Data-Driven Robust Design for a Curing Oven

Abstract

If no external control is employed, the cure performance will be fully dependent on the oven design. Design of such a system is challenging because of its inherent complexity: complex heat transfer, unknown relationship between design variable and curing performance, unknown boundary conditions and large parameter variation, and so on. In this paper, a novel data-based robust design approach is proposed to design a new curing oven with simple structure. First, a modeling method with hybrid singular value decomposition and response surface method is developed to extract the system model from experimental data. The coupling influence between design variable and model parameter is also separated to make the design simpler, because only the part affected by design variable will be considered in the design. A robust design method is then developed to minimize the influence of parameter variation on the performance. Finally, the effectiveness of the proposed method is demonstrated in the design of the new curing oven.Index Terms-Design for manufacture, design methodology, semiconductor device manufacture, semiconductor device modeling.

A KL/Subspace Modeling Method for Integrated Design and Control in Curing Processes

“…A data-driven modeling method is proposed for the curing process with the hybrid design/control variables, as indicated in Figure . It includes the following key points. Variable separation: Inspired from some separation methods, ,,, the original system is first separated into the design-variable-dependent basis function and the control-variable-dependent temporal model by the KL method. This decomposes the original modeling task into two simple subtasks: modeling respectively for the design variable and the control variable.…”

“…The ratio of the sum of the n largest eigenvalues to the total sum is where the maximal number of nonzero eigenvalues is K = min­( N , L ). Usually, the sufficient number of eigenfunctions that capture 99% of the system “energy” is used to determine the value of n . ,, …”

“…The curing oven as presented in refs ,, is shown in Figure . Its heaters, which are embedded in the heater block, first transfer heat to the heating block, and then the heating block heats the integrated circuit (IC) placed on the lead frame (LF).…”

“…Such a typical example is a curing oven in the semiconductor packaging industry. It is used to provide a required temperature distribution for setting compounds that are introduced onto electronic components. ,, Its curing performance, integrated design/control performance, , is fully dependent on its design variables, such as shape and size of the oven, and its control variable, such as power of heaters. On the one hand, the design variables are a discrete variable that is only able to take several discrete values due to the limitation of both cost and the manufacturing accuracy of the machines.…”

“…On the other hand, the control variable is a continuous variable, which varies continuously with time and may take an arbitrary value within its bounds. In the design and control of the curing process, there are many first-principle modeling methods for both the design variables , and the control variables. , However, because all of these first-principle models are usually built under many simplifications and assumptions, they are inaccurate due to unknown dynamics and unknown boundary conditions and complexities of the curing process, because it includes conduction, convection, and radiation. , However, data-driven modeling problems in complex curing ovens are usually depicted to obtain a model that contains either discrete design variables or continuous control variables, ,, while modeling them together has never been considered. This leads to difficulty in obtaining the optimal curing performance by the integration of design and control.…”