using abaqus in a manufacturing environment at honeywell federal

2004 ABAQUS Users’ Conference 1

Using ABAQUS in a Manufacturing Environment at Honeywell Federal Manufacturing & Technologies*:

The Quest for Quality Parts Under Small Lot Production Scenarios

James F. Mahoney, Jr.

Honeywell FM&T

Abstract: The diversity and complexity of current product designs require skillful manufacturers to produce the needed small volume of parts effectively. Traditional methods have been to do trial-and-error to produce good quality parts. Costs and flow times associated with prototyping are skyrocketing. Survivability for manufacturers is at stake. Computer simulations are key to solving production development issues facing industry today. The aid of high-performance computing simulations has greatly enhanced the manufacturing expertise and knowledge at Honeywell Federal Manufacturing & Technologies. The simulations augment the prototyping and development stage and aid in solving production problems. Keywords: Manufacturing, encapsulation, welding, forming, failure mechanics.

1. Introduction

Honeywell FM&T is a supplier of high-tech manufactured goods to the Department of Energy’s National Nuclear Security Administration defense systems. Honeywell is responsible for a wide variety of components and systems ranging from machined cases, forged metal parts, encapsulated high-voltage electronics, guidance and radar electronic systems, and individual plastic and mechanical parts. The volume of manufacturing is very low, on the order of 10–1,000 pieces per build cycle. To maintain the high quality of parts within the diversity of manufacturing, prototyping is necessary. With the low volume of required parts, the challenge faced is to fully understand new and existing manufacturing processes to produce newly designed components and systems.

* Operated for the United States Department of Energy under Contract No. DE-ACO4-01AL66850. All data prepared, analyzed, and presented has been developed in a specific context of work and was prepared for internal evaluation and use pursuant to that work authorized under the referenced contract. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government, any agency thereof, or Honeywell Federal Manufacturing & Technologies. Copyright © Honeywell Federal Manufacturing & Technologies, 2004

2 2004 ABAQUS Users’ Conference

Traditionally, a manufacturing process was evaluated for specific applications by both machine craftsmen and process engineers. The final output quality was formulated, in part, by the know how (or “art”) of the understanding of the specific processes. To be maintained as a leader in defense-related parts requires a high level of confidence in process operations. The combination of an aging population of experienced personnel with new “computer literate” engineers integrated with simulation solutions is the key to future success.

The objective of this paper is to illustrate the methodology and approach to integrating ABAQUS Finite Element Method (FEM) solutions into a diverse manufacturing framework. This paper will overview the challenges that Honeywell FM&T has faced over the past 20 years and suggest and forecast directions for the simulation business for the next decade. Some specific examples will be noted where traditional manufacturing is no longer able to cope with the complex receipt for success and where simulations compliment traditional production methods.

2. Science-Based Manufacturing Approach

During the mid 1980s, manufacturing work was flourishing at the Kansas City Plant. The development of both new defense systems and the maintenance of existing systems were at an all time high. The product line of mechanical, electrical, and plastics required a diverse production environment. Because of the small volume of each unique part, development costs amortized over the systems were becoming excessive. The Kansas City Plant has the responsibility of producing the wide diversity of parts at all levels of production quantities. The level of security on the parts and systems also required a closed-door environment, even to many downstream component suppliers. The traditional role of ABAQUS [1] simulations was on the evaluation of manufacturing processes that were difficult to understand. The art in manufacturing and the craftsmanship was still required, even though many operations were changed to automatically controlled systems. The population of experienced personnel was stable and aging. As time progressed to the 1990s, the cost or value of the production work was given tighter controls. The usage of Components Off The Shelf (COTS) was driving the electronics industry. The requirements for new strategic defense systems had diminished. It was important to “get it right the first time” on production builds. The formula for production success had to change. The mid to late 1990s was a time of survival, a time to regroup and evaluate production, a time to keep core processes, and a time to be reactive and agile to change. The focus of simulation, particularly with the FEM-based tools such as ABAQUS, had to take a leading role. The birth of the Science-Based Manufacturing (SBM) approach to production occurred in the late 1990s. The focus and direction were paramount : in effect, shift emphasis on utilization of the simulation tools to the front-end of process development, augment the work with the knowledge capture of the highly experienced workforce, and yield high quality parts, using COTS, in a shorter amount of time. The Department of Energy [2] suggested this approach to Honeywell FM&T via its broad definition:


“Science-Based Manufacturing is our strategy to use scientific, engineering, and computational tools to allow rapid realization of defect free product to support the nuclear weapons stockpile.”

