Efficient and safe interaction in production

Project No. IFA 5115

Status:

completed 12/2012

Aims:

Industrial production installations are increasingly expected to deliver greater efficiency and, at the same time, flexible production processes. Changes in the properties of raw materials, variable product specifications and flexible interaction with operators are registered automatically in order for the subsequent production process also to be adapted automatically. This transition from rigid to user-centric, adaptive automation concepts gives rise to new challenges regarding the safety and availability of the installation. In intelligent production environments of this kind, processes are no longer all stored within the control software; instead, the instantaneous behaviour of the machine is determined by continual changes in the underlying conditions referred to above in conjunction with the selected adaptive mechanisms. In order for safety to be maintained under these conditions, special measures are required. The far-reaching integration of the personnel into the process requires them to interact directly with moving machine parts. Whilst this has traditionally been an important measure for the attainment of installations with high flexibility, it leads to very high demands being placed upon safety. The objective of the joint EsIMiP project is for the production movements of the machine to be controlled by autonomously operating components for machine learning and the computation of adaptive algorithms in real time. In this approach, the control signals of the "unsafe" components referred to above are always checked by a deterministic "safe" control system. Should machine components approach the operator dangerously, the safe control system takes control of the machine's movements and brings it if necessary to a halt.

Activities/Methods:

In EsIMiP, an innovative control method was developed in order to permit safe collaboration between the operators and unguarded moving machines. For this purpose, realistic models of the human movement behaviour in the working environment were created and used for calculation of the control strategies in real time. Owing to the non-deterministic nature of the human behaviour, it was necessary for probabilistic human models to be used, and the algorithms for ideal and safe control strategies to be adapted to this model type. Process states critical to human safety, in which the machine would have to come to a halt, are anticipated in good time and prevented, for example by means of evasive movement. For this to be achieved, detection of the human being must be possible in his or her actual industrial environment in real time, reliably and without further aids (such as markers or background patterns), and acquisition of his or her location and working position must be possible with sufficient precision. Multiple sensors employing a range of principles were therefore used in order to prevent obstruction by objects in the working area (including, for example, other persons). Human movement behaviour was observed in a range of contexts (constant vs. variable robot behaviour). Several factors influencing the acceptance of the work system were determined and examined more closely by means of tests on test subjects. Tests involving test subjects were conducted at virtual robot workplaces at the IFA. Specialized and highly reliable industrial controls were used and developed further on the robot in order to provide the necessary technical platform. All developments were implemented at a workplace employing a collaborative robot, and subsequently demonstrated.

Results:

Within the project, a new approach was pursued in which a "strategic automation component" enhances the availability of the system by the learning of control strategies with consideration for the operators' behaviour and the safety functions. At the same time, the safety of the system as a whole is assured by an independent "operative automation component", in the development of which the IFA was involved. The environment of the robot is monitored continually. This enables the behaviour of the persons in the production environment to be predicted and the production behaviour of the robot therefore to be efficient whilst at the same time not impairing the safety of the system as a whole. The industrial partners to the project were able to achieve major further developments in safe drive, control and robot technology. For the monitoring of the working area, it was found to be essential for both the type and location of the different sensors to be adapted to the task to be performed by the system concerned. The safe ultrasonic sensors were placed by the IFA on the robot itself in order to determine the safety-related reduction in robot speed. The unsafe sensor components were implemented in the form of colour cameras located externally for control of the movement. These identify the areas of the work cell occupied by operators, in order for the robot movement to be continually recomputed. The relationship between speed and distance can be defined here for each object. The ergonomic studies of the project showed the importance of two key aspects: firstly, the study of human movement behaviour; secondly, analysis of the acceptance of the collaborative human-robot work model. The results of the study enable ergonomic requirements placed upon the system to be specified and, by implementation of the findings in the control architecture and in the design of the workplace, the probability of human error to be reduced.

Last Update:

2 May 2016

Project

Financed by:
  • Deutsche Gesetzliche Unfallversicherung e. V. (DGUV)
  • Bayerische Forschungsstiftung
Research institution(s):
  • Institut für Arbeitsschutz der Deutschen Gesetzlichen Unfallversicherung (IFA)
  • TU München - Lehrstuhl für Automatisierung und Informationssysteme
  • TU München - Lehrstuhl für Ergonomie
  • Universität Kassel - Fachgebiet Regelungs- und Systemtheorie
  • Baumüller Anlagen-Systemtechnik GmbH & Co. KG
  • Reis GmbH & Co. KG Maschinenfabrik
Branche(s):

-cross sectoral-

Type of hazard:

mechanical hazards

Catchwords:

accident prevention, machine safety

Description, key words:

industrial production installations, collaborative robots, adaptive automation concepts, ultrasonic safeguard, human factors

Contact