The invention relates to the technical field of ship construction and
artificial intelligence, in particular to a CGPINNS-based ship construction cost efficient prediction method, which comprises the following steps: step 1,
data acquisition: collecting ship design parameters, construction materials, process, labor cost and other data; 2, data preprocessing, cleaning, denoising and data normalization are carried out, and the
data quality is improved; 3, a CGPINNS model construction module replaces a derivative item in gradient calculation by using central difference and is fused into a ship construction physical equation; step 4, after the model training module is trained by an Adam
algorithm, the
cost prediction module outputs a prediction result and a
confidence interval; and 5, comparing the predicted cost with the
actual cost by a result evaluation and feedback module, and if an error exceeds a threshold value, optimizing the model by the
system. According to the method, indexes such as L2 relative errors are better,
physical information is effectively fused, the generalization ability is high, accurate
cost prediction can be provided for ship construction enterprises, scientific
decision making can be assisted, and
economic benefits can be improved.