Specific to the Kansas City Plant was the focus on applying simulation tools to solve single and multi-physics manufacturing problems. From the high level strategy, a more specific concept was fostered for the SBM principles:

An infrastructure and tool set that will allow for a more comprehensive approach to utilizing simulations before full production commences and design/manufacturing decisions are set.

This tool set included a vast array of software tools, with ABAQUS being in the forefront. The approach to favor preproduction appeared impossible compared to the existing challenges in current production operations. The strategy and approach needed additional resources. The number of dedicated analysts expanded from 5 to 10, computer systems were upgraded from 4-processor CRAY XMP machines to 128-processor high-performance computing platforms [3], and appropriate ABAQUS licenses were procured. The successful formula of today is an environment that models and simulates processes in conjunction with the captured knowledge of experienced associates. In many process areas, such as welding, metallurgists with 30 years of experience will work together with associates with only 2 years of production know-how. The acceptance and aggressive nature of the new workforce toward computer applications has made the utilization of simulation tools, or the science-based approach, a success. Prototyping costs have become excessive over the last 10 years. The time it takes for these evaluations has also escalated. The implementation of the SBM approach has been shown to save thousands of dollars in front-end evaluations. The prototyping stage still exists, mostly virtual, and in a capacity to produce hardware only on the most successful options or ideas. For example, to do a prototype run of a welding operation requires machined parts to be designed and fabricated, welding to be performed on a nonproduction schedule, parts to be sectioned, parts to be X-rayed for porosity/cracks, and parts to be mechanically loaded for strength. A lot of 20 parts for each geometry condition is not uncommon. Using the SBM approach, after initial validation and verification have occurred, sequentially coupled thermal-structural simulations are performed, varying both welding parameters and geometry. From this, only valid, acceptable production conditions are brought forward for prototyping. Sometimes if known modeling risks are evident, no prototyping is required. Safety reviews with large safety factors are of this type—known quality indexes in the simulation models and solutions. The time using simulations has also been shown to decrease the prototyping cycle from weeks to days. The success in the implementation of science-based solutions was not easy. One of the early stages was to evaluate all of the manufacturing processes. This is described as follows. Step 1: rate the processes according to the importance in the core competency of the business, the “keepers” that set the Kansas City Plant apart from other vendors. Step 2: determine if simulation tools such as


ABAQUS are or can be applicable to solve the specific process problems. Step 3: list which processes get the most attention, which ones are “low hanging fruit,” and which ones are too complex and require craftsmanship to complete. (This last step is considered to be one of the hardest steps because there can be a mismatch between what is best for business and what the FEM analyst would like to understand, explore, and solve.) Today at the Kansas City Plant the simulation solutions to manufacturing problems are categorized into two areas. The first, getting most of the attention, are the computer simulations focused on integrating the three-dimensional Computer Aided Design (CAD) information into software programs that articulate motion, or scenes. Such simulations are performed with an engineer’s eyes and with attention to interferences, states, or locations of parts, and with visual scenes, and camera settings. It should be noted that these types of simulations sometimes use the same baseline CAD information as the second category. This second area deals with FEM-type solutions, focused on applying the proper physics to the process via verified software tools. Below are descriptive examples of the two areas of simulation work: Design-Type Computer Simulations

• Tool/Part Interaction Interference, kinematics events

• Work cell placement Machine movement, collision detection

• Assembly Modeling Tool/hand placement, work instructions

Physics-Type Computer Simulations

• Structural Fit and function, forming, assembly

• Vibration and Forced Response All potted assembly, cleaning operation, equipment calibration

• Thermal Stress Environmental operations, solder stresses, welding/brazing operations

• Molding operation Injection molding/cooling operations

• Heat Transfer Soldering/brazing/welding, oven cycles

• Custom Processing Heat treatment, plating, drying

The implementation of ABAQUS as a general purpose FEM tool has gained wide acceptance for many manufacturing applications. Sometimes it appears to be unlimited in its ability to solve the physics-based problems encountered. The next section shall exploit some specific examples of the utilization of the FEM for manufacturing problems.


