Determination of Centrifugal Blood Pump Characteristics using CFD and Experimental Analysis

Abstract

Background Cardiovascular diseases are the leading cause of death throughout the developed world, attributed to approximately 17.8 million deaths worldwide in 2017 with increasing prevalence due to the aging population. Cardiovascular diseases generally result in heart failure. While the best treatment option for heart failure patients is heart transplantation, there is a severe deficiency in the availability of donor hearts. Rotary Blood Pumps (RBPs) utilised as Ventricular Assist Devices (VADs) provide an alternative treatment option. These devices are small implantable pumps that support the failing heart by providing power to augment circulation. The development of RBPs generally begins with initial designs obtained using traditional pump design methods (such as that developed by Stepanoff). However, studies have shown that this approach produces RBP prototypes far from optimal in design. Traditional theory relies on design constants derived empirically for large industrial pumps and these do not scale down well when applied to the much smaller RBPs. The initial designs are therefore generally quite poor and require an iterative build-and-test approach to obtain suitable pump prototypes – a process that is expensive and time consuming. Therefore, by improving the methodology for obtaining initial designs to better reflect the final product, development time can be greatly reduced. A popular avenue for analysing the effect of design variations and to further develop early prototypes of RBPs is to employ Computational Fluid Dynamics (CFD) simulations. These numerical simulations provide detailed data regarding the flow fields within these devices. However, a range of simulation options is available, leading to a wide range of potential predictions. In an attempt to provide a benchmark case, the FDA presented a challenge in which a pump design and test conditions were defined, allowing for direct comparison amongst different simulation approaches from a number of labs/RBP developers. The purpose of this thesis was to produce a gross design tool to provide a good starting point in RBP prototyping and a CFD simulation approach for verification that can also be used as a design refinement tool. Methods Formulating a design method for pumps requires the generation of empirical data. A number of pump design variables was identified as having an impact on pump performance, and a large number of experimental tests would have been needed to test the influence of each. Instead, a Design of Experiments (DOE) was utilised to streamline the process. The DOE outputs a relatively small number of tests required to fit a statistical model. Each design specified by the DOE was examined experimentally using a custom-built automated pump test platform to generate a number of performance measures. The obtained results were used to formulate a Response Surface Method (RSM) statistical model that showed acceptable fit to the input data. Coupled with desirability functions, the RSM model allowed for design optimisation. This tool essentially replaces Stepanoff’s traditional design methodology. The RSM model provides a robust tool that allows the user flexibility in design optimisation goals. The FDA pump was investigated in this thesis and a wide variety of simulation approaches was examined to determine which was most accurate. A range of factors were considered which included: mesh density, interface position between the rotating and stationary zones, steady vs. transient simulations, discretisation schemes, time step size and choice of turbulence model. The most appropriate option from each investigative study was selected to determine a recommended simulation approach. Final simulations were performed using these recommendations and were compared to the FDA experimental results to confirm the suitability of the suggested settings. Determination of Centrifugal Blood Pump Characteristics using CFD and Experimental Analysis iii The statistical model developed was used to design two different impellers as validation test cases. The first impeller was designed to optimise the maximum efficiency, P – Q curve slope and efficiency consistency. The second impeller was designed to mimic the approach used in traditional design methods for RBPs in setting a target design point as the primary objective and the aforementioned factors (from the first impeller) as secondary objectives. These two case studies underwent statistical performance predictions, CFD simulations, PIV analysis and experimental hydraulic testing to validate the statistical and CFD models. Results From the initial CFD study, a hybrid SBES turbulence model with full transient simulation on a fine grid with small time steps proved to be the most suitable both in terms of pressure rise generated by the FDA pump and resulting velocity fields when compared to published experimental results. From these findings the CFD modelling strategy was established. CFD results for the two validation pumps showed pressure rises matching the experimental data (8% and 1% difference for each impeller) within an acceptable range (<10% from the mean). The simulated velocity fields also closely replicated the PIV data for the majority of the flow domain. The statistical performance predictions well reflected those measured experimentally with the majority of data points falling within its confidence intervals. The hydraulic results also supported the main goal of this thesis, whereby an impeller generated using the statistical model, operated far closer to the target design point than that of a blood pump designed following Stepanoff’s methodology. Overall, both the statistical model and CFD approach provided accurate predictions and the purpose of the thesis was achieved. Final Remarks The statistical and CFD models developed in this thesis yield an effective design tool and verification methodology and show improvement over the current traditional design methods and accuracy in simulated results. Ultimately, the utilisation of these tools will lead to a reduction in the development time for new RBPs and provide a good understanding of the flow dynamics within these pumps, leading to improved pump designs reaching patients sooner. These tools are readily generalizable and could be adopted as design tools now.