Structural modelling process
Figure 1 is a diagram of the structural modelling process
The rectangular boxes represent processes and the elliptical boxes represent outcomes from processes.
The outcomes are:
- The system model is the design information for the complete real world system. This is traditionally in the form of drawings and specifications but also includes BIM models.
- The engineering model is a subset of the system model that defines the features of the structure needed for the technical assessment.
- The analysis model is a mathematical model of the structure represented by diagrams, model specifications and data for the software.
- The solution is the raw output from the computer.
- The accepted output is the part of the output that has been assessed and approved for use in the technical assessment process.
Modelling activities
Figure 2 shows a range of modelling activities.
Requirements assessment
While the requirements for a structural analysis are normally evident, it is important to reflect on them and to draw up a requirements statement for complex or non-standard contexts.
Questions for requirements assessment include:
- Stresses and internal forces: Have the locations and target accuracy been identified? Has the distinction between local and resultant stresses been clearly identified?
- Displacements: Where do the displacements need to be defined and to what accuracy?
- Natural frequencies and mode shapes: What range of natural frequencies need to be considered?
- Fundamental requirements assessment question: Have all required performance issues been identified and included in the requirements statement?
Model validation
Model validation is the process of ascertaining whether the model is capable of meeting the requirements.
A model validation is carried out by listing all the assumptions made for the model. To do this, one needs validation information i.e. information which discusses the applicability of assumptions. Such information is not readily available in conventional texts on structural analysis - but see Reference 1.
It is good practice to prepare a validation analysis for non-standard contexts. Ways of presenting these are given in references 1 and 4.
While having an initial validation exercise is very important, validation information can emerge from the results especially from sensitivity analysis. A constant lookout for information that will assist the validation should be held.
Questions for validation include:
- Assumptions: Have all the assumptions for the model been identified?
- Validation information: Is all relevant validation information available?
- Test results: Are there test results available to support the validation?
- ·Fundamental validation question: Is the model capable of satisfying the requirements?
Results verification
Results verification is the process of seeking to ensure that the model has been correctly implemented.
Both formal and informal strategies should be used to seek to identify faults in the implementation of the model. A formal checklist can be used but also regular quantitative and qualitative checks should be carried out.
Verification checks include:
- Data checking: Has there been sufficient resource applied to data checking?
- Overall equilibrium Has a check been made on overall equilibrium? For example the sum of the total vertical reactions should be checked against the total vertical load that was expected to be applied to the model.
- Symmetry: If there is symmetry has it been checked via a symmetric loadcase?
- Form of results: Does a qualitative assessment of the results show any anomalies?
- Values of results: Are the values of the results in the expected range?
- Checking models: See following paragraph.
- The fundamental verification question: Has adequate resource been allocated to minimise the risk of implementation errors in the model?
Checking model
Checking the model against another frame of reference - a checking model - can be a valuable review activity. The checking model may take the form of:
- A ‘back of an envelope’ calculation i.e. a hand calculation based on a simplified model of the system to provide a quick check
- A simplified model of the system which requires software for the solution.
- A repeat of the model using different software and/or different personnel
Simplified models should be assessed for validity. It is important that they are able to adequately represent the main features of behaviour being investigated. Also, care should be taken in relation to correlation between two models. False correlations are not uncommon.
Results interpretation
Results should be regularly interpreted to seek to develop understanding of the behaviour of the system. This can contribute to the validation and verification processes.
Sensitivity Analysis
Varying the model to assess the effect of the variations on behaviour is a very important review activity. Doing this helps to develop understanding of behaviour which can inform both the validation and verification processes.
Competence assessment
It is important to ensure that those using the software have the necessary competence.
Software assessment
Software validation The software should be validated by asking the question: Is the software capable of implementing the model?
Software verification The software should be verified by asking questions such as: Is there adequate evidence that the software has been checked for accuracy? Have benchmark analyses been carried out?
Modelling Review Process
Figures 3 shows a basic review process and Figure 4 represents a full review process. The former would be used in a standard context that is familiar to those working on the type of model. The latter would be used in an innovative situation where there might be significant uncertainty, complexity and/or where the context is safety critical - e.g. nuclear power installations. Results verification is always needed but in standard contexts validation may not need deep consideration and sensitivity analysis may not be required. In a complex innovative situation very careful attention to all the review activities may be essential.
The initial Review
Activities for an initial review may include:
- Make predictions about the expected outcomes at the outset and check these against the results as they emerge from the process
- Define the requirements for the model
- Assess competence and software
- Carry out an initial validation of the model.
- Set up the data with a simple loadcase (a checking loadcase) and carry out a results verification.
Ongoing review
Continue with development/production runs.
- Do quick qualitative verification on a regular basis and further quantitative checks as appropriate.
- Continue to consider model validation issues if the model is altered and to interpret validation information from the results if practical.
- As appropriate, carry out sensitivity analyses to develop understanding of the behaviour of the system being modelled and to contribute to the validation analysis.
- Continually challenge the adequacy of the inputs to and the outputs from the process.
Final review
Carry out final versions of the model validation and results verification and record the results as appropriate to the required QA procedures. Assess the outcomes against the initial predictions and seek to identify the sources of important differences between the two.
BibliographyMacleod I A (2005) Modern Structural Analysis Thomas Telford Ltd.
Macleod I A and Weir A (2016) Principles for Computer Analysis of Structures, Essential Knowledge Text No 14. Institution of Structural Engineers