Keywords

1 Introduction

Safety analysis for industrial robot cells is a crucial step during commissioning [1]. Every foreseeable potential hazard needs to be evaluated and minimized to ensure machine operator safety. Due to enhanced robot controller features new opportunities for the use of robot cells become more and more possible. This involves layout and task specific details, like work objects, grippers and handled objects [2, 3]. Since all these factors determine the safety concept, the demand for easy-to-handle safety systems with short commissioning times is further increasing. More versatile assembly or production systems need permanent update of the safety documentation [4]. Safety calculations are mostly based on minimal permitted distances between robot and objects or safety zones within the robot cell. Usually, safety concepts are derived based on assumptions for the position of robot and human and their respective maximum velocities. In practice minimal deviations of objects and safety equipment within the real cell can lead to a threat to the worker.

State of the art robot controllers enable complex configurations regarding safety functions. ABB’s SafeMove Pro feature, for example, allows for a case specific reduction of robot velocity, restricted working spaces by using safety zones and limiting the range of motion. Some safety zones only exist within the software, which makes them harder to evaluate within the real cell (Fig. 1). To compensate the lack of physical separating safety components, more flexible solutions like light curtains or laser scanners are used. An individual testing of the certain component may be easy to accomplish, but an evaluation of the overall safety concept is more complicated.

Fig. 1
figure 1

Example of a safety configuration for ABB robots (different colors mean different safety zones that are activated by the operator)

Full size image

Deviations in the minimal permitted spatial configuration of objects can lead to an insufficient safety concept. A proper validation concept needs to support the safety engineer to evaluate spatially complex safety zones. The following solutions are developed and presented in this work:

  • Automated visualization of safety zones and areas based on the specific safety concept.

  • Solution to compare DT to the real cell.

  • Robots as reference system.

  • Data from DT is usable to generate test program.

1.1 Safety Concepts for Robot Cells

For a variety of assembly or productions systems different approaches for safety concepts were developed. The higher demand for individualized products calls for production lines to get more and more close to lot-size 1 [5]. A well-known concepts behind this demand is the reconfigurable manufacturing system (RMS) [6].

Separating protective devices are widely used as a solution to ensure machine safety. Unfortunately, those systems lack flexibility while producing high costs if a change in cell design is needed. As an alternative, sensor-based security components such as laser scanners or light barriers are established. An example of a cell with different safety components is presented in Fig. 2.

Fig. 2
figure 2

Safety components of a robot cell [7]

Full size image

To ensure the exact position of every safety component and therefore the safety of the whole cell a safety certification must be conducted. To evaluate the risk and validate and verify the cell’s safety standards such as EN ISO 10218-1 [8] and EN ISO 10218-2 [9] are used.

2 Materials and Methods

As mentioned before the main goal of this work is reducing the validation time of safety concepts for industrial robot cells. One major part of every safety concept is to examine the distances between safety relevant objects, especially workers, and the robot. Mentioned in German national standards [8, 9] and international standards [10, 11] safety engineers must examine these aspects. The presented software VISIBLE generates an automated solution for this process. Based on a DT of the robot cell distanced between objects can easily be derived. This process generates the setpoints for the distances that now must be evaluated within the real system. A laser distance sensor mounted on the robot is used to examine the correct positioning of objects within the cell. The VISIBLE software generates a tool path planning for the laser distance sensor to properly measure the desired distances.

This method is designed to assist the safety engineer while commissioning production systems according to the mentioned standards.

Additionally, AR is used to visualize an overlap of real and virtual objects. The safety engineer uses AR-glasses (Microsoft HoloLens) to project the robot’s virtual safety zones while physically standing in the cell. A calibration of the AR-environment is performed guided by an assistant that automatically calculates the accuracy so that the user can properly calibrate the AR-glasses. The AR-glasses recognize the position of a calibration marker and align its coordinate system relative to the robot. A value benefit analysis was carried out comparing several different components such as:

  • accuracy

  • time consumption

  • necessary knowledge to use the components

  • error-proneness

  • failure probability

  • repeatability

  • safety within the process

  • projection of safety zone

  • calibration

  • documentation

The analysis resulted in AR-glasses and a distance measuring device being the most suitable solution. The used hardware is shown in Fig. 3. For the AR application a HoloLens and the HoloLens Development Edition by Microsoft is used. For distance measurement a GLM 120 C Professional by Bosch is used.

Fig. 3
figure 3

Used hardware: distance measuring device (e.g. line laser) (left), AR-glasses HoloLens (right)

Full size image

The whole concept is manufacturer independent. New installations and changes are easy to implement. The new methods of visualization can project even complex contours such as planes at different height levels. The new system links solutions for robotic and AR and decreases the barriers to entry for advanced industrial robot cells.

3 Results

The following section presents the results of this work. It is structured in different parts each regarding different aspects of the solution.

3.1 Development of a Method to Verify Safety Components

Initially literature research was conducted to get an overview of the state-of-the-art approaches to verify safety components. There are several norms and guidelines such as the EN ISO 10218, EN ISO 12100, EN ISO 13849 and several more that the approach is built on. The norms list different methods for planning, design, risk evaluation, verification, validation and documentation of safety components. The state-of-the-art sequence of the necessary tasks is presented in Fig. 4.

