Elastic And Inelastic Stress Analysis

Getting Stress Analysis Right When Things Go Plastic

I used to run everything through linear elastic analysis because it was fast and the software spat out results in minutes. That habit cost me a month of rework on a pressure vessel project back in 2018. The manufacturer had specified a material with a yield point around 350 MPa, and my elastic model showed maximum von Mises stresses hitting 480 MPa with a safety factor below 1.5. The obvious call would have been to increase wall thickness, but that added weight we couldn't afford. Instead of just scaling up, I switched to an inelastic approach with a bilinear kinematic hardening model, and the results showed the structure was actually stable well beyond first yield. The plastic zone was contained and the deformation was within acceptable limits. That was a good lesson in not blindly trusting the first output you see. Elastic stress analysis assumes the material returns to its original shape once the load is removed. Stresses are proportional to strain according to Hooke's law, and the governing equations are linear. This works fine when you stay well below the yield point and you're doing preliminary sizing or checking serviceability conditions. The calculations are straightforward, the convergence is reliable, and you get results quickly. Most FEA packages default to this because it's computationally cheap. Inelastic stress analysis comes into play when you expect stresses to exceed the yield limit. Materials don't just stop working at yield. They redistribute loads through plastic deformation, strain hardening, and in some cases, creep at elevated temperatures. Inelastic analysis captures this behavior by using a nonlinear stress-strain relationship. You're no longer solving a linear system of equations. The solver has to iterate, increment the load in steps, and check for convergence at each point. This takes more time and more careful setup, but it gives you a realistic picture of what actually happens.

The key difference isn't just academic. In elastic analysis, you're checking whether stress stays below a limit. In inelastic analysis, you're checking whether the structure can safely undergo plastic deformation without collapsing or deforming excessively. The acceptance criteria change completely. You're not comparing against yield strength alone anymore. You're looking at ultimate strength, plastic rotation capacity, strain limits, and collapse mechanisms.

Setting Up an Inelastic Analysis in Practice

The hardest part of inelastic stress analysis isn't running the simulation. It's defining the material model correctly. A lot of engineers pull default steel data from software libraries and call it a day. That's where things fall apart. Real material behavior depends heavily on the specific grade, heat treatment, and temperature. If your material curve doesn't match what the actual component is made of, the inelastic results are garbage regardless of how sophisticated the mesh is. You need the full stress-strain curve, not just yield and ultimate strength. The curve should extend well past the yield point into the plastic region. For most structural steels, that means data up to at least 15 to 20 percent strain. If you're working with stainless steel or aluminum alloys, the strain hardening behavior is different and the curve shape matters even more. Bilinear approximations are acceptable for rough estimates, but they miss the curvature that real materials show between yield and ultimate stress. That curvature can shift your predicted collapse load by 10 to 20 percent in complex geometries. Mesh quality becomes much more critical in inelastic analysis. Stress gradients in plastic zones can be very steep, especially near geometric discontinuities like holes, notches, and weld toes. A coarse mesh will smear the plastic zone and give you non-conservative results. I learned this the hard way on a bracket design where the plastic hinge was developing at a fillet radius. My initial mesh had elements around 5 mm at that location, and the analysis showed no significant plasticity. When I refined the mesh down to 1 mm elements with a proper bias gradient, the plastic zone lit up exactly where I'd suspected it would. The peak strain at the fillet was nearly double what the coarse mesh had predicted. That change alone forced a redesign of the fillet geometry before the part went to manufacturing.

Boundary conditions and contact definitions also matter more than people realize. In elastic analysis, a fixed support is usually adequate. In inelastic analysis, the way you constrain the model can artificially restrict plastic flow and create false stress concentrations. I ran into this on a flange connection analysis where I had over-constrained the bolt holes. The elastic results were fine, but when I switched to inelastic, the model wouldn't converge because the plastic deformation had nowhere to go. Relaxing the constraints to allow radial movement at the bolt circle solved the convergence issue and gave results that matched the physical test data within 8 percent.

