Integrating supervised and unsupervised learning approaches to unveil critical process inputs

TL;DR

A machine learning framework tailored to large-scale industrial processes characterized by a plethora of numerical and categorical inputs that addresses the pivotal question of the significance of each input in shaping the process outcome, using an industrial Chemical Vapor Deposition process as an example.

Abstract

This study introduces a machine learning framework tailored to large-scale industrial processes characterized by a plethora of numerical and categorical inputs. The framework aims to (i) discern critical parameters influencing the output and (ii) generate accurate out-of-sample qualitative and quantitative predictions of production outcomes. Specifically, we address the pivotal question of the significance of each input in shaping the process outcome, using an industrial Chemical Vapor Deposition (CVD) process as an example. The initial objective involves merging subject matter expertise and clustering techniques exclusively on the process output, here, coating thickness measurements at various positions in the reactor. This approach identifies groups of production runs that share similar qualitative characteristics, such as film mean thickness and standard deviation. In particular, the differences of the outcomes represented by the different clusters can be attributed to differences in specific inputs, indicating that these inputs are critical for the production outcome. Leveraging this insight, we subsequently implement supervised classification and regression methods using the identified critical process inputs. The proposed methodology proves to be valuable in scenarios with a multitude of inputs and insufficient data for the direct application of deep learning techniques, providing meaningful insights into the underlying processes.

Figures (8)

• Figure 1: (a) Examples of the coated cutting tool inserts. (b) A 3D representation of a 3-disk part of the reactor. In red: inlet perforations on the rotating inlet tube. In blue: outlet perforations for each disk.
• Figure 2: Most common measurement positions among the production data. These measurements can be used for several tasks, such as the development of CFD or ML approaches for the prediction of the process outcome.
• Figure 3: (a) Resulting dendrogram of the clusters output by the implemented agglomerative hierarchical clustering algorithm using a Ward linkage criterion. The three main clusters of interest are colored purple, red, and green. We note that by selecting a slightly higher dissimilarity threshold, the red and green clusters can be merged and viewed as a larger cluster (shown in blue). (b) The three resulting clusters, visualized in a reduced 3D space. The three clusters appear to be well-formed. PCA was used for finding the 3D reduced space.
• Figure 4: Thickness distribution in the case of: a) 2 clusters and b) 3 clusters. High average thickness and low standard deviation is a measure of process efficiency and product quality. The production runs in the "purple" cluster demonstrate superior quality characteristics.
• Figure 5: Distributions of |Nominal recipe surface area - actual surface area| for clusters 1 (in green) and 2 (in red). Cluster 2 includes relatively more observations with values larger than 5000 cm2 when compared with cluster 1.