3. Examples at the Kansas City Plant

The diversity of the product line and manufacturing operations makes the Kansas City Plant rich in opportunities for applying simulations during process development. Because of the small lot production needs and the need for high quality parts, every effort is made to develop the processes optimally the first time. ABAQUS/Standard and ABAQUS/Explicit have been used since the mid to late 1980s, with a transition toward process development before production over the last 5 years.

In the later 1980s much of the focus was on using the software tools properly, integrating the tools into a seamless environment, and building confidence in their applicability. Interest focused on comparative values—the new process was a certain percentage better, but there was no was validated confidence in the actual values. As time progressed, the development of new and updated procedures within the software tools, the update in mesh sizes and qualities, and the experience in analysts gave more credence to the actual resultant values. This direction is required today when using the tools on process evaluations where the final output will be the suggested solution to the manufacturing process. A lot of confidence is required. ABAQUS is one of the cornerstones of this effort.

Following are some site examples in specific process areas. Sometimes it is unclear whether the simulation is for design intent or manufacturing requirements. Remember that the design for manufacturability suggests integration of ideas between the two areas and a design is only robust if can effectively be produced.

3.1 Welding operations

Simulations to support weld operations fall into three categories, the first being the concern about heat and the concern about getting surrounding parts too hot. This is shown in Figure 1, where the temperature profiles (contours) are noted for a CO2 laser weld around a can assembly. The model is a pie section model. This simulation process is valuable for the review of temperatures far removed from the weld profile. Within the weld pool where vapor recoiling is occurring, temperatures are estimated based on varying thermal capacitances. The latent effects of vaporization are estimated by convection losses to the environment. Another example is shown in Figure 2, where the contours indicate areas where the laser radiation is impinging on the surface. Based on the cavity radiation and absorptivities of the materials at the laser’s wavelength, the process runs the risk of thermal cracking glass-to-metal seals. The same CAD model definition can be used for a variety of process areas. This leads to the second category, the coupling of the thermal model to the structural model. This is best described in upset forge welding.


Figure 1. Typical thermal profiles during welding process.

Figure 2. Cavity radiation of low frequency energy onto a glass seal.


In the forge welding simulation process, ABAQUS/Standard and ABAQUS/Explicit are sequentially coupled through time. An axisymmetric mesh is used in a coupled electrical-thermal procedure to produce temperatures that are read into the structural application for material softening and deformations under axial loading. The ABAQUS/Explicit mesh uses quadrilateral elements, while ABAQUS/Standard uses triangles generated from the same node map. (This technique was used to prevent ABAQUS/Standard from terminating with errors on bad elements during time recovery.) Figure 3 indicates the temperature profile over a softened deformed shape. User subroutines are used to evaluate the “time at temperature” for diffusion to occur and create a valid weld.

Figure 3. Welding of tubes into blocks: thermal profiles during upset forging process.

Another coupled physics example is the inertial welding process. Just as in the forging process, heat is introduced into the model, material is softened, and the model is formed together. But unlike the coupled electrical-thermal-structural response of the forging process, the inertia process


uses the friction user subroutine to develop frictional dissipations for heating. As ABAQUS/Standard solves the increment, the energy loss due to friction causes the angular velocity to slow down, thus reducing the heating on the next increment. Figures 4 and 5 show typical outputs for the inertia welding process where two axisymmetric parts have been formed. The later is the final deformed shape.

Figure 4. Typical run sequence for forging process.

Figure 5. Final shape section profile for inertia welding of two cylinders.


