Evaluation of teaching quality in colleges and universities is an important segment in the process of teaching management. It is a complicated and abstract nonlinear problem affected by many factors. Therefore, in this paper, we propose an adaptive Back-Propagation (BP) neural network model for the evaluation of teaching quality in colleges and universities. Firstly, this model introduces the adaptive learning rate momentum to improve the traditional gradient descent method to ensure the model's convergence speed and optimizes the network structure of the model to ensure its stability. Besides, in the input feature vector, this model adds pre-teaching preparation and teaching process as first-level indexes in the evaluation index system, then normalize the evaluation index sample datasets to improve computation efficiency and ensure a comprehensive evaluation. Finally, we test the performance of this model using a teaching quality evaluation dataset collected in the university's educational administration system and teaching expert group. Compared with other evaluation models using some indicators, such as MSE, prediction accuracy, iterations and so on, the results show that the proposed model works better in MSE and prediction accuracy and improves the speed of convergence, which validates the effectiveness of the model in teaching quality evaluation in colleges and universities.