Abstract: Operators of industrial processes face the complexity of the process usually. They have to set all appropriate parameters correctly to ensure requested performance and quality of production. This demanding task is additionally complicated by permanent changing of ambient conditions. A probabilistic advisory system was developed as a support tool for operators. The advisory system processes historical data and concentrates the information about process behavior into a mixture of probability density functions. With the help of bayesian statistics, the advisory system compares historical information to the actual working point of the process and generates advisory information on how to change current process settings in order to reach requested performance and quality of production. Large mathematical and computational potential of developed advisory system encouraged an extension of the system by advanced diagnostic functions. The aim of this diagnostics is to recognize sensor and signal malfunctions that cannot be detected by standard diagnostic tools.