Machine vision for in-process inspection on an automated peptide manufacturing platform

Abstract

Automated inspection of a manufacturing line utilizing machine vision is a powerful tool used to increase efficiency, reduce variation, and maintain a high-quality standard in modern manufacturing systems. The aim of this thesis is to develop and implement a machine vision system for Mytide Therapeutics' automated peptide manufacturing platform. Throughout the peptide manufacturing process, vials are used for liquid and solid handling of the peptide compounds through the disparate process steps. To ensure the vials of peptides are transferred between these steps properly, several functions were developed to analyze the images at key points in the process. These images are analyzed to ensure a vial is present when required, is gripped properly, and is transferred successfully by the robot to and from each station. Moreover, a function to estimate the volume of resin inside of a vial before and after peptide synthesis is developed as a method of collecting data relevant to the process that would not be collected otherwise. The proposed algorithms are designed to provide key insights to the manufacturing process, ensure the process runs smoothly, reduce overall system downtime, and inform future system improvements. This thesis presents an overview of the machine vision system, details about the algorithms developed to perform the image analysis, and the methods for implementation of the image analysis functions into the manufacturing platform.