Searching the optimal parameters of a 3D scanner in surface reconstruction of a dental model using central composite design coupled with metaheuristic algorithms

• Published in: International journal on interactive design and manufacturing 2024-12, Vol.18 (10), p.7401-7411
• Main Authors: Kaushik, Ashish, Garg, Ramesh Kumar
• Format: Article
• Language: English
• Subjects: Accuracy; CAE) and Design; Computer-Aided Engineering (CAD; Dental materials; Dentures; Design; Digitization; Electronics and Microelectronics; Engineering; Engineering Design; Factorial experiments; Genetic algorithms; Heuristic methods; Image reconstruction; Industrial Design; Instrumentation; Lasers; Light; Luminous intensity; Measurement techniques; Mechanical Engineering; Neural networks; Original Paper; Parameter estimation; Process parameters; Prostheses; Scanners; Search algorithms; Software; Standard deviation; Three dimensional composites
• ISSN: 1955-2513; 1955-2505
• DOI: 10.1007/s12008-023-01587-z

Determining the right process parameters for 3D scanning is crucial for rigorously inspecting reverse-engineered dental models. However, it is seen that various parameters, such as scanning distance, light intensity, and scanning angle, are rarely examined during preliminary experimental trials. The proposed research examines a method for estimating the ideal values of the aforementioned scanning parameters that minimize acquisition error. The face-centered, central composite design suggested twenty runs of experimentation with varying input parameter combinations. In each of these twenty scans, a physical denture model was scanned to extract a 3D CAD model, and the standard deviation of each model was calculated to investigate into the scan accuracy of the recorded data. A neural network architecture is used to train a model across input and output, and then the model is optimized by a genetic algorithm for the best results. Through a scanning distance of 208.28 mm, scanning angle of 54.1 degrees, and light intensity of 18 W/meter square, in a total of twenty trial runs, the lowest possible standard deviation of 0.2626. The standard deviation is minimized for achieving maximum accuracy using a heuristic GA-ANN algorithm with a scanning distance of 152.4 mm, scanning angle of 61.8 degrees, and light intensity of 14 watts per square meter and same has been validated experimentally.