Reaserch On Modeling, Control And Application Of Time-pressure Dispensing Process For IC Packaging

In IC packaging, time-pressure dispensing is a multivariable coupled system; the consistency of the adhesive amount dispensed is affected by dispensing time, pressure, temperature, gas compressibility, adhesive volume left in the syringe and the non-Newtonian dynamic of the fluid etc., which make it very difficult to modeling and control the process, and to improve dispensing performance. Starting with analyzing the factors affecting dispensing performance, an adhesive flow rate predictive model is developed and a vision-based online control strategy for time-dispensing is presented.A flow rate predictive model for dispensing was delivered. By analyzing the fluid dynamics, models based on typical power-law fluid and generalized power-law fluid were developed respectively to estimate the adhesive amount dispensed in some interval. Comparatively studying the models'precision by simulation with the software of FLUENT, especially studied the pressure effect of the tip entrance and exit on models precision. The method and condition used to calibrate model parameters are delivered. Both simulation and experiment showed that the model has enough predictive precision, the maximal error is not beyond 3% and the average is about 1.5%.Machine vision is used to estimate the adhesive dot size dispensed. Based on the idea of shape from shading (SFS), through developing the reflectance map of the adhesive dot, the dot surface orientation can be obtained by solving the reflectance map with linear iterative algorithm. After integral with the surface direction vectors, the dot height is obtained. Besides, with median filter and morphological method, the domain of the dot covering on the substrate is decided, and then dot volume can be estimate by quadraturing dot height in its domain. The estimated result is compared with the planar one to show that the measure error reduces from 3.76% for planar method to 1.09% for SFS method.A proportion- integral (PI) controller integrated with statistics process control (SPC) strategy is developed to control and compensate the vibrations of adhesive amount dispensed by adjusting dispensing time. In such process, the difference between the predictive fluid amount dispensed and the real one is looked on as output error. Simulations showed that the dispensing error is reduced from 43.6% with open loop control to 2.64% with the integrated control strategy developed in this document.Finally, based on the fluid flow rate model, a vision-based control method is used on a time-pressure dispensing system for online control. Experiments showed that the dispensing consistency has been significant improved compared with the condition of open control, the consistentcy error of the adhesive amount dispensed reduced from 31.11% to 4.44%, and the mean square error from 10.37% to 0.56%.