The implementation of the methodology takes the bottom path of Fig. 1. For this implementation, the design requirements are fed into a control engine, which uses a library of control algorithms and validated plant models to arrive at the control design using an iterative optimization process. The control design is then implemented using validated control architectures, which are tested automatically to guarantee that the reliability requirements are met. Finally, during the lifetime of the plant, the plant model is maintained up to date (e.g, updated with component failures or mode changes) by an on-line diagnostics system. 3. CONTROL ENGINE IMPLEMENTATION The automation of this design methodology requires the following steps: 3.1 Selection of Design Requirements Related to Control System Performance The first step is to review the design requirements for applicability to the problem at hand. The automated design system will deal only with those design requirements that relate directly to the "performance" of the control system. The discarded requirements will have to be addressed separately in a manual fashion. An example requirement is to specify that all the equipment must operate using a power supply of 110 Volts and 60 Hz. This is clearly a valid requirement, which must be specified when purchasing the control hardware, but is not directly related to control performance; thus, these types of requirements are not included in the automated control engine requirement set. An example of a valid requirement is to specify that the control system must control the reactor without scram for a particular transient. For example, we could specify that a Design Requirements