Simulation lifts design exercise to a higher level
04 January 2013
A Norwegian university research project is using computer models to predict the performance of complex offshore materials handling equipment. In the short term, the work is helping designers pick the best components for the job; ultimately, the aim is to automate more of the design process.
With offshore oil and gas drilling operations costing millions of dollars a day, crews are under pressure to get their jobs done as quickly as possible. One of those jobs involves assembling the thousands of metres of flexible pipe that make up a drill string, and doing so safely and consistently on a remote platform where space is at a premium and weather conditions are frequently challenging.
Typically, the materials handling equipment for this is hydraulically operated and, to some degree, automated. The control system design is challenging; when a crane is in motion its dynamic behaviour is dependent upon many variables - the electrical and hydraulic characteristics of control valves and hydraulic actuators, the inertia of the crane’s structure and its load, and the complex interactions between all these components.
Moreover, the chosen hydraulic components must deliver the required level of responsiveness, or ‘bandwidth’, while meeting requirements for budget, size, weight, long term reliability and ease of maintenance.
A research project at Norway’s Agder University, currently being carried out in conjunction with offshore equipment manufacturer Aker Solutions, is hoping to simplify this formidable design challenge by allowing engineers to build and run detailed simulations of equipment before they assemble a single part. Morten Kollerup Bak, a PhD research fellow in charge of the project, takes up the story:
“Our aim is to use model-based design to predict the behaviour of the finished products and to support key design decisions. For that to work, everything depends on your being able to model the entire structure and control system in sufficient detail to get a realistic idea of its performance.” For this work, Bak is using the MapleSim simulation application from Maplesoft, which is supported in the UK by Adept Scientific.
“I divide the whole system into three different models,” he says. “The mechanical structure, the hydraulic actuation system and the electrical control system. I’m using MapleSim to model the first two parts and in some cases all three.
“Our aim with this work is build the models of hydraulics as much as possible from standard catalogue data. But we quickly found that component manufacturers don’t always provide all the data you need, particularly when you are looking at the precise behaviour of their components in dynamic conditions.”
To obtain the missing data, Bak has built custom models of key components, like control valves, and validated their accuracy by conducting tests on single components operating in isolation. Once he has confidence in the performance of the custom elements, he integrates them into the MapleSim system models and uses that to evaluate the likely performance of the crane.
“With my industrial partner, we have already built a model of one of their existing cranes and demonstrated that it predicts the behaviour of the real crane accurately. Now we have begun to use the model in our design work by looking at the likely impact of substitutions or design changes to individual components.”
The ability to model such changes before doing them is obviously extremely useful for the Aker Solutions designers, but the next stage of the project has the potential to fundamentally change their roles.
“Ultimately, we want to use our models for design automation. In this approach we feed the system with the performance requirements of the finished product and with a library of options for hydraulic and mechanical components, then allow it to search for the optimum solution.”
Optimising the simulated model requires an efficient search algorithm and for this Bak is using the so called ‘Complex’ method. “In the algorithm, we populate the simulation with a number of randomly generated designs and it evaluates the performance of each,” he explains. “It then picks the poorest performing design and ‘mirrors’ it across the centroid of the remaining designs to produce a solution that should work more effectively.” This process is then repeated, with the worst performing design substituted each time, until the solutions converge on the optimum result.
Initially, Bak is using stability and accuracy as performance criteria, and, consequently, the optimum solution is the design yielding the lowest level of oscillations in the hydraulic system and with the best ability to follow the position reference fed to the control system. Later, Bak plans to add other criteria such as price, robustness and long term reliability.
Contact Details and Archive...