From the drawing board to the keyboard . . . computer-based tools provide new insight into engine performance as they cut development time
Perhaps no other area of mechanical engineering rests on as many technologies as does engine design and development. Engine design requires a thorough knowledge of structural and dynamic systems analysis, fluid dynamics, acoustics, tribology, combustion technology, the chemistry of aftertreatment, and manufacturing technology. The intermittent nature of combustion, fluid flow, and structural response of the internal combustion engine is still not precisely understood.
The explosion of computer technology in recent years has decisively transformed the field of engine design and analysis, as it has many other areas of technology. Engine design analysis, largely intuitive in the past, is now based on increasingly sophisticated predictions of component and engine performance.
Among the advantages computer-based tools provide over more traditional methods of engine design are fast, accurate compilation and analysis of information; the opportunity for simultaneous development and testing of components at multiple locations, using a variety of different software packages; and the capability for rapid fabrication and prototyping of tools and parts.
These advances are due not only to the availability of new tools but also to the contemporary economic climate, in which manufacturers must market their products in increasingly shorter time frames to stay abreast of, or surpass, the competition. In addition, the engine industry is required to meet increasingly demanding international standards for improved emissions levels, fuel economy, and engine reliability, combined with lower production and maintenance costs.
Computer-based procedural changes and tools have approximately halved the time -- from eight years to four -- needed to bring a new engine design from concept to production, and, through more accurate predictions of engine and component performance, have reduced the number of corrections needed in the developmental and manufacturing stages of production.
In the past, for example, the first stage of developing engine or component blueprints was the complete responsibility of the design engineer; the design was later passed on to development and test engineers, who would build and test a prototype. A component that broke or failed to meet required performance standards at this stage might be forced to undergo several redesign and test sequences before being handed to manufacturing engineers, who also made design adjustments to ensure that an engine could be produced in practical fashion on the factory floor.
In contrast, computer-aided design (CAD) is generally carried out by what is called an 'integrated platform team,' which, from the beginning, includes engine designers and development and manufacturing engineers, as well as service and sales representatives. Far greater effort is placed on the initial design process to ensure that an engine or component meets all stated requirements in advance. The traditional development stage is now called design verification, and is carried out only to confirm whether a design meets expectations.
The drawing board has been replaced by three-dimensional CAD and, more recently, by parametric solid CAD. Designing solid models is exacting work, as every part of the component must be completely defined. However, this comprehensive component definition stage precludes subsequent misinterpretations in the manufacturing process. Solid modeling also permits rapid adjustments to designs and allows computational grids to be generated quickly. Computational grids are used with analysis tools such as finite element analysis and computational fluid dynamics (CFD) when addressing important design problems such as air and fuel flow, and predictions of wear and lifetime for individual parts and systems.
The shift at Southwest Research Institute (SwRI) toward an integrated analytical approach to engine design began in 1990. The goal was to develop a set of engine cycle simulation codes that were thermodynamically complete, easy to use, and that permitted engine performance evaluation at an early stage in the preliminary design cycle, thus encouraging the rapid refinement of overall engine design parameters. Now known as Virtual Indicated Performance of Reciprocating Engines (VIPRETM), the heart of the program is a solution of the gas dynamics in the intake and exhaust pipe networks, coupled to a solution of the first law of thermodynamics for cylinders and manifolds. VIPRETM provides a complete dynamic model of the air cycle, predicts gas temperatures and pressures throughout the engine, and has been successfully used to simulate and optimize performance in engines ranging from single-cylinder weed trimmer motors to 16-cylinder locomotive powerplants. A recent update of the program, VIPRETM 2.0, offers comprehensive simulations of turbochargers, heat exchangers, mufflers, filters, and other flow network elements. The generation of emissions species in engine cylinders is predicted, and the species are tracked throughout the intake and exhaust systems.
This article highlights some of the computational tools that Institute scientists and engineers are using to implement a new approach to engine design. Included are discussions of finite element analysis, tribology, computational fluid dynamics, rapid prototyping, and probabilistic reliability analysis.
