ASMPT Advances Precision and Throughput of Semiconductor and Electronics Manufacturing Machines
Software-Driven Motion Control Delivers Precision and Flexibility for Next-Gen Semiconductor Equipment
“The Nonlinear ARX models we created in MATLAB allowed us to deliver accurate motion control of the bonding machines used in semiconductor and electronics manufacturing. Such precision gives us a strong technological advantage in a competitive market.”
Key Outcomes
- Achieved at least 30% reduction in tracking errors using ILC in MATLAB to compute high-precision feedforward control signals
- Expanded the range of accurately executed motion trajectories, improving precision and flexibility through AI-based nonlinear system identification models developed with System Identification Toolbox™
- Enabled cost-effective motion control by replacing expensive hardware modifications with software-based solutions built in MATLAB and Simulink®
ASMPT uses next-generation machines such as the AERO PRO Wire Bonder to create interconnects in semiconductor circuits. The machines must implement tight motion control while also maintaining high throughput of finished products.
ASMPT is a global provider of hardware and software solutions for semiconductor and electronics manufacturing. The company delivers advanced semiconductor packaging and electronics assembly equipment and software technologies that enable high-performance production across diverse applications.
To maintain leadership in a rapidly evolving semiconductor landscape, ASMPT needed to enhance the precision and flexibility of its bonding machines without incurring the cost and complexity of hardware redesigns. The goal was to improve motion control accuracy while supporting a wider range of manufacturing trajectories.
ASMPT used iterative learning control (ILC) techniques in MATLAB® to build nonlinear regression models that generate high-precision feedforward control signals. These models were integrated into the motion control systems of ASMPT’s bonding machines, enabling software-driven enhancements that outperform traditional hardware-based approaches.
To prevent repeated retuning of the ILC for different end-effector trajectories, ASMPT collaborated with MathWorks to build and apply nonlinear autoregressive exogenous models that generalize across multiple trajectories. This approach eliminates the need for retuning ILC per trajectory, enabling faster deployment and broader applicability without compromising precision.
Related Resources
