Figure 5

Expected, , and CNN-predicted, , galvanostatic behavior for representative battery microstructures. Inset (a) compares experimental data, , and traditional porous electrode theory response, as reported by Doyle and Newman⁷⁴, by using traditionally assumed values, \(\alpha =1/2\) and \(S = 3\), while CNN-predicted values are \(\alpha =2.37\) and \(S = 11.12\). In inset (b) expected values corresponds to \(\alpha =0.05\) and \(S = 1.0\), while CNN-model generated values are \(\alpha =0.12\) and \(S = 0.96\). The root mean squared, RMS, deviation in galvanostatic behavior of the CNN-model prediction with respect to the expected values, show a value of 1.5%. Inset (c) corresponds to dual porous structure with low porosity, with expected values of, \(\alpha =0.05\) and \(S = 40\), while the CNN-model generated values are \(\alpha =0.05\) and \(S = 39.05\). The RMS deviations are less than 0.05 %. Inset (d) corresponds to a distribution of highly textured (aligned, MRD > 20), morphologically anisotropic particles (c/a \(\sim\) 1/10) with expected values of \(\alpha =6\) and \(S = 1.0\), while the CNN-model generated parameters are \(\alpha =5.70\) and \(S = 1.48\). A maximum RMS deviation of 13.3% is observed. Inset (e) expected values are, \(\alpha =6\) and \(S = 40\), and the CNN-model predicted values are, \(\alpha =6.03\) and \(S = 39.22\). The RMS deviations are less than 0.15%. Inset (f) expected values correspond to \(\alpha =8\) and \(S = 5\), while the CNN-model predicted values are \(\alpha =7.65\) and \(S = 4.37\). The maximum RMS deviation is 5.5 % and the minimum RMS deviation is 0.67%.