The study explores factors affecting system performance in retrieval-augmented generation.The RAG system consists of query expansion, retrieval, and text generation modules.Systematic experiments identify critical elements and propose new configuration methods.The findings provide insights for enhancing retrieval-augmented generation performance.