When to Use Each Approach

Linear elastic analysis is appropriate for routine strength checks where you know stresses will remain below yield under all expected loading conditions. This covers the majority of machine components, structural frames, and pressure equipment operating at room temperature with ductile materials. If your design life is long and you're dealing with cyclic loading, elastic analysis with appropriate fatigue methods is still the standard approach in most codes. Inelastic analysis becomes necessary when you're designing for overload conditions, impact events, or residual stress assessment. It's also required when codes explicitly call for plastic collapse checks. ASME Boiler and Pressure Vessel Code Section VIII Division 2 uses elastic stress analysis for primary membrane stress but requires plastic collapse assessment for certain geometries and loading cases. The code allows stress intensification factors and shape factors that are derived from inelastic behavior, even when you're running an elastic analysis. This hybrid approach is common in the industry and it's worth understanding because it blurs the line between the two methods. There's a middle ground called elastic-perfectly plastic analysis where the material yields at a constant stress after the elastic limit. This is simpler than a full nonlinear material model and can be useful for rapid collapse assessments. However, it ignores strain hardening entirely, which means it tends to be overly conservative for ductile materials. For most practical engineering work, a bilinear or multilinear stress-strain curve is a better balance between accuracy and computational cost.

Common Pitfalls That Waste Time

The biggest mistake I see is trying to converge an inelastic model without proper load stepping. Automatic time stepping in nonlinear solvers can take huge initial increments and then struggle to recover when the material yields. Setting up manual substeps with smaller increments in the regions where plasticity is expected usually cuts the total solve time significantly because the solver doesn't waste iterations backing out of divergence. On a recent hub assembly analysis, switching from automatic to manual stepping with a minimum increment of 0.01 reduced the solution time from six hours to about forty minutes on the same hardware. Another frequent error is ignoring geometric nonlinearity while including material nonlinearity. Large deformations can change the load path in ways that a purely material nonlinear model won't capture. If your structure undergoes significant deflection before yielding, turning on the large deformation option is essential. Without it, the stiffness matrix doesn't update correctly and you can get wildly inaccurate stress predictions. I had a thin-walled cylindrical shell where the elastic analysis showed a stress of 280 MPa and the inelastic analysis without geometric nonlinearity showed 310 MPa. When I included both material and geometric nonlinearity, the stress dropped to 195 MPa because the shell's deformation changed the load distribution. That's a 37 percent difference from the inelastic-only result, and it would have been a serious design error if caught late. Temperature effects are another area where people cut corners. Material properties change with temperature, and yield strength typically drops significantly as temperature rises. If your analysis involves any thermal loading, you need temperature-dependent material data. Using room temperature properties for an analysis at 300 degrees Celsius is not an approximation. It's a fundamentally wrong model. The same applies to cryogenic conditions where some materials become brittle and the concept of plastic deformation changes entirely.

Limitations and Where the Method Fails

Inelastic stress analysis has real limitations that every engineer should understand before relying on it. The first is that it assumes the material model is accurate. If your stress-strain data is incomplete or represents a different condition than your actual component, the results are only as good as the input data. There's no mathematical trick that can compensate for bad material properties. The second limitation is computational cost. A full nonlinear inelastic analysis with contact, large deformation, and a multilinear material model can take hours or days for a moderately complex model. This makes it impractical for iterative design work where you need to evaluate dozens of variations. In those cases, elastic analysis with plastic collapse allowables from recognized codes is the more efficient path. The third and most important limitation is that inelastic analysis does not automatically mean your design is safe. Passing an inelastic analysis only shows that the structure can sustain the load through plastic deformation. It doesn't address buckling, fatigue, fracture, or environmental degradation. A component can have excellent plastic collapse resistance and still fail prematurely due to low-cycle fatigue if the plastic strains are cyclic. I once saw an analysis where the inelastic results looked perfectly fine for a seismic event, but the strain range over repeated cycles would have caused failure within a few hundred events. The fix wasn't a stronger material. It was adding flexure joints to reduce the strain demand in the critical regions.

For high-cycle fatigue applications or components subjected to repeated thermal cycling, elastic analysis combined with fatigue methods often provides a more relevant answer than inelastic analysis. The plastic strains per cycle are small, and the cumulative damage is better predicted by elastic stress ranges and S-N curves. Running a full inelastic analysis in those cases adds computational expense without improving prediction accuracy. The practical takeaway is that both methods are tools, not truths. Elastic analysis is fast, reliable, and sufficient for the majority of designs. Inelastic analysis is necessary when plasticity is expected and you need to understand the actual collapse mechanism or load redistribution. Knowing when to switch between them and what each method cannot tell you is what separates an engineer who runs simulations from one who understands what the simulations mean.