A structural engineer performs FEA to predict the approximate response of a structure under a given excitation, so a legitimate concern is the accuracy of this prediction.
First, the accuracy of the prediction will depend on the differences between the real structure and the finite element model.
Obviously, a perfect match between the reality and the model is impossible because many assumptions are made when the structural model is built:
Each of the above assumptions will produce a deviation between the real structure and the finite element model. However, there is really no way to determine the real errors in the analysis.
You can attempt to quantify the assumptions of the FEA to obtain a relative error. For example, analyzing bounding ranges of material properties would introduce the error in the material response.
The largest error in most FEAs is the result of bad boundary conditions. Incorrectly supporting your model can result in large errors that are transparent to the user.
FEA can yield spectacularly accurate results when compared with physical test results.
However, in a complex FEA, it is still not possible to obtain an error of less than 1% in the whole model. Generally, a target of ±10% global error allows for very good predictions, but this target does not prevent the occurrence of errors of less than 5% in some local regions of the FEA.
To conclude on accuracy, keep in mind that to produce a highly accurate FEA, the analyst has to pay close attention to many points and perfect them, especially if the model is composed of several assembled parts. A single component model will easily match real life, while complex assemblies will require more effort.
First, the accuracy of the prediction will depend on the differences between the real structure and the finite element model.
Obviously, a perfect match between the reality and the model is impossible because many assumptions are made when the structural model is built:
- A first assumption is made on the CAD when it is discretized by creating meshing of the different parts that compose the structure.
- Assumptions are made for the modeling of joints.
- Material models are selected.
- The real loadings are discretized to make them understandable by the Finite element model.
- Assumptions about boundary conditions are made.
- The analyst selects the behaviors the FEA will capture.
Each of the above assumptions will produce a deviation between the real structure and the finite element model. However, there is really no way to determine the real errors in the analysis.
You can attempt to quantify the assumptions of the FEA to obtain a relative error. For example, analyzing bounding ranges of material properties would introduce the error in the material response.
The largest error in most FEAs is the result of bad boundary conditions. Incorrectly supporting your model can result in large errors that are transparent to the user.
FEA can yield spectacularly accurate results when compared with physical test results.
However, in a complex FEA, it is still not possible to obtain an error of less than 1% in the whole model. Generally, a target of ±10% global error allows for very good predictions, but this target does not prevent the occurrence of errors of less than 5% in some local regions of the FEA.
To conclude on accuracy, keep in mind that to produce a highly accurate FEA, the analyst has to pay close attention to many points and perfect them, especially if the model is composed of several assembled parts. A single component model will easily match real life, while complex assemblies will require more effort.
