As a large-scale control system with strategic position and deterrent effect in China, the rapid and accurate judgment and repair of its faults plays an important role in ensuring that the electromechanical device remains stable and points to the target accurately and quickly. At present, most of the methods for diagnosing the system are to extract features from its analog data and identify fault types, but a large number of unstructured data such as maintenance logs and diagnostic reports of the system itself have not been effectively applied. Based on this, this paper develops the system construction of fault diagnosis for large electromechanical control system based on knowledge map. Through feature extraction and data migration, the data layer and mode layer are constructed by drawing lessons from the topology of ART2 neural network, and on this basis, the entity attribute association is realized, and the fault diagnosis model based on knowledge map is constructed, which effectively reduces manual participation and experience retrieval in fault diagnosis and improves the efficiency of fault diagnosis.