Category "Dassault Systemes"

In previous blog article on simulation in collaborative environment, we introduced SIMULIA 3D Experience platform and discussed the concept behind its inception, the reason why it exists in the industry and briefly discussed its four integrated components: 3D modeling, information intelligence, social & collaboration and simulation. All four of these collectively form 3D Compass. This blog article explains the configuration of simulation component in more detail.

In general, 3D Experience platform uses few specific terms with respect to its configuration: Personas, Roles, Apps and Extensions. These terms primarily govern how various platform functionalities are bundled, licensed and made available to users.

Personas: Defines the job responsibilities of a group of users. Every user has at-least one persona. Configuration requires estimation of number of personas and number of users in each persona.

Roles: Based on job activities, each persona has to be mapped with a specific role. The platform offers a library of different roles. Each role is a bundle of sellable license features or apps.

Apps: An app can be defined as a group of functionalities to achieve a specific task. For example, a fluid model creation app offers multiple GUI features to define fluid model such a fluid domain creation, fluid mesh, boundary layer mesh, fluid material definition etc.

Extensions: These are a bundle of top-up apps that can be added to a given role to enhance its overall functionality.

Role Categories

Remember that in case of standalone point solutions, we discussed two broad categories: designer level solutions that are primarily CAD embedded vs. expert level solutions with their own graphical user interfaces and complex workflows. In 3D Experience platform as well, Simulation roles can be differentiated into engineer roles and analyst roles. There is also a third category of roles called as process roles. Each of these types of roles require some basic platform roles as pre-requisites.

Extensions vs. Roles

Each extension can be ideally treated as a mini role. Extension, if compatible with a given role can be added to it to enhance its functionality without duplicating any app available in the role. Here is an example for demonstration.

Mechanical Analyst and FEM Specialist are roles with number of apps as shown above. Now if an SMU user needs SIMULIA model assembly design app, there are two ways to do it. An expensive approach would be to procure entire SFM role that would not only increase cost substantially but would also duplicate apps common between SMU and SFM. An economical approach is to add the SMA extension with only relevant app to the SMU role.

Simulation capacity: Tokens vs credits

A token is a governor of maximum amount of simulation that can run concurrently. More tokens mean more concurrent simulation. A token is a renewable simulation resource. Once simulation is complete, tokens are returned to the token license pool. Jobs can be submitted either on premise or on cloud. Both these options have different token categories.

A credit is a consumable computation resource that is not replenished once the simulation is complete. There is no limit on how fast a credit may be consumed during concurrent simulation. The user procures compute credits that gets consumed as simulation progresses. The rate of credit consumption is directly related to speed of simulation. Quite often, analysts prefer to use tokens as one-time investment. The credits are used to meet occasional peak demand when tokens are not enough to meet simulation capacity needed.

In future blog articles related to 3D Experience platform, we will discuss various roles available for stress analysis as well as their underlying apps, extensions and simulation capacity.


The definition of warranty varies depending on the nature of product being used by the consumer. For a product as intangible as an insurance policy, warranty may be defined as percentage of eligible claims settled by the insurance company in given time-period. However, manufacturing companies produce tangible products. In such scenarios warranty is almost always defined either in terms of time or in terms of in service load cycles product is expected to survive. Both parameters are often related to each other depending on the frequency of use. Moreover, in service load values varies a lot from one geography to the other as well as from one user to the other. Have you ever come across a car that is always driven on roads of a constant surface quality? Have you ever seen two drivers who maneuver their cars in a perfect identical fashion? If not, there is a statistical aspect to be considered as well. Nevertheless, customer do not realize how difficult it could be to define a warranty of a physical product. They treat it as a simple number on warranty card and if the product fails before the warranty expiry period, they feel cheated and their response is often something like this:

It is true that several critical parameters should be considered to define warranty of a physical product with precision. Many of such parameters are either very difficult to measure in house or to collect from external resources due to their seasonal, spatial or statistical nature. Thus, it is impossible to narrow down warranty failure instances to absolute zero. However, it is possible to reduce cost and unpleasant experience associated with such instances by reducing their frequency.

In case of a physical engineering product, often used in transportation, aviation or healthcare industry, virtual testing for durability is an effective method of containing warranty costs. Here we are talking about using finite element analysis approach using FEA codes such as Abaqus in combination with durability codes such as fe-safe.

