Machine Learning in Manufacturing
6 February 2024
Technical University of Denmark
Meeting Center room S09
Anker Engelunds Vej 1
2800 Lyngby
Uncover the Future of Manufacturing with Machine Learning!
Dive into an illuminating seminar where academia meets industry at the crossroads of innovation. Join us for a dynamic day featuring 8 captivating presentations that span the spectrum of manufacturing applications. The power of AI on quality assurance is presented through cutting-edge computer vision in production processes, in both medical industry and food production.
Don’t miss your chance to be part of this transformative exploration on the synergy between Machine Learning and Manufacturing.
Organizers
This seminar has been developed by Hans Nørgaard Hansen (Head of Department, Professor at DTU Construct) from Teknologisk Videndelings steering group for manufacturing and Yang Zhang (Senior Reseacher from DTU Construct)
Program
|
Registration and breakfast | |||
|
Welcome and introduction | |||
|
AI, Machine Learning and the perspectives of these technologies Morten Mørup, Professor, DTU Compute Artificial Intelligence and machine learning are changing many aspects of our society. In this presentation, I will cover central concepts within AI and machine learning, how an AI learns to solve tasks, as well as some of the challenges and perspectives there are with these technologies. Morten Mørup is Professor of Machine Learning for the Life-Sciences at DTU Compute, Section for Cognitive Systems. His research focuses in particular on multi-way/tensor decomposition approaches, statistical complex network modeling approaches and Bayesian inference with applications to life-science data. He has been head of studies for the AI&Data B.Sc. education and is involved in introductory to advanced machine learning courses at DTU. |
|||
|
Automated vision-based inspection of mould and part quality in injection moulding and closed-loop control Yang Zhang, Senior Researcher, DTU Construct Automated real time quality monitoring is one of the key enablers for future high-speed production. In this research, an in-process monitoring procedure based on computer vision inspection and deep learning is proposed to indicate the tool and part quality during soft tooling injection moulding. Multiple types of injection moulding defects can be detected by the proposed method. Geometrical dimensions of the part can be measured simultaneously and the uncertainty can be quantified. Based on the obtained data, automated quality evaluation can be achieved in-process and a decision signal can be sent back to the injection moulding system for process adjustment. Yang Zhang is an experienced Researcher with a demonstrated history of working in the research and industry. Skilled in Polymer injection moulding, Surface Metrology, Polymers,Laser Processing, and Project Management. Strong research professional with a PhD focused in Manufacturing Engineering from Technical University of Denmark. |
|||
|
Coffee and networking break | |||
|
Enabling AI with old school maths Stefan Karlsson, CEO, Rossaker AB Enabling AI with old school maths? In this presentation, I will explain why and how. Some highlights are as follows. Use large to big sets of process data in real time to create a trustworthy white-to-grey-box AI based on a traditional stochastic process model which supervises black-box AI optimizers. Bring high velocity to traditional big data pipelining, apply cleaning, sorting, fuzzy matching such operations to create high intelligence in the white-to-grey-box AI supervisor & better chances for black-box AI optimizers to succeed. He has a background on real time programming in areas such as graphics, mathematical statistics and simulations. He is specialised in database-backends and data models, such data models are often created with customised backends or by combining steady load and parallelism to guarantee massive scalabilities. |
|||
|
From Robots to Autonomous Systems through Machine Learning Lazaros Nalpantidis, Professor, DTU Electro In this talk, I will discuss how machine learning can provide robots with the necessary flexibility to adapt to and cope with unforeseen situations, ultimately paving the way toward Autonomous Systems. I will argue that robust perception is the key to this direction. Such autonomous systems have the potential to revolutionize our society—the way we live, work, and interact. With the ability to make decisions and perform tasks without human intervention, these systems are driving innovation and growth across industries by improving efficiency, safety, and sustainability. Ultimately, autonomous systems can help our society achieve the goal of the twin—green & digital—transition. Lazaros Nalpantidis is Professor of Autonomous Systems at the Department of Electrical and Photonics Engineering, DTU, where he is also the Head of the Automation and Control Group. Lazaros’ academic interests focus on Autonomous Systems, lying at the intersection of robotics, AI and perception. Within the field of Autonomous Systems, Lazaros has exhibited a remarkable track of achievements both in research and teaching, promoting the field within Denmark and internationally. |
|||
|
Lunch and networking | |||
|
