Automation takes many forms, but in this article, I refer to the use of industrial (serial axis) robots and Cartesian CNC machine tools. These systems can comprise depositing material – e.g. automated tape laying (ATL) or automated fibre placement (AFP), post processing the layup – e.g. edge of part trimming or drilling, or any number of other processes such as forming, pinning or bonding.
These automated systems are often complex and expensive installations required to ensure the consistency and quality levels demanded by high value manufacturing industries such as aerospace. It is difficult to quantify the performance of these automated systems, as accuracy statements are sometimes woolly or ambiguous. However, accurate performance is essential if any level of offline programming is expected. The translation from offline programming/simulation environment to the physical manufacture must be tightly controlled to ensure design intentions are fully met. Digital twinning is only really successful when the actual system is sufficiently close to its perfect digital sibling.
It is common for design tolerances to be applied only as the ‘bottom-line’ for finished parts and without consideration for manufacturing tolerances at interim stages. However, without such tolerances, it is difficult to objectively assess a system’s capability in relation to the requirements. Often system performance is stated – if stated at all – as a tolerance stack of the individual system elements such as a robot performance value plus an end-effector manufacturing tolerance plus a mandrel manufacturing tolerance. It is unlikely such models are valid; complex electro-mechanical systems have subtle interactions between tolerances that are difficult to model or are overlooked completely. For example, the stiffness of a robot with a heavy payload is difficult to quantify and compensate for (although some robots do compensate to some level) with any certainty that it is in line with the desired performance.
A more reliable approach is to measure the process directly with an independent metrology system. This circumnavigates the requirement to model the interaction of each component and subsequently calculates the tolerance stack-up. External metrology such as laser trackers and photogrammetry systems can provide dynamic measurement of automation systems. These systems dynamically measure (or track) in six degrees of freedom, the position of the automation relative to the component to better understand their interactions. The data can subsequently be correlated to the simulation model or offline program to quantify discrepancies. System performance can be assessed through all or part of the intended working volume, giving rise to an accuracy statement tailored to the use case. Once established, the system accuracy can inform the simulations and model, as well as the design and manufacturing decisions.
If higher levels of positional accuracy are required, there are a number of strategies for improving the performance of automation systems. There are two main generic approaches System Calibration or Closed Loop Control.
System Calibration can be achieved by fitting measured data to a kinematic model of the automation system (robot or Cartesian positioner). This calibration data can reduce the mechanical inaccuracies of the system by no longer using nominal values such as axis length or angular alignment. In addition, linear and angular encoders can be individually calibrated. Often the most challenging element of implementing a calibration is applying the new kinematic model to the control system. This varies between manufacturers, but the introduction of standard machine tool controllers (e.g. Siemens 840D) is improving the ability to apply calibration parameters or error maps. System calibration can be simpler than measurement of the kinematics, for example, an accurately positioned tool or mandrel will very often significantly reduce exhibited errors.
Closed Loop Control uses a dedicated metrology resource to monitor and control the automation in real-time. This has several challenges, many of which are unique to the automation and controller. For example, the measurement system requires line of sight which is dependent on the application and automation configuration. Additionally, the measurement/corrective path data needs to be received by the controller with an acceptably short latency; this is generally achievable for material deposition. However, any rapid material removable process will require very low levels of latency which may not be achievable.
Understanding how automated processes compare to the simulated path is of paramount importance for manufacturing confidence. Once firm data has been collected, tolerances can be defined, and actions can be determined if necessary. Companies such as INSPHERE have the expertise to help measure and correct complex automated systems to achieve confidence and ensure that simulation matches reality.