In any given fatigue analysis workflow, a structural FEA solver as well as fatigue solver are present. The output from FEA solver serves as one of the input for fatigue solver. The FEA simulation is carried out either by applying a unit load or the entire variable load depending on the nature of problem. The minimum needed output may be either principal stresses or combination of principal stresses, principal strains and temperatures depending on the physics of the problem. The second input for fatigue solver is the in-service load data files and it may be optional in certain cases. These are real time digital files that capture fluctuation of loads in different directions over a given time. They are created using data acquisition techniques and are compatible with well-known fatigue solvers such as fe-safe. Once they are entered in the workflow, they serve as a multiplier to respective unit load data set to generate a 3D stress cycle. Both uniaxial and multi axial load scenarios are supported along with multiple block loadings.

The third input is the fatigue material properties. This could be either a stress-life curve or strain-life curve depending on physics of the problem any type of fatigue algorithm used. A good news is that fe-safe has a well-defined material databank of commonly used metals and alloys. If needed, this material data can be customized.

Once these three inputs are defined, fatigue problem definition is complete and solver is executed to provide the output in desired forms: cumulative damage, damage per block, cumulative life in terms of hours/days/load cycles etc. The numeric values are all printed in text files while a fatigue contour can be seen in a binary file compatible with most FE post processors.

An advanced fatigue workflow could be one involving fatigue optimization. In this workflow, it is possible to define fatigue as one of design response that could be minimized using shape optimization code such as Tosca. It is further possible to incorporate various types of manufacturing constraints in such an optimization.

Though a fatigue simulation workflow is well defined, it is not easy to execute. It is still beneficial to adopt this methodology because physical testing is often time consuming and manual hand calculations are not valid for complex loading cycles. If customers understand the complexity involved, they may be able to accommodate margin of errors in product warranty cards.

Computational fluid dynamics role of 3DX is the one going through tremendous enhancements compared to other roles available on the platform. In this blog article, I am going to highlight a few key enhancements with respect to the scenario modeling and the underlying solver.

Multiple reference frames

Most of CFD users are familiar with concept of moving fluid boundaries. Traditionally these problems have been solved using coupled Eularian – Lagrangian (CEL) techniques. In simple words this technique can be defined as a combination of two fluid spaces: one near the boundary in which fluid moves with the mesh and other away from the boundary in which fluid moves through the mesh. A whole new concept has been introduced in 3DX CFD to solve such problems. It is moving reference frames that can either translate or rotate with respect to a global reference frame. Either entire fluid domain or a portion of it can be assigned to this frame of reference. The governing equations are solved in this reference frame. Interfaces are created between moving and stationary frames to maintain motion continuity.


Compressible flows

Compressible flows become a concern when speed of flow is high. While this may not be relevant for companies designing exhaust manifolds or valves, an analyst trying to study the exterior air flow drag on a fighter jet moving at one tenth the speed of sound would feel the need of compressible air flow. The new release can simulate transonic flows up-to Mach number of 1.2.

Modeling of porous media

Are you looking to model components such as catalytic converters or air filters? Well, these products have a permeable medium that allows restrictive flow of air through it. Modeling such flows require porous media functionality that is now supported in 3DX CFD.

These along with many other enhancements are now a part of 3D experience 2018x platform. If you wish to know more, please feel free to contact us.

When it comes to Abaqus structural solver, picture is clear. There is a standard (implicit) solver as well as an explicit solver. Each of these has its own merits and demerits that we have discussed in previous blog articles. However, in CFD there appears to be only one solver. So…

Is it implicit or explicit!!

Well, if you look at the underlying parameters of the solver, it appears to be hybrid. The solver talks about inner loop and outer loop convergence. That makes user feel that solver is implicit and it requires matrix based calculations. This is true for both steady state as well as transient flows. But then when we talk about transient flow parameters, solver mentions CFL number that primarily governs the step size of transient flow. Higher CFL number results in larger step sizes but beyond a certain value of CFL, the flow may become unstable. This looks more like an explicit scheme where stability plays a role. But then, the transient flow also requires convergence that is not an explicit property. So where do we stand.

The answer is somewhat mixed. The CFD solver of Abaqus is implicit by nature. However, it does not follow all the traits of structural implicit solver. One big difference is that CFD implicit solver is not unconditionally stable. While the explicit structural solver just exits if time increment exceeds its critical value, the CFD solver continues at larger than desired CFL numbers but it may give non-realistic flow results. The other big difference is the physical quantities involved. The structural implicit solver looks at force and displacement residuals for convergence. The CFD solver looks at momentum, pressure and velocity residuals.

