Parametric Studies &
Design of Experiments
Master the art of exploring multiple metallurgical scenarios virtually. Move beyond trial-and-error by systematically varying parameters to discover the perfect casting process before pouring a single drop of metal.
In traditional foundry environments, optimization is a grueling, physical process[cite: 1]. A misrun occurs; the pouring temperature is arbitrarily raised[cite: 1]. Shrinkage is detected; a riser is enlarged based on heuristic guesswork[cite: 1]. This sequential, physical trial-and-error is not only economically ruinous due to wasted material, energy, and labor, but it is fundamentally unscientific[cite: 1]. It observes the symptom without holistically mapping the problem space[cite: 1].
Enter the virtual Design of Experiments (DOE)[cite: 1]. Utilizing advanced platforms like PoligonSoft, metallurgists and casting engineers can execute parametric studies[cite: 1]. This methodology involves identifying critical variables—such as pour temperature, gating velocities, mold pre-heating, and cooling conditions—and systematically simulating combinations of these variables across a defined multidimensional matrix[cite: 1].
By doing so, we transition from asking "Did this one change fix the defect?" to asking "What is the exact mathematical relationship between variable X and outcome Y, and where is the absolute optimal operating window?"[cite: 1] This exhaustive guide will walk you through setting up, running, and analyzing a complete virtual DOE[cite: 1].
Phase 1: Defining the Parametric Space
The foundation of any successful DOE is the rigorous selection of variables[cite: 1]. In a PoligonSoft parametric study, you are not simply running a simulation; you are programming a scenario generator[cite: 1].
Consider a high-value steel casting suffering from borderline microporosity and occasional misruns[cite: 1]. Physical intuition suggests hotter metal might cure the misrun, but could exacerbate gas solubility and volumetric shrinkage[cite: 1]. To test this virtually, we establish a parameter matrix[cite: 1]. Instead of running one model, we instruct the software to permute through a range of designated inputs[cite: 1].
Common Permutation Variables:
Interactive Simulation Matrix Builder
Select parameters to vary. Watch how the total number of required simulations scales multiplicatively[cite: 1].
Phase 2: Automated Batch Execution
Once the parametric matrix is established, the computational burden begins[cite: 1]. If a single complex thermal-fluid coupled simulation takes 4 hours to solve, a matrix of 12 permutations represents 48 hours of compute time[cite: 1]. Manually setting up and triggering each of these 12 runs sequentially is inefficient and prone to human error[cite: 1].
Modern simulation environments overcome this through Batch Processing[cite: 1]. Whether utilizing built-in macro scripts or dedicated batch runner utilities found in tools like PoligonSoft, the engineer configures the matrix once[cite: 1]. The system then automatically duplicates the mesh, applies the unique boundary conditions for iteration n, executes the solver, saves the result data into a structured directory, and immediately begins iteration n+1[cite: 1]. If hardware allows (e.g., multi-core HPC or cluster nodes), these jobs can run in parallel[cite: 1].
The Batch Execution Pipeline
Phase 3: Comparative Analytics & Visualization
Executing the batch is merely the computational prerequisite[cite: 1]. The true value of a parametric study lies in the post-processing and comparative analytics[cite: 1]. Reviewing a dozen 3D color plots individually is cognitively overwhelming[cite: 1]. We must abstract the 3D data into 2D plots to identify trends[cite: 1].
By extracting specific metrics from the solver—such as Total Macro-Porosity Volume (cm³) or Percentage of Casting with Niyama Criterion < Threshold—we can plot these outputs against our input parameters[cite: 1]. This visualizes the physical response curves of the casting[cite: 1]. In the interactive chart below, explore how altering the Pouring Temperature affects Total Porosity, and how that entire curve shifts when the gating design (choke area) is modified[cite: 1].
Defect Sensitivity Analysis
Total Porosity Volume vs. Pouring Temperature[cite: 1]
Phase 4: Multi-Objective Optimization
Finding the "best" parameter is rarely straightforward because foundry engineering involves conflicting objectives[cite: 1]. A higher pouring temperature might reduce the risk of a misrun to near zero, but it simultaneously increases mold degradation, energy costs, and potentially the size of the required riser (due to higher total volumetric shrinkage)[cite: 1].
This is the domain of Multi-Objective Optimization[cite: 1]. Advanced users of platforms like PoligonSoft look for the Pareto Frontier—the set of parameter configurations where you cannot improve one metric (e.g., porosity) without worsening another (e.g., yield)[cite: 1]. For instance, automatic riser placement algorithms can iterate through hundreds of positional coordinates and modulus sizes to find the absolute minimum riser mass required to achieve a sound casting under specific thermal constraints[cite: 1].
Optimal Zone Reached!
Yield > 65% | Misrun = 0% | Soundness > 98%[cite: 1]
⚙️ Real-time Parameter Tuning
The Virtual Paradigm Shift
The power of simulation is not merely in verifying a single design, but in providing an infinitely malleable virtual sandbox[cite: 1]. Physical design of experiments involves making costly patterns, melting physical metal, pouring, destructive testing, and waiting weeks for results[cite: 1]. It restricts innovation because failure is too expensive[cite: 1].
By adopting parametric studies within environments like PoligonSoft, failure becomes a cheap, virtual data point[cite: 1]. An engineer can evaluate 50 configurations overnight[cite: 1]. They can discover non-intuitive solutions—perhaps identifying that a slightly smaller riser, combined with a 10°C drop in temperature and a faster fill time, produces a completely sound casting with a 15% improvement in yield[cite: 1]. This level of optimization is mathematically impossible to achieve on the foundry floor through human intuition alone[cite: 1].
Stop guessing. Start simulating[cite: 1].