An efficient decomposition algorithm for a workflow model based on Petri net invariants is presented in this paper. The algorithm is a new kind of simplification strategy that decomposes a complex and sound Workflow net (WF-net) to a class of simple subnets that are able to describe the business cases. The existence of T-invariants and the coverage sets of WF-net transitions are analyzed and verified in detail. The advantages of the algorithm include its complexity, and avoidance of state space explosion. It can overcome the shortage of the methods based on a depth traversal algorithm. Furthermore, it is readily comprehensible, and can be extended easily to allow parallel processing. The usability of the research results is illustrated by an example.