SIMULIA has made good efforts in 2018x release of CFD solution on the 3D Experience platform in terms of making the solver fully implicit so that it can handle large time increments. There have been other solver enhancements to improve accuracy and reduce solution time. Wish to know more about SIMULIA CFD techniques! Please get in touch.

In recent blog article on friction, I discussed about a new Abaqus functionality that allows user to define friction as surface property and Abaqus computes contact pair friction coefficients from corresponding surface friction properties. In this blog article we discuss yet another nice and recent functionality in Abaqus explicit called anisotropic friction.

The anisotropic behavior may arise from number of scenarios the most common of which is composite material that has longitudinal and transverse fiber directions. In such a scenario, the coefficient of friction between contact pair depends on the relative direction of sliding between the contact surfaces. Looking for a real-life example!!!

“The interaction of seat belt with the occupant body is an example of anisotropic friction”

he above figure shows the concept pictorially. Blue arrows indicate the direction of relative sliding. Hence these arrows are always at an angle of 180 degrees. The red lines show the direction of primary material axis. Theta is the angle between blue arrows and red lines per surface. The directional friction stress is computes as:

Both anisotropic friction as well as estimation of friction interaction from surface property are in the category of “combinatorial rules” and both are controlled by same keyword entry as follows in the .inp file.

If both the nominal friction and directional preferences are to be determined from surface property, it is not necessary to define *friction keyword.


In this article we are going to discuss an advanced friction modeling technique in Abaqus. It is based on combination rules that allows solver to compute effective friction interaction based on two contacting surfaces with different coefficients of friction. As an example, look at the following table:

If someone asks: “what is the coefficient of friction of steel?” There really is no answer to this question. The answer really depends on the other object with which steel interacts. The table shows two different values, one for steel-steel interaction and other with steel-teflon interaction. If the user has NXM matrix of materials interacting with each other and each cell of that matrix has a friction coefficient assigned to it, then modeling in Abaqus is trivial. Define surface interaction with friction coefficient for each cell and use it with corresponding surface pair in the contact property assignment. The example below highlights it.


But this straightforward approach is possible only if friction values for all cells are available. However, at times only the diagonal values are available. That means all the non-diagonal cell values are unknown. In that case contact property assignment is not possible.

Abaqus now allows users to define friction as surface property as well. For two different surfaces (A,B) with individual coefficient of friction, the effective friction for pair is computed as follows:

The default value of alpha is 0.3. In case of mixed problems, where surface property and contact property methods co-exist, either method can take precedence. Look at following example.

The approach is an approximation but its worth in situations where user has no access to friction coefficients values for all the contact material pairs. This friction algorithm is available in Abaqus explicit 2018 release and beyond.


This topic has always been very popular and this problem has always been very complicated in FEA user community since the inception of Abaqus, or any non-linear FEA code in general. In this brief article, I will highlight few simulation situations where Abaqus standard may not be a good candidate from convergence perspective. Identifying these situations early during pre-processing and working in Explicit right away may save lots of time and efforts that otherwise would be wasted in trying Abaqus Standard.

  • Look at the motion aspect: We always say that simulation is not the complete replacement of physical testing right away. In the beginning physical tests play a critical role in identifying right approach for simulation as well as in data correlation between physical and virtual tests. Look closely at the physical test. Is there a large relative motion between different parts involved? If yes, then Standard is very likely to face convergence problems, even if problem is static by nature. Standard has an option of “small sliding” and “finite sliding”. But user should remember the difference between “finite sliding” and “large sliding”. Attached is the video of wire crimping simulation that ideally is a static problem but numerically not a good candidate for Standard, primarily because of motion.
  • Clock time matters: Apart from magnitude of motion, the duration of motion matters as well. While looking at physical test, closely look at the time in which motion is completed. If too much of motion is covered in too less time, problem is indeed dynamic instead of static as inertia effects cannot be ignored. In such a situation either Standard dynamics or Explicit would be the right way to go. Which one to choose really depends on event duration. If a lot of dynamic phenomenon happens in the order of milliseconds or microseconds, Explicit is only option for this candidate.
  • Is there a severe discontinuous contact: In the status file of Abaqus Standard, there is an undesirable column called SDI’s. It’s called severe discontinuous iterations and too many of these often always leads to convergence nightmare. The reason of SDI’s is discontinuous contact, also known as “chattering”. It’s a phenomenon in which nodes between two bodies in contact continuously change their contact status from OPEN to CLOSE from one iteration to the other as analysis proceeds. If chattering occurs due to modeling errors, it can be corrected but at times discontinuous contact is the nature of problem itself. In such a situation, explicit is the only approach to be taken, even for long duration events with respect to physical time. The attached video is an example of a dynamic event that would only solve in explicit or multi body dynamics, primarily because of severe discontinuous contact.
  • Is there too much Plasticity: Abaqus has material models to capture plasticity but there is a limit on the magnitude of Plasticity Abaqus Standard can handle. If the permanent deformation becomes so high that underlying part completely loses its load carrying capacity then Newton Raphson method of Abaqus Standard would not be able to establish equilibrium and further leading to non-convergence. Ideally, there is no further need to perform simulation as it’s a classic situation of part failure but if further simulation is needed, it should be continued in Explicit using Restart options.

