Adaptive real-time networks design and simulation based on IC wedge-bonding control system

Aimed at the real-time requirement of the Ethernet in the IC wedge-bonding control system, a communication system with a neural-network-based adjustment module was presented. This new system consisted of a managing node and many control nodes. In order to reduce the real-time decline resulting from channel conflict, every control node was assigned a default priority and the control nodes with the same priority fell into one group. Control nodes communicated among themselves and with the managing node based on the priority groups they belonged to. An adjustment module based on the continuous Hopfield neural network (CHNN) was employed in the managing node. The status data gathered from the control nodes were counted as inputs of the module; thus, through the self-learning function of the module, the priorities of these control nodes could have dynamic adjustments. The adjustment module distributed channel resources by way of dynamic priorities, ensuring not only that there was less communication conflict among nodes at a certain moment, but also that adaptive adjustments of priorities could be achieved to improve transmission efficiency. The analysis and the simulation results demonstrate that the method is correct and efficient.