Fig. 4
figure 4

State-of-the-art sequence of the necessary tasks for commissioning of safety concepts

Full size image

After evaluation of the established sequence for plant commissioning a new method is developed. The implementation of the new method includes a comparison of the layout of the cell and the information of the DT. The process enables an automated documentation by picking up contours of physical and optical safety components from the planning tool in relation to the robot basis coordinate system. This feature is included in the new developed sequence presented in Fig. 5.

Fig. 5
figure 5

Method for automated commissioning of robot cells

Full size image

Further on the use of the HoloLens offers a solution to examine virtual safety zones. Zones that only exist within the robot control lack a physical counterpart in the real cell that makes them hard to evaluate. Visualizing these zones using AR offers an easy-to-handle solution for the safety engineer.

The planning and verification of safety components are to be connected to the developed software. An analysis of robot safety controls and their interfaces was carried out identifying those controls that are able to integrate safety zones. The controls are:

  • ABB SafeMove2

  • Denso Safety Motion

  • Fanuc Dual Check Safety

  • Kuka Kuka.SafeOperation

  • Stäubli CS9

  • Yaskawa/Motoman Functional Safety Unit

The safety controllers of the above-mentioned manufacturers were also evaluated if they can connect measuring or visualization devices. The controls of ABB and KUKA fulfill the requirements the most.

3.2 Development of the Planning, Communication and Verification Software

The overall approach for a software architecture is presented in Fig. 6. The concept illustrates the interaction of planning, communication and validation functions. After a detailed analysis the methods of a “line laser” and “laser projection” were evaluated as inappropriate.

Fig. 6
figure 6

Architecture of the software for planning, communication, and verification

Full size image

The developed tool converts the parametrized safety zones from the robot controller into mesh-geometry for further use. The converted objects are easily visualized via the AR-glasses. The desired and the actual state can be displayed.

To calibrate the AR-glasses within in the robot cell a marker and an assistant were developed. An iterative algorithm calculates the position of the glasses relative to the robot position based on predefined points. The accuracy of the calibration process increases with every calibration step. A stop criterion is derived that stops the iteration if the accuracy does not increase any further. This process is a state-of-the-art process for calibrating AR-glasses. For the given scenario several different marker positions were examined by the principle of trial-and-error.

Both simple measuring points and complex contours can be defined within the software. The measurement of these points and contours can be performed automatically using the software by direct control of the laser measuring device and the robot controller. After the measurement, the results are automatically summarized in a report. Figure 7 shows the user interface of the planning software.

Fig. 7
figure 7

User interface of the planning software

Full size image

In order to be able to move the robot to the defined points or to have the robot move along the defined contours, a system was introduced that does not require inverse kinematics. The system is based on a rough pre-positioning of the robot by the commissioning engineer. The pre-positioning can then be used to calculate the final poses for the defined points and contours. Subsequently the check points and their distances to the robot as well as any deviations between measured (in real cell) and calculated (in virtual cell) distances are summarized in a technical report.

3.3 Development of a Modular Measuring- and Visualization System

Prior to using the planning software two calibration processes are needed. First the operating point of the distance measuring device must be described within the robot basis coordinate system. Second a connection between the coordinate system and the AR-glasses must be established. This leads to three functional components that must be mounted on the robot:

  • A device for travel time measurement to perform measurements on the real robot cell.

  • A calibration marker to link the coordinate of the AR-glasses and the robot basis.

  • Two test prods to calibrate the robot tool.

The components are mounted on a plate with standardized hole pattern to mount it directly on the robot flange. The modular setup allows a case specific mounting of the components on the tool. Figure 8 shows the tool.

Fig. 8
figure 8

Modular robot tool

Full size image

The accuracy of this method for calibration was tested using an ABB IRB 4600. In case of the distance measurement device calibration, an angular error between 0.09 and 0.2 degree between the measured robot tool and the laser beam was measured. For more acute angles between laser beam and surface an increase of the angular error is observed. The error for the alignment of the light point (distance of 1000 mm, beam angle of 90° to the surface) is less than 4 mm in an unforeseeable direction. Therefore, check points must have a minimum distance to corners and edges to ensure that the desired object is hit. The precision of the test can be further increased by using a more accurate robot.

To calibrate the connection between robot basis and AR-glasses the calibration marker is used at different calibration positions. For every position the connection between both systems is known and therefore the pose of the robot coordinate system and the AR-glasses system can be derived. The highest accuracy was achieved, when the marker was spectated (by the AR-glasses) along an arc of 90°. This method allows a visualization with an accuracy of 5 mm between virtual safety zone from the DT and visualized safety zone in the augmented reality.

4 Summary and Conclusion

A new system for a safety analysis of robot cells was presented. Usually the process of safety analysis is very time consuming while commissioning new robot cells and has great potential for automation. A software was introduced that partly automated chooses points within the digital twin of a robot cell. The software then calculates poses for the robot in which the distance of the checkpoints from the virtual model can be measured in the real work cell. Within the performed tests an accuracy of 5 mm was achieved. However for greater distances the angular error further increases.

The increasing computing power of robot controllers allows to use more and more complex virtual safety zones within the safety setup. This makes it harder for the engineers to evaluate safety concept. Therefore, AR-Glasses were used to display virtual safety zones in the real cell. To calibrate the system, a calibration method to connect the coordinate system of the AR-Glasses to the system of the robot was developed using a calibration marker. This method allows a display of the holograms of the safety zones with an overall accuracy of 5 mm.

The presented system contributes to the aim of decreasing commissioning time and assists the safety engineer to evaluate more and more complex safety concepts.