Set-up

We have based our PDUQ examples on the Cu-Mg example from ESPEI, so please look at that first if you want to understand what is required to use PDUQ for your own system. In short, ESPEI takes the raw data, Gibbs energy models, and phase descriptions and computes the parameter values for those models. What makes ESPEI different from traditional CALPHAD tools is that it finds the parameters through Bayesian methods and can therefore provide a distribution for each parameter instead of a single, deterministic value. This enables all of the uncertainty quantification tools that PDUQ provides.

In all of our examples, we will be using the outputs of the trace.npy file. trace.npy simply contains all of the Gibbs energy parameter sets generated from a single ESPEI run. trace.npy has three dimensions, the first for the number of “walkers” in the MCMC algorithm ESPEI uses, the second for the number of total MCMC iterations, and the last for the number of paramters.

As an example, one trace file might have the following shape

import numpy as np
print(np.load('trace.npy').shape)

(150, 350, 15)

meaning that there are 150 walkers, 350 iterations, and 15 parameters. We will be using this trace file along with the CU-MG_param_gen.tdb file for the remaining examples.