The use of AI and Big Data in the production of advanced vertical guide plates

Tuohy, Simon, McDonnell, Michael, David Tom, Ferguson, Alan, Praeger, Matthew, Grant-Jacob, James, Mills, Benjamin and Karnakis, Dimitris (2022) The use of AI and Big Data in the production of advanced vertical guide plates. In SWTest 2022. (In Press)

Abstract

Two extremely powerful fields of Computer Science are harnessed, namely Artificial Intelligence (AI) and Big Data to better improve the manufacturing accuracy of advanced vertical guide plates.
With the continual advances in semiconductor chip design, there is an increasing demand for higher performing Probe Cards with a reduced time to market, this in turn puts additional manufacturing and/or performance demands on the guide plates within the probe heads in these devices.
Using Big Data techniques, we can discover trends and insights that only become apparent when an analysis of millions of laser drilled holes across multiple laser tools is carried out. Before Big Data techniques can be utilised however, preparations for the storage and retrieval of vast quantities of data is required. At Oxford Lasers, such a custom built data warehouse has been created. A simple example of the power of data will be shown, where laser drilled hole position accuracy can be improved, resulting in extremely tight positional tolerances within a given DUT array.

Laser drilling the “perfect guide plate hole” is possible but requires laborious process optimisation over typically twenty hole measurement metrics. This usually involves finding a global optimum across fifteen different input parameters, where each of these inputs can range between fifty different input values. Many of these conditions are not independent and influence each other complicating the search. Therefore, the engineer tasked with finding the optimal process recipe for guide plate laser drilling has a potential 1X1025 different combinations that she could try which is not practical.
Given the size of the parameter search space, Oxford Lasers and the University of Southampton have collaborated on a project to develop AI, specifically Neural Networks, and investigate whether such Neural Networks can virtualise this drilling recipe search. Very encouraging findings will be reported in this presentation.
This has allowed us to gain insights previously undiscovered by humans, to find better laser drilling process recipes whilst reducing development cycle time and improving the time to market of such guide plates for production.

This record has no associated files available for download.

Contributors

Author: Michael, David Tom McDonnell

Author: Matthew Praeger

Author: James Grant-Jacob

Author: Benjamin Mills

Download statistics