In previous blog articles on 3D Experience simulation roles, we primarily discussed platform configurations, concept of personas and roles as well as simulation capacity of the platform. In this blog article contains detailed information about three primary structural simulation roles: MDS, DRD and SMU.

To begin with, lets recapitulate that simulation roles are categorized in groups based on personas of users working on such roles. In terms of complexity and functionality, offerings range from based to intermediate to advanced.


Engineer profile: The is the simplest and easiest to use simulation offering primarily meant for designers with low to intermediate simulation knowledge. Their primary job is product design and they perform simulations very occasionally. Roles for this profile are CAD centric and are associated with a guided workflow. Simulation tokens are embedded in the role.

Analysis engineer profile: This profile is one level above the engineer profile and is suitable for structural analysis engineers associated with product engineering. Their simulation knowledge is of intermediate level which means they understand simulation process in terms of meshing, BC, Loads, result visualization etc. but don’t have any hands-on experience of advanced simulation tools. Usually there is no guided workflow. Simulation tokens are embedded in the role.

Analyst profile: This role is for full time analysts who primarily perform intermediate to advanced level simulations. They have in depth expertise in at-least one simulation domain and often hold Masters or Doctorate level credentials. This role requires extensive knowledge of pre-processing, solver terminologies such as statics, dynamics, non-linearity, convergence schemes, as well as post processing etc. There is no guided workflow. Simulation tokens are procured separately.

 Research Specialist profile: This is a complex simulation offering primarily for experts who develop novel simulation workflows and processes. The simulation requirements often span across multiple physical domains and involves advanced Physics such as vibrations and noise. The pre-processing aspect may include complex meshing of assemblies and assemblies of meshes.

Let’s look at one role from each of first three profiles:

Stress Engineer role (MDS)


It’s a role from engineer profile and has a guided workflow. The snapshot shows apps available in MDS role. It performs routine strength and deflection calculations under static loading conditions. It can also compute product fatigue life for very simple loads. The CATIA and SOLIDWORKS associativity is well maintained. Local solver execution up to 4 cores is included.

Structural analysis Engineer role (DRD)


It’s a role from analysis engineer profile that has no guided workflow. It is used to access the structural integrity of products subjected to wide range of loading conditions. The snapshot above shows available apps in this role. It works on MSR concept available in advanced simulation tools i.e.  Model-Scenario-Results. Many advanced settings are exposed to the user. This role can perform multi step simulations. Local job execution of up to 8 cores is available.

Mechanical Analyst role (SMU)


It’s a role from analyst profile and it does not include a guided workflow. The snapshot above shows available apps in this role.  It uses advanced finite element techniques to simulate and validate complex engineering problems. It offers multiple advanced meshing techniques such as Octree, surface, sweep and RBM. Both single step as well as multi step scenarios are included. Supported analysis steps include static perturbation, non-linear static, frequency, buckling, implicit dynamics, explicit dynamics, steady state heat transfer, transient heat transfer etc. Most of the non-linear materials and complex engineering connections are included.

While we discussed one prominent role from each profile, the south quadrant of 3D Experience platform offers numerous simulation roles. To know more, please contact us.

One thing common between SIMULIA roles of 3DExperience platform and the standalone Abaqus products is that both require an Abaqus solver to perform computations. It further means that both solutions require Abaqus tokens to complete or speed up the computation part of the simulation. For standalone abaqus product, we know that the calculation is straight forward. Abaqus requires a minimum of five tokens to execute a single core non-linear job. Large models require more number of cores to solve in real time and more number of cores require more tokens as follows:

The computation capacity of 3D Experience platform, however, cannot be defined by a single equation. Unlike Abaqus solver, that is available as an integrated all-in-one license for all types of simulations such as standard, CFD, explicit etc., 3D Experience offerings are in form of roles. Each role is a sellable license that includes either some or all Abaqus solver capabilities. Offers are made further flexible by on premise vs. on cloud offerings. Let’s have a look at solver offerings in different configurations and roles.

