XPRESS 42 Months Publishable Executive Summary
Thesis: Formal methods for reconfigurable assembly systems
Author: Tiago Oliveira Ribeiro
Retrieval of Process Methods from Task Descriptions and Generalized Data Representations
Authors: Jürgen Pollak, Alireza Sarveniazi, Norbert Link
Intelligent machines are supposed to set process parameters automatically when faced with a task to be processed. The intelligence is often realized by databases linking the task with process parameters. This paper introduces a flexible and portable (to various processes) system to find (optimized) process parameters, which force the process outcome to predefined quality under given, variable conditions. The process knowledge is extracted from experiments and represented by a generalization in the form of so called goal-functions based on support vector methods. A level set method is proposed to find all process parameters which fulfill a given task. We also provide experimental results based on numerical simulations applied to Resistance Spot Welding to demonstrate the usability of the concept. Promising numerical results concerning the accuracy of process parameters and the process outcome approve the concept for the implementation on real process machines.
Journal: The International Journal of Advanced Manufacturing Technology
Transport with Conveyors in the Factory of the Future
Authors: Almeida, F. L.; Terra, B. M.; Dias, P. A.; Goncalves, G. M.;
The project XPRESS introduces a completely new scalable concept of an manufactronic networked factory, which is composed by a co-ordinated team of specialized autonomous objects (manufactrons), each knowing how to do a certain process optimally. This paper looks in particular to the transport manufactron implementation, which is an manufactron responsible to carry material at the shop-floor. It is suggested an architecture based in a multi-agent approach that uses a conveyor belt transport manufactron to perform this task. Additionally, it is presented an implementation of the conveyor belt transport manufactron that uses the .NET framework integrated with the manufactronic framework.
Journal: Automation Quality and Testing Robotics (AQTR), 2010 IEEE International Conference,28-30 May 2010 at Cluj-Napoca, Romania.
Recognizing User Interface Control Gestures from Acceleration Data Using Time Series Templates
Authors: Pekka Siirtola and Perttu Laurinen and Heli Koskimäki and Juha Röning
This study presents a method for recognizing six predefined gestures using wrist-worn tri-axial accelerometer. The method is designed to control a simple user interface by matching the shapes of wrist movement to predefined time series templates. In this study matching is done by using three different trajectory distance measures and the results show that weighted double fold distance measure is the most accurate. The results show that the proposed method can be used realtime. User dependent version of the method recognizes gestures with accuracy of 94.3% and the recognition rate of the user independent version is 85.5%.
Journal: ICINCO 2010 - 7th International Conference on Informatics in Control, Automation and Robotics
Indoor 3D Position Estimation Using Low-Cost Inertial Sensors and Marker-Based Video-Tracking
Authors Bastian Hartmann, Norbert Link
A lot current research utilizes vision-based systems for indoor object position tracking applications. An advantage of vision-based tracking systems is that they are relatively accurate for tracking objects in spaces of small dimensions, such as rooms. However, because of the fact that these systems utilize vision-based sensors, they are limited to the line of sight and thus suffer from occlusions. Moreover, computer-vision algorithms have comparable high computing costs. In this paper, a system for indoor 3D position estimation based on a low cost (MEMS) IMU is presented. The system is aided by position measurements from a marker-based video tracking system, which allows precise position tracking of pre-defined marker tags. In order to acquire accurate timestamp information, sampled measurements from IMU and video sensors are synchronized in hardware by using triggerable cameras. For estimating 3D positions, an Extended Kalman Filter is utilized. With the combination of inertial sensors and vision-based position tracking the presented system is able to accurately track objects in space-limited environments while compensating short-time occlusions (with position errors in the order of centimetres after a few seconds) and providing higher sampling rates at reasonable computing costs. The focus of this application is on estimating 3D position information for human activity recognition such as hand gesture recognition. However, the system is not limited to this application and may be used for indoor object tracking applications in general.
