Improving Performance of the Stirling Converter: Redesign of the Regenerator with Experiments, Computation and Modern Fabrication Techniques

Quarterly Report for the Period September – December 2000

Mounir Ibrahim – CSU

Terry Simon - UMN

David Gedeon - Gedeon Associates

Roy Tew - NASA Glenn RC

Cleveland State University

1983 East 24th Street

Fenn Tower 1010

Cleveland, Ohio

Date Published – January 2001

PREPARED FOR THE UNITED STATES

DEPARTMENT OF ENERGY

Under Financial Assistance Award

No. DE-FC36-00G010627

1

Background

This is the first quarterly report on fundamental research that intends to advance Solar Dish/Converter Technology by making significant improvements to a crucial component of the Stirling engine converter, the regenerator. There is great incentive for improving the regenerator heat transfer while reducing pressure drop. The project will be conducted over a three-year period and is a collaborative effort among experimentalists, researchers of Computational Fluid Dynamics (CFD) and system analysts.

Alliances with Stirling Engine Development Companies

We have been working with one Stirling-engine development company, STC, since the proposal writing stage for this project. More recently, we have approached another Stirling-engine development company, Sunpower. Both companies will help us identify the important issues they face concerning the regenerator and provide specific engine parameters required to design our experiments.

Our associations with these two companies goes a long way back. More than twenty years, in fact, during which time we have worked together as fellow collaborators with several key employees, as well as served as project managers or research advisors, on several government-funded Stirling-engine research projects.

STC

Stirling Technology Co. is a company of 32 employees in Washington State. Their President, Kelvin Colenbrander, has expressed willingness for STC engineers to provide detailed information on the regenerator and associated manifold of a representative engine.

In December, Professor Simon visited STC and viewed the engine components, assembly area, assembled engines and testing facilities. He spent the day with former student Songgang Qiu, who has recently been promoted to the position of Director of Engineering, and Maurice White, a Principal Engineer and co-founder of the company. During the visit, Dr. Qiu and Professor Simon identified the engine parameters and operating conditions that will be used to establish the dimensionless parameters of the test and, therefore, the operating parameters and conditions of the test. They also discussed at length the geometry of the engine in the vicinity of the regenerator and how that geometry ought to be simulated in the test. According to Maurice White, Principal Engineer, STC:

“If improved regenerator performance and/or cost is indeed achieved,

it could aid in the commercialization of stirling engines that are currently underway.”

Sunpower

Sunpower Inc. is a small company in Ohio engaged in Stirling engine development. In December, David Gedeon spoke with president Neill Lane, who was eager to support our work in any way he could. Sunpower has identified regenerator inefficiency as one of the chief obstacles in commercializing of Stirling engines. Their support would likely take the form of information about specific regenerator applications in Stirling engines and in-situ testing of any regenerators we might develop. Sunpower’s role would be much like that of Stirling Technology Co. Neill Lane has tentatively agreed to send a Sunpower representative to our next project review meeting in Cleveland, planned for March, 2001.

Selection of Random-Fiber Regenerator Materials

At our kick-off meeting in October, 2000, we decided to focus our attention on regenerators of the random-fiber type, rather than woven screens or wrapped foils, which are the two other main contenders sometimes used in Stirling engines. Random fiber felts are much cheaper to manufacture than woven screens and perform about as well. And they generally outperform wrapped foils. According to idealized one-dimensional theories, it should be the other way around, with foil regenerators providing the best (minimum) pressure drop performance. But, in practice, foil regenerators are difficult to manufacture in such a way that the theoretically good performance can be achieved. For example, it is difficult to maintain spacings between the foils (required to achieve uniform flow), the foils need to be “interrupted” to disturb the boundary layer and improve thermal performance and the foils need to be “penetrated” to permit flow redistribution. The Stirling-cycle literature is littered with cases of failed foil regenerators. For this reason we chose random-fiber regenerators as the most promising for commercialization.

Random fiber matrices also offer many opportunities for improvement. They are amenable to changes in porosity, packing structure, fiber shape and fiber orientation — all of which are likely to affect performance. Our job is to discover exactly how.

Involvement of Random-Fiber Manufacturers

Early on, we recognized the importance of involving in our research an actual manufacturer of random fiber materials. If we are going to recommend changes to the structure of random fiber regenerator matrices, they should be changes that can be made. So we set about to enlist the help of one of the several companies doing work in this area, with the hope that they would:

1)Provide feedback as to whether any matrix modifications we come up with are possible to manufacture.

2)Make suggestions on promising alternatives we might investigate, based on their own research.

3)Fabricate one or more test matrices for us based on the outcome of items 1 and 2.

We are hoping to set up an informal working arrangement (without exchange of money) on the grounds that we can both benefit by the end-result of an improved regenerator matrix.

Bekaert Fibre Technologies

We initially identified Bekaert Fibre Technologies (a Belgium corporation) as a promising company to work with. Bekaert manufactures random-fiber metal felt for filtration purposes and claims to be the world’s largest independent wire and wire product manufacturer. Their filtration division has a North American office in North Carolina.

Bekaert bundle-draws circular fibers down to 2 m diameter from various metal alloys (typically 316L stainless steel) at reasonable prices. They also produce non-circular “shaved” fibers at even more reasonable prices. And they have developed their own methods for laying down the fibers into a homogeneous mat. They also routinely compress then sinter the fiber mat into a semi-rigid structure. All to rigid quality-controlled standards.

This quarter we made contact with Johan Saelens, general manager of their North American division, who expressed interest in supporting our work but had to first discuss the matter with the main office in Belgium. He has not yet gotten back to us with the results of that discussion.

U.S. Filter

Should we fail to reach agreement with Bekaert, we plan to contact the Fluid Dynamics division of U.S. Filter, another manufacturer of random-fiber materials. U.S. Filter was not our first choice because it represents a conglomeration of acquired companies. It will probably be difficult to find someone in the corporate hierarchy with enough authority to help us, yet sufficient understanding of the random-fiber process and market to appreciate what we are doing.

Auburn University

Not exactly a manufacturing company, Auburn University (in Alabama) has a funding agreement with NASA which provides them with funding for technology work in several areas. One area of interest is porous random fiber materials for heat pipe wicks and, possibly, for regenerator materials. They haven’t developed porous material specifically for regenerators before but have expressed interest in working in that area. They believe they have a fabrication technique that may result in low cost random fiber material fabrication. There appears to be some possibility that they might be interested in fabricating a regenerator or regenerators for us (but this would likely require additional funding from NASA). This possibility will be explored further, just in case agreements with Bekaert and U.S. Filter fail to materialize.

Experimental Design

In order to start characterization of a simulator of a Stirling engine, we needed to cast the “representative engine” parameters into dimensionless parameters. To do so, we sought help from STC in identifying the raw data that would represent an engine regenerator of interest to them. We also extracted dimensionless parameters from the NASA Contractor’s Report “A Survey of Oscillating Flow in Stirling Engine Heat Exchangers” written by Simon and Seume at the University of Minnesota.

We have prepared the oscillatory flow test facility for the experiment. This involved changing some oil, greasing and softening some seals.

Graduate student Yi Niu will make this project her Ph.D. thesis topic at the University of Minnesota. She arrived in early January and has read introductory material and has become acquainted with operation of the facility.

In the fall, Federico Ottaviani and Terry Simon assembled a list of geometric and operational parameters that must be chosen, or at least considered, in designing the new test section. These were sent in the form of an email message to Songgang Qiu at STC on the 22nd of November. Since then, he used that list to determine operating points that are achievable with our facility and are of interest, as determined from the NASA-CR mentioned above. We are now relating these points to the STC data.

