In the context of the integration of man, machine and objects, the intelligent manufacturing of intelligent demand engineering systems is facing problems such as difficult to determine demand, low production efficiency and high production costs. To solve the above problems, this study conducts demand modeling based on knowledge graphs, constructs knowledge graphs in the field of human-machine-object fusion, and integrates the knowledge and relationships of different subjects to understand the needs more comprehensively. First, this paper implements multi-source data fusion and analysis to collect multi-source data from humans, machines, and objects, including sensor data, personnel operation records, product design documents, etc. Then, this paper uses random forests to fuse and analyze these data to mine hidden demand information. Finally, this paper introduces intelligent agent technology, which can coordinate and communicate between humans, machines and objects, automatically query knowledge graphs according to user needs, find suitable manufacturing resources, and interact with operators to convey needs and manufacturing instructions. The results show that the demand identification method based on knowledge graph shows high accuracy in multiple production processes, with the lowest accuracy of 90.2%. The production time is reduced under multi-source data fusion and analysis, and the maximum production cost is reduced from 644,000 to 389,000, and the lowest cost is reduced from 417,000 to 202,000. Intelligent agent technology has improved the coordination between humans, machines and objects in practical applications, and has demonstrated good performance in the manufacturing process, improving production efficiency and quality. These achievements have not only promoted technological progress in the manufacturing industry, but also provided valuable references for future research directions.