Defect Estimation Using Coupling Matrix and DPSM

The patent segments the defect detection problem by dividing the object into multiple depth layers (voxels) and processing each layer independently through iterative inversion. This segmentation separates the superimposed defect signals from different depths, allowing accurate reconstruction of individual defect locations and characteristics without interference from other defects in the superposition.

The patent applies mathematical inversion techniques to reverse the forward modeling process. Instead of predicting sensor readings from known defects, the method inverts the relationship to estimate defect characteristics from measured sensor data. This inversion approach, combined with regularization, recovers the original defect distribution from the superimposed and disturbed sensor signals.

The patent introduces depth as an explicit dimensional parameter by discretizing the inspection volume into voxels along the depth axis. This dimensional transformation allows the method to resolve defects at different depths independently, converting the superimposed 2D sensor readings into accurate 3D defect reconstruction even when sensors are placed on only one side of the object.

The patent introduces a forward model and iterative inversion algorithm as intermediaries between the sensor measurements and defect reconstruction. This intermediary computational framework processes the superimposed sensor signals, separates contributions from different depth layers, and reconstructs accurate defect images without requiring physical access to both sides of the object.

The patent makes the computational system self-service by implementing an iterative inversion algorithm that automatically extracts depth information and reconstructs 3D defect images from single-angle sensor measurements. The algorithm internally performs the mathematical operations equivalent to multiple observation angles, eliminating the need for physical repositioning of sensors while achieving 3D reconstruction.

The patent replaces the mechanical system of physically moving sensors to multiple observation angles with a computational system performing mathematical inversion. This substitution uses algorithms to simulate multi-angle measurements from single-angle data, reducing device complexity and measurement time while maintaining accurate 3D defect reconstruction capability.

The patent transforms the defect characterization from qualitative visual inspection to quantitative parameter estimation by inverting the forward model to directly calculate defect properties such as depth, size, and location. The method outputs numerical values for defect parameters rather than images, enabling precise quantification of defect characteristics from the sensor measurements.

The patent replaces conventional image-based qualitative assessment with a computational inversion system that directly calculates quantitative defect parameters. This substitution transforms the measurement process from producing visual images to generating numerical estimates of defect characteristics, achieving accurate quantification without information loss.