In order to improve the accuracy and reliability for detection of grains cleaning loss, a piezoelectric crystal sensor array was proposed in full width of distribution to realize the online multipoints detection. The dynamic model for grain collision sensor array was evaluated by the method of experimental model analysis. The distribution of low-level vibration and deformation was tested for the sensitive elements of detection array using dynamic signal tester. The location and quantity of the piezoelectric crystal units were determined. The simulation was done based on the software ANSYS for the impacts of grains on the sensitive elements. A quantitative analysis was developed to correct the positions of the piezoelectric crystal units and the construction was adjusted according to the deformation contours. The exact positions were quantitatively analyzed and corrected based on the numerical results for the deformation placement, which was used for optimizing the array construction. Then, the grain impact experiment was established on the test bench. The results showed that the performance was stable for the piezoelectric sensor array, and output signal amplitude was about 2.5 V, with uniform sensitivity in the whole range. The present array sensor helps to provide a technical foundation for detection of grain cleaning loss.