Journal: IEEE/ION Position Location and Navigation System Conference (PLANS 2010), Indian Wells (Palm Springs), California May 3-6, 2010, in press
IEEE/ION Position Location and Navigation System Conference(PLANS 2010)
Manufacturing Process-Learning From Experimental Data Applied to Resistance Spot Welding
Authors: Alireza Sarveniazi, Jürgen Pollak, Norbert Link
This paper describes and evaluates the use of support vector regression to determine a statistical model for a nugget diameter function associated with a spot welding process between two materials A and B. The nugget diameter function plays a central role in the task to method transformation part of the Xpress-Project, where a process method is automatically derived from a task description. We are using a goal function, where the level set provides the task-related methods. Based on the training data produced by a simulation or an experimental measuring, we are using support vector regression to extract a nugget diameter function (our goal function) of five variables: current, welding time, force, sheet thickness of material A, sheet thickness of material B armed by a Welding-Manufactron. According to this nugget diameter function, we develop an explicit description of our optimized method needed by an intelligent welding machine.
Journal: World Academy of Science, Engineering and Technology
Volume 58, Part IX
Activity recognition using a wrist-worn inertial measurement unit: A case study for industrial assembly lines
Authors: Heli Koskimäki, Ville Huikari, Pekka Siirtola, Perttu Laurinen, Juha Röning
While wearable sensors are becoming more common their utilization in real world applications is also becoming more attractive. In this study, a single wrist-worn inertial measurement unit is attached to active wrist of a worker and using acceleration and angle speed information it is decided which activity the worker is performing at certain time intervals. The activity information can then be used for proactive instruction systems or to ensure that all the needed work phases are performed. In this study, the selected activities are basic tasks including hammering, screwing, spanner using and using power drill to screwing. In addition, a null activity class consisting of the other activities (moving around the post, staying still, change of the tool) was defined. The performed activity can then be recognized online using sliding window method to divide the data into two second intervals and by overlapping two adjacent windows by 1.5 seconds. Thus in every half a second the activity is recognized. The method used for the actual recognition is k nearest neighbor method with a specific distance boundary to classify completely new events as null data. In addition, the final class was decided using the majority vote for classifications of three adjacent windows. The results showed that using this kind of setting almost 90 percent accuracy can be achieved where the activity-specific accuracies for hammering, screwing, spanner using, power drilling and for null data are 96.4%, 89.7%, 89.5%, 77.6% and 89.0%, respectively. In addition, in case with completely new null events the usage of the specific distance measure improved the accuracy from 68.6 percent to 82.3 percent.
Journal: Proceedings of 17th Mediterranean Conference on Control and Automation
Worker Behavior Interpretation for Flexible Production
Authors: Bastian Hartmann, Christoph Schauer, Norbert Link
This paper addresses the problem of recognizing and interpreting the behavior of human workers in industrial environments for the purpose of integrating humans in software controlled manufacturing environments. In this work we generalize the methodology of behavior interpretation in order to give a generic approach for finding solutions for concrete applications in a flexible top down manner. This conceptual worker integration approach offers a versatile basis without being restricted to a rigid concept, or being focused on a specific problem or application. We instantiate our concept in a spot welding scenario, where the behavior of a human worker is interpreted when performing a spot welding task with a hand welding gun. For this application we acquire signals from inertial sensors, video cameras and triggers and recognize atomic actions by using pose data from a marker based video tracking system and movement data from inertial sensors. In a further step, the recognized atomic actions are analyzed on a higher evaluation level by a finite state machine, which represents a model of the spot welding task decomposed into interdependent task steps.
Journal Proceedings of World academy of Science, Engineering and Technology
Volume 58
Pages 828-836
ISSN 2070-3740
World Academy of Science, Engineering and Technology website
Adaptive manufacturing by encapsulation of functionality and expertise (english version)
Anpassungsfähige Fertigung durch Kapselung von Funktionalität und Expertise (german version)
Authors: Michael Hoffmeister and Roland Wertz
Within the EU research project XPRESS the Fraunhofer IPA realizes together with their project partners a new, innovative manufacturing architecture, where “Manufactrons” adopt the task-orientated planning, simulation and manufacturing execution in a distributed way. These Manufactrons can embody installations, plant components and operators as independent units within a collaborating factory. Manufactrons thereby enclose functionality and expertise. They can execute generalized task descriptions in the manufacturing process and are able to automatically choose the optimal mode of processing under given boundary conditions. Resulting quality data will be evaluated and integrated in future planning.