In both the prior examples, caution has been given to the evaluation of friction and coupling energy across individual parts. Actual sections of baseline parts correlated to computer runs have allowed for the determination of these parameters.

The last category is one in which the strength of the weld is in question. This is based on the depth of penetration, the materials in question, and the overall load path to failure. Figure illustrates where a weld section has indicated porosity on the weld pool area. Computer models have been constructed parameterizing the voids. Figure 6 illustrates a 4o slice of a cylindrical part with a lower cylindrical void subjected to axial loads. The simulations use power law hardening models for the various materials and shear failure damage to direct the shear path and suggest failure loads. Heat-effected zones are estimated based on metallurgical section input.

Figure 6. Equivalent plastic strain (damage) to weld joints with porosities.

3.2 Mechanical actuator operations

This area of simulation has been the most challenging, and the approach varies between each application. Most of this work done in the industry is focused on applying rigid body dynamics to


the structures. The Kansas City Plant is mostly interested in the effects of tolerance and manufacturing features on the operation of assemblies. This suggests that elastic collisions and dynamic forces are important. Most of the assemblies today are formulated from both rigid bodies and deformable bodies. Figure 7 shows a typical mesh of a working mechanism that operates within 1–5 milliseconds.

Figure 7. Sample mechanism mesh using rigid/deformable bodies.

This method of solution has gained popularity for a number of reasons. The advantages to this method follow:

• Derived elastic collision (restitution) parameters. • FEM models built directly off CAD systems. • Parts can be switched between elastic and rigid easily. • Visual presence is very marketable. • Mass/Center of Gravities extracted directly from parts. • Allows for elastic deformations to be captured.

The main disadvantage of this method is the computational time. It is not uncommon to get time constants on the order of a nanosecond to capture wave speeds through such a small mesh on a steel part. A sample output might be the deformation and stresses in the center pin or the tolerance shift in the bearings, as shown in Figure 8. Note that this response can only be found in either this


comprehensive method or breakout models in which forces are transferred to individual sub models.

Figure 8. Typical output results from a complete assembly model.

The use of entirely deformable bodies is also employed and, in some cases, processed with ABAQUS/Standard. Figure 9 shows an output response of an actuator subjected to an induced strain simulating a piezo effect. This type of modeling is unambiguous—it has no assumptions to the dynamics, collisions, and force outputs.


Figure 9. Actuator launching mass—output stress contour plot.

The use of hybrid models will continue to aid in the solution times. The compromise of using rigid bodies is small. There has been more attention focused on using all rigid body dynamic simulations to extract baseline information than on specific areas for further evaluations. The first problem is to determine how springs are going to be evaluated. In some cases the spring contacts the surrounding parts, yielding a stiffness constant that is not ideal. Development work is being done on the determination of proper surface behaviors for the penalty contact formulation.

3.3 Forming and swaging operations

Classical metal forming with elastic rebound is just one area where the simulation tools are applied. The development of FEM models toward these applications is well understood and used commonly today. The current developments have been toward the damage models, or understanding the effects of material breach. One example is gas analysis and review. To open a steel container that has been in service requires a needle valve to be placed on the outside surface. The valve breaches the surface and sniffs the contents. A computer simulation was performed on various needle valve arrangements to review the piercing operation. Figure 10 shows the results of one such case.


Figure 10. Needle valve piercing steel case—output contour plot.

Another application using material damage and failure was the evaluation of notch features in burst discs. The current method of making scores (or stress risers for failure patterning) is to etch the patterns chemically. The new proposed process called for v-grooves to be stamped in the section and then annealed. A section schematic of this geometry is shown in Figure 11. The depth of the groove and its shape were optimized for the new manufacturing process. Figure 12 shows a typical output prior to failure for the original etched geometry.

Figure 11. Section detail showing model and loading response.


Figure 12. Initiation of pressure loads on ½ section model.

The Kansas City Plant has a vast amount of resources to apply to each production problem. In the past large models (on the order of 0.1 million–1 million nodes) being processed, under transient or transient dynamics procedures, with highly nonlinear constitutive models for the materials were considered “hero” runs. The runs tended to allow for an understanding not possible otherwise. Today these type runs are typical to a point where parametric studies are processed, running sometimes 50 runs on a single case.

