The maximum absolute average error when using the GRNN model on the same data set was approximately 6%%.
Comparisons between mechanistic model predictions and the acquired experimental data show a maximum absolute average relative error of approximately 7% for the cross-section void fraction.
The maximum absolute average fitting error is 0.11% for the fresh nanofluids while it is 0.24% for the 3-day fluids.
The Peng-Robinson EOS with the alpha correction by Mathias and Copeman, the mixing rules by Wong and Sandler, and the NRTL excess Gibbs energy model is fitted to the data in the present work, with a maximum absolute average deviation of 0.01 in mole fraction.
In addition, the following aberration filters were employed: minimum number of probes in region = 2, minimum absolute average log2 ratio for region = 0.5, maximum number of aberration regions = 10,000.
Table 4 Maximum absolute error and average absolute relative deviation to indicate the difference between the predicted values and experimental data Method Maximum absolute error Average absolute relative deviation LS-SVM model 5.42 1.5433 Emera and Sarma method 91.2154 66.92 Simon and Graue method 125.70 56.87.
When the modulus of the wave is less than 0.95, the maximum error and absolute average error are acceptable for engineering.
From the profile ordinate data read at 1 μm intervals, a number of surface roughness parameters – centre line average, root-mean-square roughness, maximum height, maximum mean peak height, maximum mean depth, and absolute average asperity slope were calculated using a computer program.
In addition, the largest average maximum absolute principal stress is observed to occur along the particle circumference at the center-plane of the particle.
Six statistical indicators (e.g., percent relative error, maximum absolute percent relative error, average absolute percent relative errors, percent bias, index of agreement, and Nash Sutcliffe efficiency) and relative computer computation time are used for assessing the model performance.
The average maximum absolute power difference for the considered tests across the phenotype models were 39.8%, 43.9%, 25.8% and 31.8% for MDMR, SKAT, KBAT and U-test respectively.
