Background: A talar body prosthetic implant may be indicated after a severe fracture in the talus bone resulting in avascular necrosis with collapse. This process is patient-specific, in which the geometry is copied from the healthy talus of the opposite foot to create the implant. More recently, an ‘off-the-shelf’ non-custom talar prosthetic was proposed, consisting of a standardised shape in 10 different sizes. Methods: The generic-shaped talus was determined by creating 3D models via image processing software MIMICS and further refined using Geomagic, from raw data of CT scan imaging. This study evaluated the intra- and inter-operator reliability in 3D modelling of talus bone from CT scan imaging. Results: Four operators created 3D models using CT scans of four subjects via the documented protocol. The average deviations were well within the acceptable value, and although the extreme values were large, the distributions showed that the critical deviations were either few points, or small areas. Intra- and inter-operator differences were not statistically significant in most cases. The talus bone was found to have larger deviations than the tibia due to rougher surface. Conclusions: From our results, the method of converting 2D CT images to 3D models used to develop generic talus implant is reliable within the acceptable tolerances.
A computationally faster and reliable modeling approach called a physics-based artificial neural network framework for adsorption and chromatography emulation (PANACHE) is developed. PANACHE uses deep neural networks for cycle synthesis and simulation of cyclic adsorption processes. The proposed approach focuses on learning the underlying governing partial differential equations in the form of a physics-constrained loss function to simulate adsorption processes accurately. The methodology developed herein does not require any system-specific inputs such as isotherm parameters. Accordingly, unique neural network models were built to fully predict the column dynamics of different constituent steps based on unique boundary conditions that are typically encountered in adsorption processes. The trained neural network model for each constituent step aims to predict the entire spatiotemporal solutions of different state variables by obeying the underlying physical laws. The proposed approach is tested by constructing and simulating four different vacuum swing adsorption cycles for post-combustion CO2 capture without retraining the neural network models. For each cycle, 50 simulations, each corresponding to a unique set of operating conditions, are performed until the cyclic steady state. The results demonstrated that the purity and recovery calculated from the neural network-based simulations are within 2.5% of the detailed model's predictions. PANACHE reduced computational times by 100 times while maintaining similar accuracy of the detailed model simulations.
Large scale physics-based reservoir models are employed routinely in the prediction of the behavior of steam assisted gravity drainage (SAGD) processes under different operational situations. However, parametric uncertainty persists in these models even after history matching with production data. This uncertainty, and the computational cost associated with the full-scale reservoir simulations, makes it challenging to use reservoir simulators in closed-loop control of reservoirs. As an alternative strategy, we present in this work a dynamic proxy model for the reservoirs based on system identification and the prediction error method using only injection and production data. These proxy models are validated against field data from a SAGD reservoir and simulated synthetic reservoir data and shown to be appropriate for use in model predictive control. We also provide evidence that the predictive power of these models can be improved by the appropriate design of input signals (injection rates and pressures).
In this work, we detail an automated reaction network hypothesis generation protocol for processes involving complex feedstocks where information about the species and reactions involved is unknown. Our methodology is process agnostic and can be utilized in any reactive process with spectroscopic measurements that provide information on the evolution of the components in the mixture. We decompose the mixture spectra to obtain spectroscopic signatures of the individual components and use a 1-D convolutional neural network to automatically identify functional groups indicated by them. We employ atom-atom mapping to automatically recover reaction rules that are applied on candidate molecules identified from chemistry databases through fingerprint similarity. The method is tested on synthetic data and on spectroscopic measurements of lab-scale batch hydrothermal liquefaction (HTL) of biomass to determine the accuracy of prediction across datasets of varying complexities. Our methodology is able to identify reaction network hypotheses containing reaction networks close to the ground truth in the case of synthetic data, and we are also able to recover candidate molecules and reaction networks close to the ones reported in the previous literature studies for biomass pyrolysis.
While ammonia synthesis and decomposition on Ru are known to be structure-sensitive reactions, the effect of particle shape on controlling the particle size giving maximum turnover frequency (TOF) is not understood. By controlling the catalyst pretreatment conditions, we have varied the particle size and shape of supported Ru/γ-Al2O3 catalysts. The Ru particle shape was reconstructed by combining microscopy, chemisorption, and extended X-ray absorption fine structure (EXAFS) techniques. We show that the particle shape can change from a round one, for smaller particles, to an elongated, flat one, for larger particles, with suitable pretreatment. Density functional theory calculations suggest that the calcination most likely leads to planar structures. We show for the first time that the number of active (here B5) sites is highly dependent on particle shape and increases with particle size up to 7 nm for flat nanoparticles. The maximum TOF (based on total exposed Ru atoms) and number of active (B5) sites occur at ∼7 nm for elongated nanoparticles compared to at ∼1.8−3 nm for hemispherical nanoparticles. A complete, first-principles based microkinetic model is constructed that can quantitatively describe for the first time the effect of varying particle size and shape on Ru activity and provide further support of the characterization results. In very small nanoparticles, particle size polydispersity (due to the presence of larger particles) appears to be responsible for the observed activity.
Process models that are affected by uncertainties need a robust mechanism to account for them in the model based design of experiments (DOE). The aim of this study is to design a set of experiments to estimate the parameters of multiscale kinetic models for the catalytic decomposition of ammonia. Along with uncertainties in the model, the problem is challenging due to constraints on experimental conditions. A stochastic D-optimal design is used to find the optimal experimental conditions using maximization of the expectation of properties of the Fisher information matrix (FIM). The expectation of FIM is calculated by sample average approximation (SAA) based on Monte Carlo simulations. Particle swarm optimization (PSO) is used to perform stochastic optimization to find the optimal set of experimental conditions. A novel method based on the rescaling of velocities is proposed for handling of equality and inequality constraints in particle swarm optimization.
'/%3e%3cuse xlink:href='%23a' transform='rotate(60)'/%3e%3ccircle r='17' fill='%23000095'/%3e%3ccircle r='15' fill='%23fff'/%3e%3c/svg%3e)
'%3e%3ccircle r='20' fill='%23008'/%3e%3ccircle r='17.5' fill='%23fff'/%3e%3ccircle r='3.5' fill='%23008'/%3e%3cg id='d'%3e%3cg id='c'%3e%3cg id='b'%3e%3cg id='a' fill='%23008'%3e%3ccircle r='.875' transform='rotate(7.5 -8.75 133.5)'/%3e%3cpath d='M0 17.5.6 7 0 2l-.6 5L0 17.5z'/%3e%3c/g%3e%3cuse xlink:href='%23a' transform='rotate(15)'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='rotate(30)'/%3e%3c/g%3e%3cuse xlink:href='%23c' transform='rotate(60)'/%3e%3c/g%3e%3cuse xlink:href='%23d' transform='rotate(120)'/%3e%3cuse xlink:href='%23d' transform='rotate(-120)'/%3e%3c/g%3e%3c/svg%3e)