“Deep Learning techniques applied to industrial automation problems”

“Deep Learning techniques applied to industrial automation problems”

Engineering for the Information Society and Sustainable Development

Dr. Alberto Tellaeche Iglesias – University of Deusto. Google Scholar: https://scholar.google.es/citations?user=LHifZmAAAAAJ&hl=es Research Gate: https://www.researchgate.net/profile/Alberto_Tellaeche

The University of Deusto is one of the oldest universities in Spain, founded in 1886. More than 40 years ago, Deusto Engineering faculty was the first in the country to offer Computer Science studies. Nowadays this faculty covers a wide spectrum of engineering branches such as Computer Science, Automation, Electronics, Mechanics, Industrial Technologies, etc. Today, the Faculty of Engineering counts with more than 100 researcher profiles, with more than 600 research projects in the last 20 years, summing up more than 60 million euros. The Research Unit that will hold this proposal is Deusto for Knowledge (D4K), recognized by the regional government since 2010, focusing its activity in Applied Computational Intelligence. This group has experience in applying soft computing techniques to industrial processes , and continues focusing its efforts on giving solutions to advance manufacturing applications and to all the Industry 4.0 challenges that may arise.

Applicants for positions must hold an MsC degree in Computer Science, Electronics, Automation or Industrial Engineering. Experience in developing industrial applications interfacing with automation equipment such as robots, cameras, PLC, etc, will be highly valuable. Good knowledge of Embedded Systems, Linux, C++, Python and MATLAB. Strong mathematical foundations and knowledge of statistics are also needed. Fluency in English (understanding, speaking and writing) and ability to work in a team and collaborate with people in complementary disciplines are essential.

• Information Sciences and Engineering (ENG)

Industrial applications in the digital industry or Industry 4.0 seek to carry out tasks in unstructured environments just as a human worker would do in a similar situation. Adaptability of the production systems and rapid adoption of new conditions are among the most important advancements that this new paradigm is bringing to the manufacturing companies. This flexibility, adaptability and productivity leads to a much more complex situations in terms of engineering processes. Because of this, the artificial intelligence (AI) plays a fundamental role in the new industry 4.0. Robotics applications for picking objects, inspection lines to detect texture failures in production, etc, need an advanced processing of data using AI to fulfill requirements. The applicant will consider advanced robotics and machine vision problems in industry, using advanced AI techniques such as Deep Learning algorithms to obtain a solution for them that will overcome the current state of the art. Taking all this into account, we are looking for a student motivated enough to develop research activities in this area. Specifically, the project will involve, but it is not limited to, the following aspects: Development of machine learning and deep learning algorithms for industrial robotics applications, such as Bin Picking. Use of AI in advanced industrial machine vision algorithms: texture recognition, failure detection, etc. Integration of these algorithms in systems using real industrial hardware.

Maurtua, I., Fernandez, I., Tellaeche, A., Kildal, J., Susperregi, L., Ibarguren, A., Sierra, B. (2017). Natural multimodal communication for human-robot collaboration. International Journal of Advanced Robotic Systems. Vol 14, Issue 4. SAGE Publications. Tellaeche, A., Arana, A. (2016) Robust 3D object model reconstruction and matching for complex automated deburring operations. Journal of Imaging 2(1), 8. Tellaeche, A. Maurtua, I., Ibarguren, A. (2016). Use of machine vision in collaborative robotics: An industrial case. In 21st International Conference on Emerging Technologies and Factory Automation. IEEE. Pastor-López, I., Santos, I., Santamaría-Ibirika, A., Salazar, M., de-la-Pena-Sordo, J., & Bringas, P. G. (2012, July). Machine-learning-based surface defect detection and categorisation in high-precision foundry. In Industrial Electronics and Applications (ICIEA), 2012 7th IEEE Conference on (pp. 1359-1364). IEEE. Santos, I., Nieves, J., Bringas, P., & Penya, Y. (2010). Machine-learning-based defect prediction in high precision foundry production. Structural Steel and Castings: Shapes and Standards, Properties and Applications, 259-276.

The scientific in charge of this proposal, Dr. Alberto Tellaeche, holds a PhD degree in Computer Science, obtained in 2008, with research focused on machine learning applied to natural images. For the last 11 years he has been focused on research projects that comprise machine vision, artificial intelligence and robotics. The most significant ones are cited below: PIROS – EUROC (Challenge 1 Robotic pallet preparation using force and vision control). Competitive European Project. Budget: 173.160,0 € FourByThree (Highly customizable robotic solutions for effective and safe human robot collaboration in manufacturing applications). Competitive European Project. Budget: 1.001.240,0 € DEBUR ( Automatic Deburring System of injected 3D components). Competitive European Project. Budget: 260.875,0 € A4BLUE – Adaptiv Automation of collaborative part mounting to enhance workers´ satisfaction. Competitive European Project. Budget: 1.011.750,0 € Projects representative of the research unit D4K, are: VESSEDIA:Verification Engineering of Safety and Security Critical Dynamic Industrial Applications. Competitive European Project. Budget: 352.125 € HORUS: Hornos de recalentamiento inteligentes para procesos siderúrgicos competitivos y sostenibles. CIEN Spanish Call. Budget: 122.890 € FIPS: Future Intelligent Production System. Private project with the U.S. Navy. Budget: 137.000 €

