Maintaining minimal levels of geometric error in the finished workpiece is of increasing importance in the modern production environment; there is considerable research on the identification, verification and calibration of machine tool kinematic error, and the development of Postprocessor implementations to generate NC-code optimised for machining accuracy. The choice of multi-axis positioning function at the controller, however, is an often-overlooked potential source of kinematic error which can be responsible for costly mistakes in the production environment. This paper presents an investigation into how mis-management of the positional error parameters that define the rotary-axes’ pivot point can lead to unintended variations in multi-axis positioning. Four approaches for kinematic positioning on a Fanuc-based controller are considered, which reference two separate parameter locations to define the pivot point – managing the kinematics within the Postprocessor itself, full five-axis positioning with a fixture offset, full five-axis with rotation tool centre point control and 3+2-axis with a tilted workplane. Error vectors across four sets of rotary-axis indexations are simulated based on the theoretical kinematic model, to highlight the expected differences in geometric error attributable to mismatched pivot point parameters. Finally, the simulation results are verified experimentally, demonstrating the importance of maintaining a consistent approach in both programming and operation environments.

Maintaining high levels of geometric accuracy in five-axis machining centres is of critical importance to many industries and applications. Numerous methods for error identification have been developed in both the academic and industrial fields; one commonly-applied technique is artefact probing, which can reveal inherent system errors at minimal cost and does not require high skill levels to perform. The primary focus of popular commercial solutions is on confirming machine capability to produce accurate workpieces, with the potential for short-term trend analysis and fault diagnosis through interpretation of the results by an experienced user. This paper considers expanding the artefact probing method into a performance monitoring system, benefitting both the onsite Maintenance Engineer and visiting specialist Engineer with accessibility of information and more effective means to form insight. A technique for constructing a data-driven tolerance threshold is introduced, describing the normal operating condition and helping protect against unwarranted settings induced by human error. A multifunctional graphical element is developed to present the data trends with tolerance threshold integration to maintain relevant performance context, and an automated event detector to highlight areas of interest or concern. The methods were developed on a simulated, demonstration dataset; then applied without modification to three case studies on data acquired from currently operating industrial machining centres to verify the methods. The data-driven tolerance threshold and event detector methods were shown to be effective at their respective tasks, and the merits of the multifunctional graphical display are presented and discussed.

No abstract

No abstract

NC-Checker is a software tool used for monitoring and validating the geometric performance in modern machining centres. Threshold settings allow the Manufacturing or Maintenance Engineer to customise the tool based on specific job or industry tolerance requirements. In order to perform effective long-term monitoring, this has the potential to skew the perceived health state of the machining centre as presented in the NC-Checker benchmark reports. This study brings attention to this fact and its relevance in the pursuit of enhanced levels of automation for geometric performance monitoring tools, in preparation for the machine shop's transition to Industry 4.0. A sense-check function is proposed to identify unusual alterations based on historical data, utilising a support vector machine methodology to develop a predictive classifier. The models achieved predictive accuracy scores of 87.5% during validation, acquisition of a suitable testing set is under way and the predictive models will be evaluated upon completion.