Dataset for Self-adjustable tool center point of industrial robots based on neural network models

This dataset is the official supplementary material for the article "Self-adjustable tool center point of industrial robots based on neural network models." It contains the raw, unshuffled computational and experimental data required to reproduce the training and validation processes of three modular neural networks (NN1, NN2, and NN3). These models are designed for the predictive identification and real-time kinematic compensation of filler wire deviation in Gas Metal Arc Welding (GMAW) based Wire Arc Additive Manufacturing (WAAM).

Data Structure:
The dataset is structured into three separate sheets/files, each containing the complete, continuous dataset corresponding to a specific predictive module in the framework:

NN1 Data: Contains raw data for predicting the unconstrained wire deviation (D, mm) based on the initial radius of curvature (R, mm) and the current geometric wear of the contact tip (W', mm).
NN2 Data: Contains raw data for evaluating the normal contact force (F, N) exerted by the wire, determined by its inherent radius of curvature (R, mm), effective flexural rigidity (EI, GPa*mm^4), and current wear state (W', mm).
NN3 Data: Contains raw data for modeling the progressive wear (W, mm) as a function of the consumed wire length (L, m), normal contact force (F, N), and material microhardness (H).

Methodological Note (Data Processing & Leakage Prevention):
The provided files contain the raw, sequential data arrays prior to any preprocessing.

Furthermore, please note that the physical validation data acquired using the 308LSi stainless steel wire is deliberately excluded from this provided training dataset. The 308LSi physical trial served strictly as an isolated, unseen test set during the final printing experiments to rigorously prevent data leakage and prove the true generalization capabilities of the developed models.