Verification and validation are the essential procedures required to assess accuracy and credibility of numerical analyses. The terms “verification” and “validation”  are often used interchangeably in the current engineering practice. However, the two procedures have entirely different meanings and applications. The differences are outlined below:

Verification is meant to identify and remove programming errors in a computer code and verify numerical algorithms. It deals with the mathematics, only.

Validation is meant to assess the accuracy at which a numerical model represents reality and includes the essential features of a real model. Unlike verification, validation deals with the physics.

As an example, let’s assume that you have implemented Mohr-Coulomb constitutive model in a finite element (FE) computer program. You need to verify the program to ensure that the developed code is correctly incorporated into the FE program and its mathematics is in entire accordance with the theory. Thereafter, you need to validate the program to ensure that the model is capable of simulating the reality which, in this example, is soil stress-strain response.

Generally, the end users do not need to verify commercially available softwares like PLAXIS, FLAC, ABAQUS, etc. It is the responsibility of the program developers to take care of the verification and ensure that their product is mathematically correct and is free of programming errors (so-called bugs). On the other hand, validation should be generally performed by the end user. It is the primary responsibility of the end user to create a numerical model that represents the real physical model by adopting appropriate boundary conditions, constitutive models, elements, etc.

Validation of a finite element and/or finite difference program includes (Brinkgreve and Engin, 2013) :

• Validation of constitutive models and parameters: The capability of the model in simulating stress-strain responses must be evaluated by modeling lab test results. For instance, Oedometer test results if the soil model is meant to simulate consolidation. In this process, soil parameters are iteratively varied to make a “best fit” to the test data. This exercise is often referred to as calibration of constitutive models.
  XXXX
• Validation of boundary conditions: Lateral and base boundaries are required to limit the extents of a finite element/difference model and to optimize the analysis execution time. Validation is required to ensure that the selected boundaries do not influence the output of the analyses.
  XXXX
• Validation of meshing and spatial descretization: The finite element mesh has to be sufficiently fine so that the analysis outputs are remain about the same when  a finer mesh is analyzed.
  XXXX
• Validation of time-integration steps (for dynamic analysis): The time step must be sufficiently small to ensure that the analysis outputs do not change considerably by using a smaller time step.
  XXXX
• Validation of analysis outputs: After each of the model components, as listed above, is individually validated, it is necessary to assess the validity of the entire numerical model by comparing its results against the measurement of a real physical model. Center for geotechnical modeling, at University of California, Davis has provided a valuable database for various types of geotechnical systems. The experimental data (digitized format) and documentations of all centrifuge tests since 1997 are available online here in this link.

Experimental test results of three different soil conditions are uploaded herein to facilitate validation of the numerical models that are developed for free-field site response analysis in level grounds:

Dry sandy soils (Centrifuge test by Gohl, 1991):
Gohls-Test-1991-Test-12.zip

Saturated sandy (liquefiable) soils (Centrifuge test by Wilson et al,1997):
Wilsons-Test-1997-CSP-2_Event-F.zip

Soft clay (bay mud) underlain by saturated sandy soils (Centrifuge test by Wilson et al,1997 ):
Wilsons-Test-1997-CSP-5_Event-C.zip

Note: The three centrifuge tests referenced above were primarily meant to analyze seismic soil-pile interaction. Single piles  as well as pile groups were present in the tests. However, measurements at the free-field (far from the piles) are of interest herein. The measurements provided in the “zip” files above are those recorded sufficiently far from the pile elements.