Efficiently compressing three-dimensional (3D) point cloud data (PCD) without sacrificing crucial physical properties is essential in many fields, including computer graphics, robotics, and simulation. This paper presents an innovative approach for compressing 3D PCD that integrates geometric restrictions into neural networks analyzed with physical loss in input and uses principal component analysis (PCA) in neural network feature selection, all while protecting important physical properties in 3D PCD. The proposed novel method combines geometric restrictions, physics loss, and topological optimizations in a mixed model architecture. In particular, the proposed method utilizes the input to the neural network and use dense layers to impose geometric constraints with PCA-reduced features during the compression process. Encapsulating the intricate linkages included in the PCD while preserving the fundamental physical laws guiding its formation, it is meticulously constructed. approach minimizes a composite loss function by methodically training and optimizing, striking a balance between reconstruction fidelity and respect for geometric and physical restrictions. The approach lights the effectiveness in maintaining the integrity of 3D PCD while achieving efficient compression of size into Kilobytes in proposed PINN-COMP model is demonstrated by validation and evaluation and the proposed work is lossless technique. With significant implications for a wide range of areas needing precise 3D modeling and simulation, the suggested approach presents a viable option for effective data representation and storage.