The unstructured historical data available in the databases of Computerized Maintenance Management Systems represents a wealth of diagnostic knowledge. In this paper, a methodology for converting the maintenance log data into formal knowledge graphs is presented. The methodology uses text analytics techniques, in combination with human-assisted thesaurus development methods, for generating a formal thesaurus, or knowledge graph, that encodes the semantic relationships between multiple maintenance entities. The knowledge graph proposed in this work uses Simple Knowledge Organization System (SKOS) standard. A java-based tool is developed that uses the generated knowledge graph as the input and conducts light-weight reasoning to support smart maintenance diagnosis.