Digital twins are the key to success for deep learning (DL) projects – especially DL projects that involve processes that are dangerous, expensive, or time-consuming. A digital twin is a digital replica of an actual physical process, system, or device that can be used in simulations.
Digital twins are used in simulations to create the right amount of the right kind of simulation data to train DL networks successfully. There are several reasons to use digital twins in simulations to create DL training data:
DL employs neural networks to perform advanced pattern-matching. In semiconductor manufacturing, DL has already been applied in areas such as defect classification and anomaly detection. As companies start to explore DL and how it can help them, many are finding two things: first it’s easy to get to a prototype; and second, it’s harder to get from “good prototype” results to “production quality” results. Why?
DL applications are “programmed” by presenting neural networks with data that represent a target to matched. Masses of data train the network to recognize the target (and to know when it’s not the target).
The predictive ability of any DL application depends on the depth and breadth of the data set used in training. If the training data set is too small, too narrow, or too “normal,” it will not be more predictive than standard techniques. It’s important to train networks with masses of data representing every possible state or presentation in equal volumes.
The difficulty for some fields, such as autonomous driving, or semiconductor manufacturing, is that some of the most serious anomalous conditions occur very rarely. But, if you want a DL application to recognize a child darting in front of a car – or a fatal mask error – you have to train the networks with a multitude of these scenarios…which don’t exist in any great volume in the real world. Simulation is the only way to create enough anomalous data to properly train the networks to recognize these conditions.
At the 2019 SPIE Photomask Technology conference, D2S presented a paper demonstrating the creation of two digital twins – a SEM digital twin, and a curvilinear ILT digital twin – using DL techniques. These digital twins have been used both for DL training and validation.
Linyong Pang; Suhas Pillai; Thang Nguyen; Mike Meyer; Ajay Baranwal; Henry Yu; Mariusz Niewczas; Ryan Pearman; Abhishek Shendre; Aki Fujimura, “Making digital twins using the Deep Learning Kit (DLK),” Proc. SPIE 11148, Photomask Technology 2019, 111480B (21 October 2019); doi: https://doi.org/10.1117/12.2538508
If you’ve already started down the road with your DL projects but have encountered issues due to the DL data gap, D2S can help you build the digital twins you need to augment and tune your data sets for DL success.
If you’re starting to gather your resources for DL, D2S offers TrueMask® DLK, a deep learning kit that includes GPU computing, modeling, simulation, and validation technologies, as well as neural networks trained for semiconductor manufacturing, and is available with some digital twins, such as SEM and inspection digital twins.
If you’re new to DL and would like comprehensive support for your pilot DL project(s), you can join the Center for Deep Learning in Electronics Manufacturing (CDLe) www.cdle.ai.