RAMLAB – Removal of Excess Metal

Students: Alexander Ramp, Wouter van Leeuwen, Guus Figdor, Willem van der LindenSjoerd Kaptein & Marc Groeneveld. (Spring 2017)



“Industrial spare parts should always be available wherever they’re needed, whenever they’re needed.” A phrase in which RAMLAB sees additive manufacturing as potential to enable a wide scale availability of on-demand metal parts.

At the moment RAMLAB is researching the use of wire-arc additive manufacturing to make parts for the maritime industry. Consequently, the surface of the printed parts is too rough (as the image below illustrated) and needs to be grinded and smoothened until within product tolerances. Which done manually, is a time-consuming process.

Therefore, RAMLAB approached the Smart Manufacturing and Robotics minor in Delft to research the possibility and develop a robotic grinding process, starting with simple flat shaped objects. Besides grinding, the robotic process also must measure and validate the part’s deviations.


The objective

As the above introduction somewhat indicated, the set objective stated the following;

“Within 8 weeks a robotic grinding and measuring method for finishing a wire-arc-printed product must be researched, developed and demonstrated. Capable of finishing the product within 1 millimetre tolerance.”

To accomplish the set objective, the following main tasks were set:

  • Develop software to automatically generate robot toolpaths for different shapes and parts.
  • Develop an end-of-arm-tool with grinding and measuring capabilities.
  • Develop software to map measured deviations.
  • Create a user interface to input the parameters of a printed parts.

 The result

After 8 weeks, an Universal Robot (UR10) is capable of grinding a simple shaped object within 1 millimetre tolerance, this along with a predetermined grinding sequence and pattern, generated based on given parameters of an object.

User interface
These parameters of an object are manually set in the User Interface, which is the face for a communication program running on a raspberry pi. The parameters are dimensions and can be divided into the following categories:

  • Calibration point; the length (x) and width (y) from a zero-point to the unfinished object.
  • Object dimensions; the length (x), width (y) and height (z) of the object.
  • Product tolerance; the preferred tolerance length (x) and width (y).

When confirmed, the given data is sent to UR10 to generate a customized toolpath. The communication between both devices is done over UTP.

Toolpath generating
The toolpath is generated in URScript (software of the UR10) which divides the determined movement pattern into steps of 10 millimeters over the length and width of an object. The movement pattern is an iteration of the grinding tool’s down-up motion depending on the object’s height. Moreover, after completion of a full surface run, the movement pattern moves object-inwards with steps of 0,5 millimeters until the tolerance dimension is met.

Tolerance inspection
Due to abrasive removing of material, it can’t be assumed that the object is within tolerance when the tolerance dimension of the toolpath is reached. Therefore, the end-of-arm tool is equipped with an optical distance sensor (ODS) to inspect surface deviations with respect to the set tolerance.

The ODS measures the distance in a 0-10 voltage range and is calibrated to a workspace of 20 millimetres, what results in a theoretical accuracy of approximately 0,01 millimetres. The inspection is done with point measurements along the whole surface, covering 75 square millimetres per point (during tests).

During an inspection, the measurements and coordinates are sent to the raspberry pi, which virtually plots the distances along the width and length of the object. In continuation of this project, this information can be used to determine where the surface needs more material removal. Below such a virtual measurements plot is shown.


Concluding this project, the developed robotic grinding method is suitable to finish simple wire-arc printed parts, meeting all set client requirements. In comparison with milling (researched by Autodesk), this method will be less expensive and create a more smooth (mirror-like) surface. Being ideal for maritime products such as bows and propellers.

Unfortunately, the mirror-like surface creates too much noise for an ODS to make a reliable tolerance inspection. In order to prevent this, the surface roughness must be increased, possible with the use of a coarser abrasive disk, faster up-down motion or slower grinding speed.

Furthermore, a feedback loop based on sensor information to the toolpath-generator still needs to be realised. Along with force/pressure controlling software to optimise the performance of the abrasive disks. However, this was all out of scope.

As a result of this project, the possibility/potential is proven, though further research, development and integration are still necessary to be implementable for finishing of complex shapes.