Machining process automation is critical to improving productivity and unlocking novel design and manufacturing processes. High-accuracy automated drilling is fundamental to demanding assembly tolerances typically seen in Aerospace, Nuclear energy and Wind energy. High-accuracy automation in machining processes can be prohibitive from a cost, complexity and size perspective. If not solved, these technical limitations will become a bottleneck in advancing the field of high-accuracy drilling.
Typically, in industrial machining processes, the absolute accuracy of a machine is often
the focus of research, despite a machine’s repeatability outperforming its stated accuracy by an order of magnitude. The repeatability principle is exploited in high-precision metrology through comparator gauging or the substitution method to obtain precise measurements. The repeatability method can be applied to an industrial drilling process with supporting fixturing to achieve a high-accuracy drilling process by taking advantage of the machine’s repeatability. This thesis introduces a novel high-accuracy drilling process that significantly reduces deviation errors, in the X − axis by approximately 1217% and in the Y − axis by approximately 1488%, representing a major advancement in the field.
A systematic process is developed and presented to achieve a high-accuracy drilling process using the repeatability of a machine. A template, workpiece, backing plate and fixturing were designed to integrate with the machine system and the Coordinate Measurement Machine (CMM). Key features were integrated within the overall setup to ensure the integrity of the experimental setup, repeatable drilling process and measurement process. A repeatable probing process using the machine’s On Machine Probe (OMP) and a template sized at 0.5m × 0.5m was performed to assess the machine’s repeatability. The result provided experimental validation that the machine is repeatable to a high degree. The High Accuracy Positioning (HAP) drilling process was then developed using a template, backing plate, workpiece and fixturing. The HAP drilling process produced relatively high positional accuracy holes compared to literature. A conventional drilling process was also performed on the same machine to compare with the repeatable drilling results and validate the theory.
The HAP process is tested and validated for a large-scale template and workpiece sized
at 3m × 1m. A single-position repeatability experiment was performed to characterise the
machine’s repeatability over an extended period in a large working area. The HAP drilling
experimental trials at a large scale were then performed. This process included both dry and Tool Change (TC) drilling. The TC drilling process consisted of cutter changes to include seven hole diameters thereby requiring seven tool change operations. A correlation matrix was produced for the HAP dry drilling and the TC drilling processes. The HAP drilling process achieved significantly lower deviation errors than any results seen in the literature for a process at this scale.
Artificial Intelligence techniques are used to develop a capability to predict drilling accuracy based on probing data. The large-scale, repeatable probing and drilling data are
pre-processed to build a pipeline for the regression models. The data was split into input
parameters and output required. The features from the repeatable probing data are used
as the input parameters to predict the output from the repeatable drilling process, namely the deviation errors in the X − axis and Y − axis. Four regression models are used - linear regression, random forest, gradient boosting and extra trees. The models were executed in their native format to provide a baseline evaluation and with MinMaxScaler to evaluate the influence of data normalisation on the model’s performance. Following this, the hyperparameters of the two best-performing models were optimised. The models were cross-validated to ensure generalisability to an independent data set.
