Accurate and reasonable cement take prediction is of great significance for effective control of dam foundation grouting quality and cost. This article combined the previous research results and engineering practice to explore the different influencing factors of cement take, and conducted parameter correlation analysis to determine the input parameters for prediction. Then, an intelligent prediction model for cement take based on improved extreme learning machine (ELM) is proposed, which uses the grasshopper optimization algorithm (GOA) to optimize its input weights w and hidden layer thresholds b. Finally, taking sets of cement take data from a real dam foundation project as an example to verify the performance of the proposed prediction model, the results illustrates that the proposed model has good prediction accuracy and can assist grouting engineers in adjusting grouting construction design and controlling grouting quality, which has a wide application prospect.