3.4 General fixturing

As expected, there are many custom mounting fixtures and holding devices that are created for the various manufacturing processes. In the area of general fixturing, there are evaluations for the functional performance based on service loads and the evaluation of a fixture for safety concerns. The first example is one in which evaluations were performed on a chuck jaw fixture to determine if the bolt torque preloading was adequate to maintain contact and restrain service loads. Figure 13 shows the computer model used in this example.


Figure 13. Typical jaw chuck fixture loading.

The computer model was constructed directly from the CAD assembly in the I-DEAS Software called Master Series [4]. The mesh was composed of parabolic tetrahedron elements. The preloading of the structure was via a thermal strain induced only to the bolts. This strain caused part preloading. Both chucking pressure loads and centrifugal loads were applied to the model. With ABAQUS/Explicit, the centrifugal loads were applied via body forces proportional to the element centroid location (part radius). Results for the loading were reviewed. Figure 14 shows the output due to the pressure service loading on the master jaw part for both single and dual bolting arrangements. Contact forces were monitored to determine if service loads were exceeding bolt preloads. A series of runs were processed with varying preloads, chuck bolt quantities, and friction factors to give confidence in the solution.

Figure 14. Stress output contours on chuck jaw system.


Other examples include the evaluation of the fixtures for service loading. Figures 15 and 16 display stresses in a compound joint as a result of large dead load mountings. Redesigns were processed on the joints to optimize the manufacturability while retaining enough strength and safety margins for adequate designs.

Figure 15. Stress in fixture joint as a result of heavy loading.

Figure 16. Stress in support stand under offset loading.


One of the largest areas of evaluations and simulation work within the Kansas City Plant comes in the review of fixtures during shock and vibration testing. Models are constructed, usually tetrahedron meshed FEM models from the based CAD assembly, tied together using the ∗TIE, ∗EQUATION, or other Lagrangian methods and solved. Eigenvalue extractions, as shown in Figure 17, are evaluated. Masses are added, the stiffness of members changed, and restraining locations are modified to either offset the natural mode or move it beyond the frequency range.

Figure 17. Typical resonance response output showing modes.

4. Simulations Using Advisors

The use of computer simulations in a diverse manufacturing environment is an ever-changing process. For example, as one looks back on FEM solutions processed two years ago, one would make changes if the same analysis were run today. This may include the usage of a viscous constitutive material model that now includes the rate effect, or it may include a finer detailed mesh to account for the quality aspect. The evaluation of properties with American Society for Testing and Material (ASTM) specifications may not have been available for the validation of a material subroutine. One thing to note is that the simulations, even projects being worked on


