This paper describes derivation of an operator advisory system (OAS) for detecting agglomeration based on the multivariate signal analysis of an industrial scale fluidized bed. Agglomeration is an intermittent concern in many fluidized bed processes, which may lead to production upsets, including problems with product quality, blockage of the product withdrawal lines, and even complete defluidization of the bed. There is a large incentive then, both in terms of lost production and clean up costs, to maintain a stable operating regime and avoid any major process upsets. Unfortunately, process operators cannot actually see agglomeration within the fluidized bed. The OAS is capable of distinguishing four levels of operational stability: normal, abnormal, dangerous, and critical in relation to the production rate of agglomerates within the fluidized bed. The system described monitors hundreds of process signals on-line and in real time with an overall average success rate of 90% (improvement over chance). (c) 2007 Elsevier B.V. All rights reserved.