Reliable Manufacturing of Heavy Copper Wire Bonds Using Online Parameter Adaptation

Usage of copper wire bonds allows to push power boundaries imposed by aluminum wire bonds. Copper allows higher electrical, thermal and mechanical loads than aluminum, which currently is the most commonly used material in heavy wire bonding. This is the main driving factor for increased usage of copper in high power applications such as wind turbines, locomotives or electric vehicles. At the same time, usage of copper also increases tool wear and reduces the range of parameter values for a stable process, making the process more challenging. To overcome these drawbacks, parameter adaptation at runtime using self-optimization is desired. A self-optimizing system is based on system objectives that evaluate and quantify system performance. System parameters can be changed at runtime such that pre-selected objective values are reached. For adaptation of bond process parameters, model-based self-optimization is employed. Since it is based on a model of the system, the bond process was modeled. In addition to static model parameters such as wire and substrate material properties and vibration characteristics of transducer and tool, variable model inputs are process parameters. Main simulation result is bonded area in the wire-substrate contact. This model is then used to find valid and optimal working points before operation. The working point is composed of normal force and ultrasonic voltage trajectories, which are usually determined experimentally. Instead, multiobjective optimalization is used to compute trajectories that simultaneously optimize bond quality, process duration, tool wear and probability of tool-substrate contacts. The values of these objectives are computed using the process model. At runtime, selection among pre-determined optimal working points is sufficient to prioritize individual objectives. This way, the computationally expensive process of numerically solving a multiobjective optimal control problem and the demanding high speed bonding process are separated. To evaluate to what extent the pre-defined goals of self-optimization are met, an off-the-shelf heavy wire bonding machine was modified to allow for parameter adaptation and for transmitting of measurement data at runtime. This data is received by an external computer system and evaluated to select a new working point. Then, new process parameters are sent to the modified bonding machine for use for subsequent bonds. With these components, a full self-optimizing system has been implemented.