today, offer value and understanding to the manufacturing process: to be able to evaluate a forming process and gain reaction loads, to be able to judge the time required to heat a complex system under varying environments, to be able to see the stress state of a billet being hot formed. Although not perfect or ideal, these simulations aid in the development of the manufacturing process for specific parts. The number of projects processed each year at the Kansas City Plant is not growing at the rate needed to take full advantage of the available technology. The balance between analyst and manufacturing process engineers suggests no escalated growth in personnel toward dedicated computer simulation work. Some of the best fits for the software tools are in the hands of the manufacturing personnel close to the process or process area. For example, a welding engineer knows what a weld bead profile will look like before it is processed based on experience—sometimes interpolating, sometimes extrapolating on past manufacturing events. On the other hand, a dedicated FEM analyst can only extrapolate as to the correctness of the simulation process—checking loads, elements, and underlying physics. The analyst can simulate many processes that are impossible to fabricate. One’s first thought is to cross train. Educate manufacturing and give them the software, train them, and make them take on the analyst role. Push the FEM analyst toward an internship in a manufacturing process. Both of these concepts have been tried and mostly failed. It is not a technological issue. Computer systems can be placed and are used in all areas of the factory. However, social, education, and short-term needs and requirements make success difficult. For example, it is nearly impossible to have an electrical engineer process a soldering simulation using viscoelastic materials and recover effective plastic strains and fatigue indicators. It is difficult to have mechanical engineers grasp the effect of high-voltage gradients, joule heating, and the propensity of electrons cascading from a charged cathode. The growth for simulation work and the exploitation of simulations in the factory have driven the Kansas City Plant into the development of a framework called “Simulation Advisors.” This area of development targets specific applications in which prior simulation work has been known to be successful and has been validated and the need exists to continue to process runs. An initial development project was used to determine the interconnects between the applications on the front end and the high-performance computers in the background. This pilot project suggested many solution options. The one selected was based on standards and pseudo standards. The standard advisor is built around a Hyper Text Markup Language (HTML) web page. The advantage to this technology is the proliferation of the web browsers in the plant. This concept would alleviate the requirement of learning a new interface, such as ABAQUS/CAE. The page is constructed with the analyst-determined valid parameters, such as machine forces or boundary conditions, and the extent to which these parameters can be made to give valuable results. This data collection may only manipulate an ABAQUS input stream or update a script that runs ABAQUS/CAE in the background. The web page is the means to create an input stream for processing. Figure 18 displays a screen shot from a system that evaluates the insertion of a pin into a plate of material. This example illustrates an advisor that will take user inputs; update geometry within ABAQUS/CAE; create a proper input stream; spawn the job to a broker on the HPC system; process results in the form of strains, force plots and animations; and return the output information back for review.


Figure 18. Sample Simulation Advisor input/output pages.

All scripting is done using Python [5]. Although the initial pilot used many different system languages and was more efficient by the selection, it was too expensive to maintain. One of the goals of the project was to retain the development for the advisors within the mechanical engineering (analyst) community and develop all of the advisors in a common, consistent framework. As a note, this work has grown into a new development called “Technical Advisors.” Although these cases do not run FEM solutions in the background, they do process routine calculations such as coil spring designs, bolt loading of joints, and kinematic motion of object being dropped. More information about the Simulation and Technical Advisors, including the applicability and coding required to initiate the baseline scripting, can be found in reference [6].

5. Conclusions

The use of ABAQUS for the solution of manufacturing problems at the Kansas City Plant has been very successful over the past two decades. The diversity of production processes requires a set of flexible, general-purpose tools to be applicable. Early work was focused on encapsulation—thermal stresses in electronics packaging and forming—metal shaping with elastic rebound. As discussed, today’s demands are more on precision, more on achieving quality within the processes that will yield small lot parts. Requirements today are for more simulations on each process and the support of more processes.


The direction of a two-tied approach to manufacturing simulations is being deployed and used. The focus of the simulation group is dedicated to evaluating, understanding, and accessing manufacturing processes. The advisor technology has allowed for the exploitation of the simulation work to be performed in the background by a process engineer. This has greatly enhanced the effectiveness of the simulation business, augmented the prototyping process, and enhanced the readiness to manufacturing decisions.

Next stages include the incorporation of all process variables into advisors. This methodology would require no “wrappering” of the boundary conditions around the meshes, just the mesh introduction to the advisor environment. With the ever-increasing quality and automation of meshes, the view in the next 10 years is one in which application advisors create a true virtual manufacturing environment.

6. References

1. Hibbitt, D., B. Karlsson, P. Sorensen, ABAQUS Version 4.5–6.3-1, 1984–2004. 2. Beck, D., Deputy, DP-20, Department of Energy, Washington, D.C., 1999. 3. Silicon Graphics Incorporated, Origin 2000 series parallel platforms, 1998. 4. EDS, Inc. (formally UGS, formally SDRC, Cincinnati, OH), I-DEAS Master Series, Version

10, 2003. 5. Stichting Mathematisch Centrum, Python programming language, Version 2.1, Amsterdam,

The Netherlands, 1991–1995. 6. Seaholm, A. J., “Web Based Simulation Advisor Using ABAQUS and Python,” 2002

Midwest ABAQUS User’s Group Meeting, West Lafayette, IN, October 2003.

using abaqus in a manufacturing environment at honeywell federal

Documents