiment results