Wireless sensors in structural health monitoring are useful to simplify deployment and to realize a remote control of deployed sensors. However, wireless sensors operate in significant constraint on power consumption. Reduction techniques of radio communications in sensor networks are useful to extend the battery life. This study is intended to a method to utilize sensor grouping so as to reduce amount of data by radio communications between sensors. The sensor grouping is decided by considering correlations between sensors, and then the original system matrix is reconstructed into a sparse-like system matrix by replacing components for the weakly correlated sensor in the system matrix with zero. To verify the validity of the proposed approach, the sensor grouping is examined by means of simulations and field experiments. Observations demonstrated feasibility of system identification utilizing the sparse-like system matrix.