Abstract:Population balance equation (PBE) models have potential to automate many engineering processes with far-reaching implications. In the pharmaceutical sector, crystallization model-based design can contribute to shortening excessive drug development timelines. Even so, two major barriers, typical of most transport equations, not just PBEs, have limited this potential. Notably, the time taken to compute a solution to these models with representative accuracy is frequently limiting. Likewise, the model construction process is often tedious and wastes valuable time, owing to the reliance on human expertise to guess constituent models from empirical data. Hybrid models promise to overcome both barriers through tight integration of neural networks with physical PBE models. Towards eliminating experimental guesswork, hybrid models facilitate determining physical relationships from data, also known as 'discovering physics'. Here, we aim to prepare for planned Scientific Machine Learning (SciML) integration through a contemporary implementation of an existing PBE algorithm, one with computational efficiency and differentiability at the forefront. To accomplish this, we utilized JAX, a cutting-edge library for accelerated computing. We showcase the speed benefits of this modern take on PBE modelling by benchmarking our solver to others we prepared using older, more widespread software. Primarily among these software tools is the ubiquitous NumPy, where we show JAX achieves up to 300x relative acceleration in PBE simulations. Our solver is also fully differentiable, which we demonstrate is the only feasible option for integrating learnable data-driven models at scale. We show that differentiability can be 40x faster for optimizing larger models than conventional approaches, which represents the key to neural network integration for physics discovery in later work.

jinns: a JAX Library for Physics-Informed Neural Networks

SNNAX -- Spiking Neural Networks in JAX

SynJax: Structured Probability Distributions for JAX

JAX-SSO: Differentiable Finite Element Analysis Solver for Structural Optimization and Seamless Integration with Neural Networks

TensorNetwork: A Library for Physics and Machine Learning

FeNNol: an Efficient and Flexible Library for Building Force-field-enhanced Neural Network Potentials

JAX-SPH: A Differentiable Smoothed Particle Hydrodynamics Framework

SOLAX: A Python solver for fermionic quantum systems with neural network support

Slax: A Composable JAX Library for Rapid and Flexible Prototyping of Spiking Neural Networks

jaxsnn: Event-driven Gradient Estimation for Analog Neuromorphic Hardware

Computational Synthetic Biology Enabled through JAX: A Showcase

IDRLnet: A Physics-Informed Neural Network Library

JPC: Flexible Inference for Predictive Coding Networks in JAX

Modern, Efficient, and Differentiable Transport Equation Models using JAX: Applications to Population Balance Equations

JAXbind: Bind any function to JAX

OmniLearn: A Method to Simultaneously Facilitate All Jet Physics Tasks

JAXFit: Trust Region Method for Nonlinear Least-Squares Curve Fitting on the GPU

A physics-informed deep learning framework for inversion and surrogate modeling in solid mechanics

Efficient Bayesian inference using physics-informed invertible neural networks for inverse problems

TensorDiffEq: Scalable Multi-GPU Forward and Inverse Solvers for Physics Informed Neural Networks

BlackJAX: Composable Bayesian inference in JAX