    Engineer role vs. Analyst role

While most of design engineer roles have embedded Abaqus tokens, most of the analyst roles do not have any compute capacity at all. The number of tokens embedded in designer role depends on the level of simulation complexity a role can accommodate. For example

Stress Engineer role has 8 embedded tokens to accommodate up-to 4 cores job

Structural analysis engineer role has 12 embedded tokens to accommodate up-to 8 cores job

It is possible to submit jobs on more number of cores than what embedded solver permits but in that situation external tokens need to be utilized and embedded solver takes no credit at all.

                Tokens vs. Credits

In case of analyst roles such as stress analyst, fluid mechanics analyst etc., the role itself does not have any compute capacity which should be procured either in the form of tokens or credits. Tokens are renewable form of compute capacity which means they can be used over and over. 3D Experience uses tokens in a very similar fashion as does standalone Abaqus. The token consumption with respect to number of cores is the same for Abaqus as for 3D Experience platform. On the contrary credits are a non-renewable form of compute power. It means that credits, just like the talk time over phone, can be consumed only once.

               Why credits at all!!!

In general credits is an expensive preposition for customer but there are exceptions. Credits are utilized to meet unexpected and rare increase in peak usage. This is somewhat more common in engineering consulting firms that can face high demand of simulation capacity due to influx of many short duration simulation projects at any time. To meet this sudden spike in demand, one-time credit bundle offering makes more sense than increase in perpetual tokens. Once peak demand is over and credits are consumed, simulation capacity is returned to normal levels.

On premise vs. on cloud

Design engineer as well as analyst roles are available in on premise as well as on cloud formats. There are three ways of utilizing cloud resources: store the models on cloud, stores the results on cloud and solve on cloud. The first two offerings require only cloud storage and are available at no additional charge with cloud based license. However, the third offering requires cloud compute resources that consumes compute credits in addition to cloud based license.

Need to know more about SIMULIA 3D Experience platform compute capacity! Please approach us and we are ready to help.


Organizations invest huge sums of money in simulation software to avoid expensive and disruptive physical testing processes. But how long it really takes to make this transformation happen! One thing is sure; it does not happen in a day. The flow chart below explains the reason pictorially. The last two blocks “compare and improve model” and “compare and improve theory” make this transformation a longer process than expected.


Let’s explore the reasons behind it. Comparison is needed to make sure that simulation results mimic the physical testing results before latter can be discarded, partially or fully. The difference in results can be due to three main factors: lack of user competency, limitation of software used, lack of sufficient input data.

Lack of user competency: FEA analysts are not born in a day. The subject is complex to learn and so are the software associated with it. The ramp up time really depends on analyst background along with complexity of problem being simulated. Organizations usually make a choice between hiring expert and expensive analysts who can deliver the results right away or producing analysts of its own through class room and hands on trainings. First option saves time while the second saves money. CAE software development companies are also making big stories these days by introducing CAD embedded simulation tools that require nominal user competency. Nevertheless, the competency builds up over time.

Limitation of software used: Initial investment in simulation domain is usually small. It means two things: either number of users are less or software functionality is limited. With time, complexity of problems goes up but the software remains the same. A common example I have seen is of a customer starting with simple linear simulation workbench in CATIA and over period trying to simulate finite sliding contact problems with frictional interfaces in the same workbench. Users don’t realize that their problem complexity has exceeded the software capacity to handle and it’s time to upgrade. It’s always recommended that analysts get in touch with their software vendors whenever they anticipate an increase in simulation software capacity or functionality. A certified simulation software vendor is a trusted advisor who can really help.

Lack of sufficient input data: “Garbage in – Garbage out” is a very common phrase in simulation world. However, at times it is very difficult to get the right input for software in use. The complexity of input data can arise either from complex material behavior or from complex loading conditions. Example of complex material may be hyper-elasticity or visco-elasticity observed in elastomeric materials. Examples of complex loading may be real time multi block road load data to estimate fatigue life. Sometimes simple metallic structures exhibit complex behavior due to complex loading. Examples are high speed impact or creep loading. With time many material testing labs have come into existence that can perform in house testing to provide right input data for simulation.

Conclusion: You will come out of the vicious loop of physical and simulation results comparison after couple of iterations if you have three things in place: right people, right software product and right input data. If you need help in any of the three aspects, we are always available.

© Tata Technologies 2009-2015. All rights reserved.