Third generation development of an 11-watt Stirling converter

This paper describes recent design enhancements, performance results, and development of an artificial neural network (ANN) model related to the radioisotope Stirling generator (RSG), an 11 W converter designed for remote power applications. Design enhancements include minor changes to improve performance, increase reliability, facilitate fabrication and assembly for limited production, and reduce mass. Innovative modifications were effected to increase performance and improve reliability of the vacuum foil insulation (VFI) package and linear alternator. High and low operating temperature acceptance testing of the engineering model demonstrated the robust system characteristics. These tests were conducted for 1 week of operation each, with rejector temperatures of 95/spl deg/C and 20/spl deg/C, respectively. Endurance testing continues for a complete Stirling converter, the developmental model (DM), with over 25,000 hours of maintenance free operation. Endurance testing of flexures has attained over 540 flexure years and endurance testing of linear motors/alternators has achieved nearly 27,000 hours of operation without failure. An ANN model was developed and tested successfully on the DM. Rejection temperatures were varied between 3/spl deg/C and 75/spl deg/C while load voltages ranged between engine stall and displacer overstroke. The trained ANN model, based solely on externally measured parameters, predicted values of piston amplitude, displacer amplitude and piston-displacer phase angle within /spl plusmn/2% of the measured values over the entire operating regime. The ANN model demonstrated its effectiveness in the long-term evaluation of free-piston Stirling machines without adding the complexity, reduced reliability and increased cost of sophisticated diagnostic instrumentation.