Internetware system are widely used in the chemical and pharmaceutical industry, effective acquisition of transition probability matrix is directly related to internetware reliability computation based on Markov chain model. The construction of Markov chain and the acquisition of transition probability of internetware are studied; the Markov chain of internetware reliability is constructed. Transition probability matrix computation based on the smallest quadratic difference is presented by using the occupancy of component executing the transition as the sample statistics to calculate transition probability. The approximation algorithm based on projection gradient is put forward, and it effectively guarantees the transition law of Markov chain and the characteristics of transition probability matrix. The experiment proves that the presented method and the designed algorithm can effectively compute transition probability with great value in internetware reliability computation