Most of current structural management systems make decisions based on previous maintenance records. In these systems Markov evaluation with transition probabilities between different condition states are used to predict future performance of deteriorating systems. However, the values of transition probabilities have to be modified according to material and structural differences. As a result, the same deterioration transition rule cannot be used for different types of structures. Besides the transition probabilities are very subjective when it is combined with condition states based on unreliable assessment methods such as visual inspection. Recognizing this problem, a new method has been developed to construct more objective performance models for steel box bridges. The new method also gives enough flexibility and extendibility for the next generation of BMS. Based on this newperformance regression model, a bridge management system has been developed also.