Massive-Parallel Algorithms for Identifying the Production Batches of Semiconductor Devices

Abstract

The modern development of rocket and space technology sets stringent requirements for hardware developers - improving the overall spatial and weight characteristics, increasing the hardware functionality and increasing the period of active existence. The range of electronic devices used in spacecraft on-board equipment must have the following set of specific conditions: (1) a large set of tasks performed by rocket and space technology leads to the necessity of using a wide variety of electronic devices standard types with an extremely small quantitative requirement; (2) the wide electronic devices functional range requires the use of various technologies in production, many of which are unique and absent in the Russian Federation; (3) stringent requirements for the active existence period in the absence of repair opportunities lead to super-stringent requirements for reliability and resistance to outer space destabilizing factors. Due to the fact that the Russian Federation does not produce specialized electronic devices for rocket and space technology, the problems of completing on-board equipment are solved by using foreign-made electronic devices and domestically made electronic devices processed by specialized test technical centers with the total incoming control, additional screening tests, diagnostic non-destructive testing and selective destructive physical analysis. To reduce the computation time for the selection of potentially unreliable electronic products for space application, we propose parallel algorithms with a greedy agglomerative heuristic procedure for solving problems of automatic grouping with large amounts of data, adapted to the CUDA architecture.