Accidents Control and Safety Management of Information Systems. N. I. Borodianskii and V. A. Lysenko, V. Glushkov Institute of Cybernetics Ac. Sc. of Ukraine, Glushkov Avenue 40, 25220 7, Kiev 20 7, Ukraine
It is common knowledge that every technological and organizational system can go out of order. However for complex system it is very difficult to determine the causes and the nature of possible failures. Therefore it is hard to propose certain control procedures allowing to avoid all types of failures. Nevertheless we can analyze and estimate probability of certain classes of failures within definite systems. This brings an opportunity for preliminary risk minimization while designing the system by the way of main failure causes and consequences description. The mathematical approach for construction of classification function of failure is suggested. This allows to recognize the most probable failures and accidents with the hardest consequences. This analysis is minimizing the deadlocks' probability, i.e., in case of extreme danger of system failure, to control it in such a way that to avoid irreversible consequences. This technique is based on methods of mathematical logic and mathematical modeling. It operates with new statistical criteria for detection and definition of preaccidental states of the system which allow to transfer the system into the state either controlled by certain save procedures or, in case the parameters are not recognized, demanding the conservation in the last condition aimed at expectation of dispatcher interference for determination of new procedures. This article considers information systems in different applications as a system class element. Saving of certain information is one of the main aims of failure control in such systems. The subject of saving should be as users' information so the series of parameters bearing specific system information. The other goal of failure control is system adaptation to new state under which the consequences are not so serious. As an example we consider such known failures as a number of health management information systems.