Traditional trial-and-error springback compensation methods have the problems of high scrap rate, low efficiency, high cost of fixtures and operator experience dependency. The method presented in this paper uses on-line measured springback data from the same batch to predict and compensate for springback. Because there are no springback data for the first bend, bend-rebend control is used to make the first bend to eliminate trial tubes. In addition to springback, relaxation and radial growth are also estimated and compensated for to make a bend more accurate. A process control method is developed to optimize the overall control strategy such that the overall tube error is minimized without increasing the required hardware accuracy. The optimal process control strategy has significantly higher accuracy than the traditional trial-and-error method. The details of statistical analysis of tube tolerance and adaptive bend correction algorithm are presented in part 2 of the paper.