Case Study: Automated Optimisation of Connecting Rods
IDAC was required by their client to develop automated macros to carry out design optimisation of connecting rods using the ANSYS program. Due to market competitiveness it was important for the client to be able to produce an optimised preliminary design (with a minimum weight in as short a time as possible). The objective of the work was to develop a customised procedure for the client to carry out rapid optimisation of preliminary connecting rod designs. The optimisation procedure was performed in three phases as follows:
- Phase 1 – optimise the small end and shank of the rod
- Phase 2 – optimise the big end, cap and bolt design
- Phase 3 – develop a customised procedure for the client to carry out an accelerated optimisation of a preliminary connecting rod design; a special feature of the accelerated optimisation process was that model parameters from the past were used to derive a new design
Phase 1 and 2 simulation details
For each phase, two main macros were created. The first was a model file, which created a parametric model of the design. The second was the design optimisation control file, which specified the variables to be optimised (design variables), optimisation constraints / limits (state variables) and dependent variable to be minimised (objective function). An advanced zero-order method called the subspace approximation technique was employed for the optimisation process.
The basic geometric model for the entire connecting rod when the shank and small end from Phase 1 is joined to the big end and cap from Phase 2. The geometric model from each of the two phases was constructed within the ANSYS program. The finite element model was then created with using higher order structural elements. The piston pin and the crank pin were assumed to be rigid. Rigid-to-flexible contact surfaces were created between the ends and the pins using the surface-to-surface contact elements. Three load cases were carried out to analyse the pin interference fit, tensile inertia loading and compressive inertia loading.
The design optimisation control file was used to carry out the design optimisation process by varying the design variables, executing the model file, and evaluating the state variables and objective function. Typical design variables for Phase 1 were the width, depth and taper ratios of the shank, and the wall thickness of the small end. Typical design variables for Phase 2 were the radii of some of the big end features, the depth of the bolt hole, the radius and height of the cap web.
For both phases, the state variables were the maximum and minimum stresses, and the objective function was to minimise the total volume of the design. A number of random designs would be generated by varying the values of the design variables within the specified limits before the optimiser ‘homed’ in to the best design.
The macros were used to carry out single design stress analyses and also full design optimisation analyses for a number of connecting rod designs.
Phase 3 – accelerated optimisation details
The work for the accelerated optimisation project was divided into two parts. The first part was devising a knowledge database containing all optimisation variables and their values for past optimised connecting-rod designs. The knowledge database was created such that it could be updated with minimum effort each time a new connecting rod is designed and optimised. The speed and accuracy of deriving a new preliminary design would depend on the number of data sets contained in the knowledge database and the proximity of the existing data points to the required parameter values.
The second part of the work was the development of an accelerated optimisation control file, in which the design variables, state variables and objective function were defined. They were set to be the same as those in the first two phases. The accelerated optimisation control file also read in the previous design data sets from the knowledge database.
A number of special-state variables were set up as the criteria for selecting the best results sets from the database. These were the engine design parameters, namely the required big end and small end radiuses, big end centre to small end centre distance, maximum tensile force, maximum compressive force and its offset loading angle. These past design instances would then provide the closest approximations or starting values for the optimiser to perform the actual optimisation analysis. The subspace approximation technique was used for the optimisation process.
IDAC has developed the macros for generic use, which provide a number of very useful options for an accelerated optimisation project. For the quickest solution, the user can request the ANSYS optimiser to select and output the closest existing design set from the knowledge database. This is a viable approach if a rapid draft design is sought and a similar design exists in the past.
Although Phase 3 of the project (accelerated optimisation) did not include a full finite element analysis, it is an option that the macros could be further developed to include a finite element analysis, so that the stresses and total volume could be fully evaluated. This would still give significant cost savings compared to the standard, non-accelerated optimisation analyses, since the closest data sets from the knowledge database are used as first approximations to the current design.