Dimensional features extraction by semantic segmentations on point clouds Shuo Shan, PhD student, DTU Construct Deep learning-driven semantic segmentation of point clouds offers substantial support in the extraction of dimensional characteristics. Once point-level semantic data is assigned, it enables the swift and precise identification of target point clouds, facilitating subsequent computations for extracting the necessary dimensional attributes. This presentation takes the geometric shapes of 3D printing as a case study and introduces the application of point cloud semantic segmentation in dimensional feature extraction. Shuo Shan is a PhD student at DTU Construct. His research concerns digital fingerprint in advanced manufacturing. |
|||
|
Creating “Dynamic Composition” within a standalone meeting room video-bar utilizing machine learning models Lionel Plato Kuhlmann, PhD, Manager of Advanced Video Processing Group, GN Audio A/S At GN Audio we have extended the functionality of a meeting room appliance from a “simple end of room” camera device to an intelligent device, where AI or Machine Learning models have been trained to identify meeting room participants. Where the initial video-bar created a high resolution 180 degree 4K image of the entire room, the new functionality was developed to also include separate video channels of individual presents and active meeting participants at high resolution. This enables a completely new experience for remote meeting participants – as if all meeting participant had their own personal camera. In this presentation we will highlight key multimodal machine learning models comprising the heart of the added functionality. How sound AI and video AI is combined, and how these models are tuned for speed and size to fit within an “IOT” device. In addition we will highlight a few other active research areas within GN Audio, where novel machine learning techniques are utilized for new and existing products and services. Lionel Plato Kuhlmann hold a PhD from DTU Computer within Statistical Image Processing from 1995. He works as a manager of an R&D group within GN Audio called “Advanced Video Processing Group”, and serves as a systems architect for next generation video meeting room devices. He has been active in image processing and applied machine learning for more than 25 years and is currently main mentor for an industrial post-doc project between GN Audio and DTU-Electro (Photonics). |
|||
|
Coffee and networking break | |||
|
Modern Machine Learning in Manufacturing - What Does it Take? Henrik Aanaes, Senior Project Manager & Principal Researcher, 3Shape What does it take to apply and deploy modern day machine learning technology? Where “modern machine learning “is understood as very data intensive, in particular deep learning? I will give some pointers and perspectives. I will be drawing on my research in 3D computer vision, e.g., applied to manufacturing, as an Associate Professor at DTU, and my job the last five years of making AI algorithms for 3D intra oral scanner at 3Shape. I will e.g., touch upon why deploying deep learning is harder than many people think and not for the most obvious reasons. Also, I will argue that metrology needs to be more present inline in the production line. Henrik Aanæs is Senior Project Manager & Principal Researcher at 3Shape, where he works with AI for constructing 3D models from sensor data. Before this he held the chair in 3D computer vision at DTU, where he e.g., had considerable activities within manufacturing. |
|||
|
Machine Learning in the food Industry - Benefits, Challenges and Applications Dennis Brandborg Nielsen, Head of Section for Data Analysis, Danish Technological Institute Machine Learning (ML) solutions have the capability to effectively streamline arduous and time-consuming procedures, including tasks such as ensuring food quality, sorting, and packaging. ML can enhance the utilization of raw materials by automating the interpretation of various processes. This entails harnessing machine learning algorithms to sift through extensive datasets, pinpointing distinctive attributes indicative of processes or quality markers. The integration of ML within the food industry not only serves to reduce substantial labor expenses but also accelerates production cycles. The employment of ML algorithms can span a wide array of applications, encompassing the prediction and management of debris and food particles throughout various operational stages. Dennis Brandborg Nielsen is the Head of Section for Data Analysis in the Centre for Sustainability and Digitisation at the Danish Technological Institute. The Danish Technological Institute develops digitisation solutions for sustainable food production. |
|||
|
Panel discussion and closing remark |
Registration fee
DKK 2,750 | Members of Teknologisk Videndeling (former ATV-SEMAPP) and promoting partners listed in the registration form |
DKK 3,280 | Non-members |
DKK 1,075 | PhD Students |
DKK 325 | BSc and MSc students (Membership is free of charge – register here. Early bird discount does not apply) |
All prices are exclusive of 25 % VAT.
Early bird discount of DKK 300 when registering before 12 January 2024.
Registration
Register – CLICK HERE
Binding registration
Registration is binding, however substitutions are accepted at any time.
Questions
Please do not hesitate to contact ATV-SEMAPP by e-mailing teknologiskvidendeling@construct.dtu.dk
Promoting partners