Finite Element Analysis
Finite element (FE) analysis has been used for more than 30 years in the evaluation of engineering structures to determine stress distribution and heat flux. The FE method divides a structure into a mesh of discrete subdomains, within which a polynomial function serves to represent the displacement or temperature field. In the simplest form, linear functions are used, although quadratic or higher polynomials can also be used. Four possible element shapes are employed to build a finite element mesh in three dimensions - brick, tetrahedron, wedge, and pyramid. Brick or hexahedral elements are preferred because they require less memory and are generally more accurate than tetrahedral elements for a given structure. Tetrahedral elements are less accurate because they are unnaturally stiff. This problem can be overcome by using quadratic rather than linear displacement functions, but these also consume disk space and solution time.
Complex meshes containing hexahedral elements are, however, time consuming to construct. Until automatic mesh generation software became available, it was necessary for an engineer to manually create a mesh, a process that could consume several weeks in the case of a complete cylinder block. Current FE software packages that automatically and rapidly generate meshes are an improvement but still not ideal. For example, they make it easier to generate tetrahedral, wedge, and pyramid elements rather than hexahedral elements in complex structures. However, the time needed to create a mesh has been reduced from weeks to hours or even minutes, a clear benefit. Not only are analyses completed faster, but in greater detail and with more accuracy. In addition, the designer has access to improved tools for understanding and communicating analysis results.
Tribology -- the science of friction, lubrication, and wear of interacting surfaces in relative motion -- is a complex discipline that has been integrated into sophisticated software tools for design analysis, particularly in the area of predicting engine bearing wear.
It is often impractical if not impossible to determine the component wear rates of engines under development. Nevertheless, predicting the lifetime and maintenance needs of certain categories of engines, such as large industrial or marine engines designed to run continuously for periods of five years or more before any appreciable wear occurs, is of crucial importance.
Two forms of wear that commonly occur in this class of engine and that need to be forecast with accuracy are sliding and contact fatigue wear. While combustion engines have become lighter, they are still required to retain equal or greater horsepower, with the result that engine bearings are subjected to increased loads, which can result in significant deformation. Tribological models developed by Institute engineers are being used successfully to predict critical parameters in bearing design.
For example, as shown in the illustration above, the SwRI bearing model can rapidly predict oil film thickness and dynamic film pressure. From this information, engineers can determine whether a bearing is overloaded or if it exhibits evidence of cavitation that could cause failure. The film pressure is compared with the known material limit of the bearing. If the film pressure exceeds the material limit, the designer knows that the bearing has a high risk of contact fatigue and pitting.
In one case history, a model was developed to predict contact fatigue wear in bearings typically found in large marine engines. First, a representative profile of the surface roughness was used to calculate surface and subsurface pressures on the bearings. A crack propagation model was then employed to predict crack formation and the pits that would subsequently form on the surfaces of the highly loaded bearings. If several such pits form, the load-carrying capacity of the surface is substantially reduced and can lead to rapid failure.
The Institute-developed dynamic bearing model can be used to assist clients in predicting problems that arise from bearing deformation, oil film leakage, oil film cavitation zones, and crankshaft wear.
Computational Fluid Dynamics
Institute engineers have gained extensive experience in the use of CFD as an aid in design analysis, particularly in understanding and implementing combustion process refinements. To date, the SwRI Engine Design Department has developed a state-of-the-art three-dimensional CFD modeling capability for all engine components and processes, including intake ports and manifolds, in-cylinder motion, mixing and combustion, as well as coolant, fuel, and lubricant flows.
The numerical methods necessary to build accurate, useful CFD models are computer-intensive and require considerable expertise. Small engine companies are often unable to afford this kind of analysis. The fluid flow problems of greatest interest in engine design are complex -- the nature and interaction of turbulence, multi-species mixing, fuel spray and combustion, piston motion, the timing of valve opening and closing, and many other such variables. Several internal combustion processes can only be described in multidimensional and transient terms, making the processes difficult to understand and to describe mathematically.
Precise measurement of combustion events is also elusive. Much of the experimental work conducted in the past could only provide broad performance data, which could not be evaluated in sufficient detail to make the modifications needed to enhance a particular design.
