A quality prognostics scheme for semiconductor and TFT-LCD manufacturing processes

A quality prognostics scheme for predicting the product quality during the semiconductor or TFT-LCD manufacturing processes is proposed in this work. This scheme considers the current production equipment parameters sensed during the manufacturing process and several previous quality data collected from the measurement equipment to predict the quality of the product in the future. This quality prognostics scheme is composed of a conjecture model and a prediction model. The conjecture model itself can also be applied for the purpose of virtual metrology. With this two-tier arrangement, the quality prognostics scheme becomes more applicable. The algorithms adopted by the conjecture model can be various and exchangeable. Basically, numerous algorithms of artificial intelligence and statistics may be applied. The selection criteria are based on the physics of the equipment and the property of the measured parameters. The prediction algorithms used in the prediction model are weighted moving average, neural network, or any algorithm with prediction capability. Besides, the quality prognostics scheme possesses a self-searching mechanism and a self-adjusting mechanism. Upon the new conjecture or prediction algorithm being selected, the self-searching mechanism will be activated. The self-searching mechanism will not stop until the best combination of various parameters/functions used by the conjecture algorithm or prediction algorithm is found. Then, the conjecture or prediction model enters the normal running mode. After a period of time, if the prediction accuracy exceeds an acceptable bound or the equipment properties are altered due to scheduled maintenance or part change, the self-adjusting mechanism is launched to tune the system parameters to bring the prediction accuracy within the acceptable bounds.