The authors use bond graph (BG) modelling, a unified multi-energy domain method, to build dynamic models of process engineering systems; BG sub-models for various common process engineering devices are developed. The structural and causal properties of BG models are exploited for supervisory systems design. Controllability and observability are elegantly derived from BG models by following the simple algorithms presented here. The design and resource optimization of new supervision platforms is aided by analysing model structure to determine how easy it will be to detect and/or isolate some specific sub-space of the process. Real-time evaluation of system constraints followed by various decision steps integral to the supervisory environment are used to extract meaningful data from process state knowledge. Several applications to academic and small-scale-industrial processes are discussed and the development of a supervision platform for an industrial plant is presented with experimental validation.