In this paper we describe a tools environment which automates the validation and maintenance of a requirements model written in many-sorted first order logic. We focus on: a translator, that produces an executable form of the model; blame assignment functions, which input batches of mis-classified tests (i.e. training examples) and output likely faulty parts of the model; and a theory reviser, which inputs the faulty parts and examples and outputs suggested revisions to the model. In particular, we concentrate on the problems encountered when applying these tools to a real application: a requirements model containing air traffic control separation standards, operating methods and airspace information.