In contrast, CFD can provide enough detail over a full spectrum of engine events to help perfect a design. The accuracy of the model is of course critical. A good model provides baseline analytical information that can be used for validation and comparison purposes. A set of parametric designs can be built up to compare important design characteristics. Parametric studies allow a trend to be detected and, after several experiments, can suggest an optimum design.
Although some models are widely used in industry, there is no consensus as to which is best. Much development work remains to be done. For example, the current model of turbulence within the cylinder may not be adequate, but at present there is no practical alternative when one considers design time and budget constraints.
Another feature of modern engine design is direct access to software and machines for rapid prototyping of components and parts with minimum human interpretation or interaction. There are two main methods of carrying out rapid prototyping -- numerically controlled machining based directly on a CAD design and stereolithography using laser equipment. However, a number of other methods have emerged, including solid ground curing, laminated object manufacturing, and fused deposition modeling.
Rapid prototyping by numerically controlled machining is made possible by the generation of machine tool paths from a solid or wire frame CAD model. There are disadvantages to this form of prototyping. A component can be made from any machinable material using this method, but if it is necessary to remove a lot of material, the process can take several days. In addition, because the component is made by cutting with a mechanical tool, shapes cannot be as complex as those created using laser-based methods. Finally, tooling costs can be high, as operator time to oversee the machines must be taken into account. Laser-based methods can be carried out largely unattended.
Stereolithography is the most widely used laser-based rapid prototyping method. In this process, a solid model of the component is reconfigured by computer into horizontal slices. The information is relayed to a machine that traces the outline of each slice with a laser beam focused on the surface of a bath of photosensitive polymer. The laser light solidifies the polymer, and the solid piece is lowered incrementally, allowing a fresh layer of liquid polymer to flow across the surface.
Rapid prototyping can reduce the time needed to produce parts by half and the costs by up to one third. In addition, rapid prototyping techniques incorporate features that can result in significant improvements to the design process:
Probabilistic Reliability Analysis
Engine components are intended to last, and the longer they last, the more costly they are. For an engine company, containment of costs means profitability; therefore, an engine designer must balance durable designs against the real need for durability. Given a target component lifetime, probabilistic reliability analysis provides the means to design a component with the confidence that it will meet that lifetime, without including costly materials, processes, or features that are not necessary to meet the lifetime.
To use probabilistic reliability analysis, a design engineer must quantify the uncertainty of various parameters in terms of distribution functions such as Gaussian, linear, exponential, etc. The parameters can be in the form of geometric tolerances assigned to dimensions or finishes, material properties, and the like. For instance, crankshaft bearing diametral clearance is normally specified to be within a certain range.
Next, the engineer must determine failure mode sensitivity to variations in the parameters, which can be done through finite-element stress analysis or other techniques. Rather than designing for minimum and maximum conditions of the tolerance bands, the probability distribution function reveals how likely a parameter is to deviate a certain amount from its nominal condition, and the cumulative analysis allows the engineer to quantify the probability of component failure after a given time in service.
Without this tool, engineers had to design for the most demanding expected conditions and allow a factor of safety to ensure a component would survive those conditions, a process often leading to components that were expensive and difficult to manufacture. Conversely, without a quantitative estimate of the likelihood of failure, it was possible to miss important parameters that might lead to premature failure. Engineers in the Materials and Structures Division at SwRI have developed probabilistic reliability analysis into a general purpose software tool that is now being applied to improve the designs of engine components.
Computer-aided design and analysis represents a tremendous improvement over traditional methods of engine development and introduces new challenges as well. The engineer who formerly was pressed to find ways to estimate the behavior of engine components in service is now faced with the potential to generate more information than he can manage. There is no reduction in the need for well-trained engineers with the ability to make sound decisions and to skillfully and judiciously apply the new tools. Organizations that effectively handle this transformation -- from minimal information to information overload -- will be rewarded by reduced development cycles, improved product quality, and improved productivity. Southwest Research Institute is committed to remaining at the forefront of engine design in terms of both a highly trained, responsive staff and the latest in technological advances.
Published in the Spring 1996 issue of Technology Today®, published by Southwest Research Institute. For more information, contact Joe Fohn.