Abstract: Conventional well logs, such as porosity and resistivity, are the comprehensive reflection of lithology, physical characteristics and fluid properties. These key parameters, such as porosity and hydrocarbon saturation, are calculated under several assumptions or empirical models during conventional log inversion. Big errors usually occur to the tight sandstone reservoirs composed of complex minerals, leading to the reduction in parameter calculation accuracy and fluid identification consistency. Nuclear magnetic resonance (NMR)logging is only used for analyzing the reservoir physical properties and pore structures. Although some problems in porosity-permeability evaluation of tight reservoirs can be solved in that way, a lot of reservoir geological information needs to be investigated. Based on the analysis of conventional logging theoretical response models, in combination with the modular automated processing system(MAPS) images and EDS analysis of tight sandstone reservoir in Member 7, Yanchang Formation, Ordos Basin, the main reservoir mineral types and the pore development degrees of different minerals are identified. A new method of joint inversion between NMR porosity data and conventional logs is also proposed. The optimization algorithm is applied to quantitatively identify and evaluate the types and contents of clay and major rock forming minerals in tight sandstone reservoirs. The calculation accuracy of mineral content by joint inversion method is approaching to that by element scanning logs, and the application results of mineral assemblages in rock type identification and core analysis are equivalent, which can be used for analyzing reservoir lithology and lithofacies.