This paper concerns the use of static analysis for debugging purposes of declarative object-oriented equation-based modeling languages. We propose a framework where over- and under-constraining situations present in simulation models specified in such languages are detected by combinatorial graph transformations performed on the flattened intermediate code and filtered by the semantic transformation rules derived from the original language. This is powerful enough to statically detect a broad range of errors without having to execute the simulation model. Debuggers associated with simulation environments for such languages can provide efficient error-fixing strategies based on the graph-based representation of the intermediate code. The emphasis, in this paper, is on detecting and debugging over-constraining equations, which are present in some simulation model specifications. We discuss various ways in which we have extended our approach to allow static global analysis of the original modeling source code.