When Dresser-Rand's new valve-in-piston (VIP) gas compressor hit the market, it was an immediate success. The VIP design was a radical departure from the traditional compressor design. Rather than having suction and discharge valves built into the cylinder, the VIP compressor, as its name implies, incorporated valves directly into the piston. This simpler design with fewer moving parts created a gas compressor that was less expensive to operate and easier to maintain.
Dresser-Rand initially introduced a single model of the VIP compressor, but since it was so popular, the company wanted to release ten additional models (covering a range of pressure ratings) as quickly as possible. Normally this process was anything but quick because Dresser-Rand tested design changes before releasing new products to market. Building and testing the original VIP prototype had taken six months. Testing ten additional prototypes, even if much of it was done concurrently, would have meant missing the window of opportunity with this product.
To reduce the time needed to get new compressor models to market, Dresser-Rand engineers opted to evaluate design changes in its FEA system, ANSYS, from ANSYS Inc. (Canonsburg, PA), and perform physical tests on only two of the ten models. Ajay Garg, a senior mechanical engineer in Dresser-Rand Company's Engine Process Compressor Division (Painted Post, NY), estimates that this approach made it possible to evaluate the ten new models in six months compared to the two to three years testing would have required.
ANSYS has been part of the division's new product development process for several years. The company uses the software for structural, thermal, and vibration analysis. Analysis results are applied early in the design cycle to evaluate new compressor and engine designs, then again later when a design is changed to study the effects of the change.
Garg summarizes the use of FEA this way: "We use it at first to make sure a machine works for the application it is being designed for. Analysis results then become the basis for future development of that machine. Marketing might ask for an increase in operating conditions, such as an increase to 1500 psi from 100 psi. Or manufacturing might ask us to replace a 1/8-inch fillet with a 1/16-inch fillet. Since test results for the original design are correlated with FEA, we don't have to run tests for every subsequent design change. This shortens our development cycle and reduces costs as well."
Early analyses are performed long before a design has been defined as a CAD model. "These analyses are done to ensure that a concept is valid for a particular market or that it can withstand the specified operating conditions," says Garg. When analyses are performed at this stage of the design cycle, analytical models are defined in ANSYS using the software's parametric modeling capability. This feature makes it easy to alter the model for what-if studies or to evaluate design changes later on.
For every new product, Dresser-Rand also performs a full series of physical tests. The costs and time required for physical testing limit its use to an evaluation of the final design. Another drawback of testing is that it can be difficult or impossible to check stress levels in certain areas of a compressor where it is not possible to place strain gauges. Dresser-Rand's solution has been to instrument test experimentally. "This lets the engineers make sure they don't miss any hot spots or potential problems," Garg says.
One component the engineers needed to evaluate was the circular steel plate that acted as the piston, specifically its response to stress at all the different sizes. The piston was a uniformly loaded plate containing drilled holes and milled grooves. It deformed as a cantilevered plate. The engineers expected to see maximum stresses at the 0.03 inch fillet of the innermost milled groove.
The plate operated under cyclic loads which were determined by laboratory testing. Garg and his colleagues first performed a structural analysis of a 3D finite-element model of the plate that included the holes and grooves. This 46,176 degrees of freedom model took about three weeks to create.
Results of the analysis showed that bending the plate as a cantilever created compressive stresses at the holes and tensile stresses at the milled grooves. These stresses at the milled grooves were chosen as the governing factor for the design but since the finite-element model did not contain the radii of the grooves, there was a singularity at the point of interest. Engineers modified the design (increased plate thickness) to minimize tensile stresses by minimizing compressive stresses at the surface of the drilled holes. This was confirmed by another analysis and the percentage reduction in compressive stresses at the holes was assumed to present a similar stress reduction at milled grooves.
In the interest of analyzing the nine other plate sizes as quickly as possible, the engineers decided to investigate the possibility of replacing 3D structural analysis with 2D analysis. "In a static as well as a fatigue environment, tensile stresses are more harmful than compressive stresses," explains Garg. "Therefore, the drilled holes could be ignored and the geometry could be represented by a 2D axis-symmetric finite element model."
Parametric modeling of a 2D plate went very quickly. The model, which included the milled grooves, took about one hour to prepare for analysis. Less than two minutes of CPU time was needed to run the analysis. Results of 2D analyses for several plate sizes were correlated with testing. Some plates were redesigned to improve reliability and the 2D analyses were repeated. All additional plate sizes were analyzed in 2D.
Garg cautions that corrections and compensations in 2D were
critical to simulating 3D behavior and validating analysis. Also, the 2D FEA
method was verified through analytical and experimental methods. But the effort
paid off. As Garg says, "Once the 2D analysis approach was established, design
and optimization cycles for the VIP compressor were reduced by an order of
magnitude in cost and time."