This proposal is presented to progress beyond the state of the art in new manufacturing problems of recent appearance, related with the new Industry 4.0 paradigm. The following knowledge areas integration will be needed: Industrial Automation: Programing of robots, use of sensors and cameras. Mechanics: For scenario setup. Electronics: Connection of elements, etc Computer Science: Fundamental knowledge area for IA algorithm development.

Yes. Co-direction of the thesis will be carried out by another scientific of the D4K group.

While the research is mainly focused on Computer Science and AI algorithm development, there is also a demand of construction of real industrial setups to test the algorithm developments, after a first stage of algorithm modeling and analysis. Industrial setups and real interaction with the hardware (cameras, sensors, robots, etc) will require from the candidate the use and integration of different engineering disciplines, such as mechanics, electronics, etc.

The 2018-2020 work program focuses efforts on four principal topics (https://ec.europa.eu/programmes/horizon2020/en/h2020-section/cross-cutting-activities-focus-areas), one of them being «Digitising and transforming European industry and services». The budget for this area is 1.8 billion Euros, which gives an idea of the importance the European Commission (EC) gives to this specific subject. Expanding the idea more, the EC seeks a development in «Digitisation of products, services and processes will transform industry and provide solutions to several major societal challenges.» The research area proposed fully aligns with this idea. According to the 2030 agenda for Sustainable Development this project will contribute directly to advance in the following goals: 5th (Gender Equality), 8th (decent work and economic growth),9th (Industry, Innovation and Infrastructure), 11th Sustainable Cities and Communities and 12th (Responsable Consumption and Production), having also a positive indirect impact on the rest.

Initially there is no international co-direction or «co-tutelle» planned.

The work intended to develop as a result of this research project has direct applicability in industrial companies in the manufacturing sector. Companies interested could easily contribute as external stakeholders for this project, starting with the Etxe-Tar group, an international known group of machine tool campanies, that is going to collaborate closely in the research activities proposed in this work.

The research topic presented in this proposal has direct applicability to industry. Historically, the Basque Country region has a long and strong industrial tradition, focused on the production of machine tool goods, parts for the automotive sector, etc. This region is one of the strongest in Europe in this field, and machines and solutions built in the Basque Country are exported worldwide. This proposal comes directly to cover the obvious necessity of the basque industrial manufacturing sector, detected by Dr. Alberto Tellaeche in his more than 15 years of real industrial experience. Outcomes of this research will have scientific quality for the obtaining of a doctorate degree, and also direct applicability in the solution of present industrial problems. As stated in the previous section, the Etxe-Tar group (http://www.macarbox.com/es/etxetar-group) is going to be the main beneficiary of the findings of this proposal, as agreed in the collaboration of the company with this project.

Smart Manufacturing is a priority in the actual industrial research. Although in the last years there have been steps forward to obtain more flexibility in the production lines, there is still a big gap between the real achievements and the theoretical objectives that could drive to a real new production scenario. A good example of this situation in the operation of Bin Picking in robotics, that is, picking of several different objects from an unstructured environment (a bin) using a robot for machine tending, for example. Many attempts have been done so far to solve this problem, but it is still in a very early stage of development. Another example is the production quality control. There are many situations, such as reflective surfaces or surfaces with special textures, where scratches, breakages or other typology of errors can not be detected with machine vision algorithms present in the state of the art. The work proposed in this research topic aims to find a real implementable solution (algorithms + real implementation) to give a solution to these type of problems in industry.

Research and development of new algorithms for solving challenging industrial proposed problems in manufacturing, using IA. Real implementation of these innovative solutions to assess their validity in a real production situation. Propose a real implementation of an innovative solution to the above mentioned problems.

The inclusion dimension will be addressed indirectly by helping to solve problems that nowadays are very physically demanding for the human workers. The current situation in problems such as the Bin Picking is that, although some approaches exist to ease this operation, in the real world, industrial workers have to pick manually parts for machine tending. This parts can be particularly heavy. By apporting intelligent automatic solutions to industrial problems, the role of the shopfloor workers will change, preventing injuries, accidents, dedicating their work to other healthier activities. In the same way people that now has physical limitations to work in industrial manufacturing, will now be able to contribute to industry in different new ways.