To improve the accuracy of palm surface reconstruction in flexible robot grasp sensing, we conduct a COMSOL simulation to select a ring arrangement comprising of 7 fiber Bragg grating (FBG) flexible sensors packaged with polydimethylsiloxane (PDMS) on a 436 mm×436 mm×2 mm polypropylene plate. Assuming that the center and two corner ends of the plate were subjected to stress, respectively, we collected sensor data using a fiber grating demodulation instrument during the experiment. The data was continuously interpolated using cubic spline interpolation. Several planes Y intersected with the fitting ring which created a three-dimensional surface. We calculated the point function to obtain the point set and achieve a fitting visual display of the spatial surface. When the center of the end of the surface is under stress, the plate experienced a minimum relative error of 0.549% in end displacement, with a maximum relative error of 8.300%, the minimum absolute error of 0.051 cm, and a maximum absolute error of 1.255 cm. When both corners at the end of the plate are under stress, for the end displacement of the plate reconstruction, the minimum relative error is 2.546%, and a maximum relative error is 14.289%, the minimum absolute error is 0.005 cm, and the maximum absolute error is 0.729 cm. These experimental results provide a foundation to implement palm grip sensing in flexible robots.