Optimize Pump Design With Simulation and AI

Pumps are deemed essential equipment in nearly every industry, from manufacturing and healthcare to energy and automotive. Two major pump types exist: positive displacement (PD) pumps, which trap and force liquid, and dynamic pumps, also known as kinetic pumps, which are powered by impeller movement. Each category includes variations, such as centrifugal pumps — one of the most widely used dynamic pumps, with a global market size of $44.9 billion in 2025, projected to reach $74.78 billion by 2034, according to Fortune Business Insights.

PD pumps move fluid by trapping it and forcing it out, making them effective for thick, high-viscosity fluids. Centrifugal pumps, on the other hand, rely on a rotating impeller to increase pressure, which is more effective for thinner, low-viscosity fluids. Consequently, centrifugal pumps are pivotal to sewage systems, water treatment systems, chemical processing, pharmaceutical manufacturing, and oil and gas installations.

To manage liquids precisely and efficiently across varied flows and pressures, the design of centrifugal pumps requires more data and variables than PD pumps. This presents challenges for original equipment manufacturers (OEMs), including cost of prototyping and design cycle time. Ansys, part of Synopsys, offers solutions that combine multiphysics simulation, optimization, and artificial intelligence (AI) to help OEMs overcome these design challenges and increase efficiency.

Improve Pump Design Workflows

Key components of centrifugal pump design include how many stages it has, impeller construction, bearing configuration (for example, overhung or between bearings), orientation, and driver mounting. Impellers endure cyclic fatigue from pressure fluctuations and centrifugal forces, which can lead to failures, particularly at vane-to-shroud junctions in closed impellers. Over-design increases weight and forces, while under-design risks failure. The goal is to use current designs as a reference to improve safety, endurance, and performance.

One challenge that OEMs face during product improvements is converting complex geometries into simulation-ready models, such as those from legacy designs, scanned prototypes, worn or damaged parts, or as-manufactured parts. Synopsys Simpleware 3D image processing software provides this value. Simpleware software specializes in transforming 3D scan data from computed tomography (CT), micro-CT, or other imaging methods directly into high-quality meshes for computational fluid dynamics (CFD) and finite element analysis (FEA). This presents several strategic opportunities for pump OEMs.

Left: Running industrial computed tomography (CT) image-based simulations requires reliable integration between imaging and simulation platforms. Right: Simpleware software provides advanced model validation to ensure that CT-derived FEA and computational fluid dynamics (CFD) models maintain their accuracy and are optimized for simulation.

The obtained mesh can be used in CFD simulations to refine design variables, such as vane count, angles, and thickness, and meet hydraulic targets. The resulting design and pressure data are used in FEA to assess stress and fatigue, which supports material selection and early design decisions. This enables efficient, durable pump designs before prototyping, saving time and cost.

OEMs can integrate Ansys CFD or FEA solvers with Ansys optiSLang process integration and design optimization software to explore pump designs running closest to the best efficiency point (BEP). To determine BEP, engineering teams run simulations to assess fluid height, known as head capacity, and obtain the flow rate. After constructing a design database, these designs can be used in the Ansys SimAI cloud-enabled AI platform to build a physics-based generative model, enabling the creation of new designs more quickly.

During centrifugal pump design, the best efficiency point (BEP) is determined by assessing fluid height, known as head capacity, and obtaining the flow rate, the rate at which fluid is received measured by brake horsepower. Source: ESP Pump Performance Curves and Effect of axial forces.

Observe the Tools in Action

Let’s walk through a sample workflow featuring a practical application by a major OEM that needed to upgrade its pump line for compliance with new European Union (EU) Energy-related Products (ErP) regulations. The regulations mandate a minimum efficiency index (MEI) of 0.4 or higher, leading to 40% of models being eliminated due to low efficiency, according to the European Commission.

The OEM needed a 4% efficiency improvement at BEP to satisfy the MEI requirement with a margin, but pump casing changes were not possible due to tooling costs, and the manufacturing process needed to remain the same. Therefore, the OEM focused on optimizing the impeller. The approach involved 3D scanning the manufactured part to create a CFD mesh, running CFD simulations, and adjusting the turbulence model until the results matched test data in the desired range.

This enabled the OEM to explore and optimize design features, such as impeller exit angle, vane count, and vane thickness, without needing tooling changes. In the CFD-optiSLang interface, these features serve as inputs while outputs depend on design goals, such as maximizing the BEP, keeping the head within 3% of the target, and reducing the net positive suction head (NPSH) required. NPSH is the pressure needed at the pump’s inlet to stop the liquid from boiling and prevent cavitation. It’s found by subtracting the vapor pressure from the inlet pressure. Cavitation causes pitting and damage to the internal surface of the pump. The pressure fluctuations from the CFD cavitation analysis can also be used as an input for a structural simulation and to assess the components for stress and durability hot spots.

Reduced-order modeling is a key feature of optiSLang software, enabling metamodeling for faster and efficient design analysis. The software’s automatic machine learning (AutoML) algorithms, including the metamodel of optimal prognosis (MOP) and adaptive metamodel of optimal prognosis (AMOP), automatically search for the most robust design configuration. After the initial design of experiments (DOE), AMOP identifies areas for improvement and runs additional simulations to enhance the model. As a result, the software’s DOE analysis provides information to train a data‑driven AI surrogate modeling platform, such as the SimAI platform, and test designs in minutes rather than weeks. With this workflow, an extensive analysis is done just once and can be reused indefinitely.

In addition, the part-level pump simulations can be used as reduced-order models (ROMs) within complete system-level simulations to verify pump sizing alongside motor and piping data, accounting for losses. The simulations also predict water hammer, a pressure spike from sudden flow changes, which may damage pipes and equipment.

Additional tools to support this workflow include the Twin Builder simulation-based digital twin platform and ModelCenter model-based systems engineering (MBSE) software. Twin Builder software integrates AI techniques and enables you to deploy FEA and CFD as ROMs for system-level validation and operational health monitoring applications. The ModelCenter platform, which is part of the Ansys Connect product family of digital engineering enablement solutions, automates workflows and tracks changes across the product design cycle, helping OEMs mitigate cost impacts.

Insights from simulation and other digital engineering solutions empower OEMs to make early and crucial design decisions by accounting for as much uncertainty as possible from individual parts to system-level performance. This reduces late-stage surprises, tooling changes, and motor sizing issues. Additionally, by verifying workflows throughout the product development cycle, OEMs can more easily validate products for certification and accelerate time to market.

Upgrade Your Pump Designs Today

OEMs struggle with cost, time, and reliability in the design of centrifugal pumps. Ansys tools use advanced simulation, optimization, and AI to improve performance while reducing expenses and development time. These solutions address uncertainties early, improving efficiency, and accelerating market readiness.

Find tools to support your pump designs at Industrial Processes and Equipment Simulation Software Solutions from Ansys.

To discover optiSLang optimization software firsthand, request a free trial.