Our basic test section will consist of a duct of simple geometry (cylinder or rectangle) with holes drilled in the sides through which straight wires can be inserted completely through the test section. Initially, we will test a crossed-rod arrangement of wires consisting of uniformly spaced rows of wires, each layer rotated 90 degrees from the previous layer. Later, we will introduce randomness to the wire orientation by varying the locations of the holes drilled through the duct wall according to some computer aided design process, followed by computer-aided machining of the actual holes. With the precise location of the holes known, we will be in a position to completely specify the positions of all the wires inside, for purposes of (1) developing mathematical characterizations of the "randomness" of the wire spatial distribution and (2) generating computational grids for the supporting CFD analysis. In the area of matrix “randomness” we plan to investigate the effects of local clumping of wires, local flow channels (opposite of clumping) and fiber orientation (whether transverse to the flow direction or somewhat parallel).

Results from a study of the flow behavior at the entrance/exit of the regenerator will be valuable toward learning how to minimize flow distribution losses associated with getting the flow into and out of the regenerator. One option is to partially block the entrance to our experimental regenerator matrix and then observe the results. In doing so, we can introduce the flow through a duct section whose centerline is asymmetric to the regenerator centerline and observe this redistribution under oscillatory flow conditions.

Preliminary CFD Modeling

During our kick-off meeting in October, 2000, we discussed two options for CFD modeling. The first option would generate a computational grid fitted to each individual wire of the test section (that is a grid fine enough to resolve all the details of the flow around the individual wires). With this approach we would need relatively less in the way of confirming experimental measurements. Ensemble averaged gas temperatures at various locations in the matrix might suffice. However, many wires would require modeling, which might make the computation too costly in terms of time and memory. The second option would model the test matrix with a much coarser computational grid using a set of equations developed for porous flow. The flow around individual wires would not be resolved but rather the subgrid-scale transport of momentum and energy would be correlated from local matrix flow conditions and structural geometry. With this approach we would need relatively more in the way of experimental measurement support. Instantaneous measurement of fluctuating velocity components at a large number of points in the matrix would likely be required.

At a meeting at Cleveland State University in December, we decided that limited detailed modeling was the only viable choice. Detailed modeling provides the basis for macroscopic equation formulations, not the other way around. If we were to use a macroscopic equation set in advance, we would have no way of resolving the detailed differences among random-fiber matrices, as we hope to do.

We also decided to restrict our CFD modeling to steady flows for the time being, adding the complication of time-varying flow only later, if necessary. Under this plan we would model a regenerator matrix over a range of steady Reynolds numbers representative of different instants during the periodic cycle, assume that the essential fluid dynamics of each case is nearly the same as that of the time-varying solution at the corresponding instant. Modeling time varying flows right off the bat would be too costly in terms of computation time. Especially considering we will already be pushing the computational envelope to the limit by modeling as many wires as possible.

In December, Tony Bougebrayel (former CSU student), working with Mounir Ibrahim and Roy Tew, began simulating a single cylinder in cross flow using the Fluent CFD software. The solver used was the "Implicit Coupled, 2D, Steady State, Laminar flow" with 2nd order discretization. The grids used were bricks with B.L. growth over the cylinder and 17,777 cells for ReD= 87. The plot below shows the velocity vectors colored by velocity magnitude.

Work has just begun to model an array of nine cylinders consisting of three rows of three cylinders each, each row rotated 90 degrees from the previous — crossed-rods, in other words. The objective of this preliminary modeling is to compare the results of CFD modeling against correlations for pressure-drop and heat transfer in an identical geometry published by Kays and London.[1] Before we can do this, we will first have to develop a method of reducing the large number of detailed solution outputs into overall correlations for matrix friction factor (pressure drop), Nusselt number (heat transfer) and effective gas conductivity (axial conduction) as functions of Reynolds number. Once done for arrays of crossed-rods, it should be easy to extend the methodology to more complicated arrays (non-uniform spacings, orientations, etc.) representing random-fiber matrices. Then to use these correlations in one-dimensional computational models.

1

[1] W.M. Kays & A.L London, Compact Heat Exchangers, third edition, McGraw-Hill, (1984) — Figures 10-98 through 10-100, pp. 274 – 276.