再设计是产品运行使用一段时间之后的改进设计，是产品开发和更新换代的主要手段，也是航天型号产品研制的主要模式。本项目将性能衰退分析引入到产品再设计过程中，通过大数据分析，揭示产品性能时变规律以及时变性能与设计参数内在关系，从而识别出导致性能异常衰退的设计薄弱环节，为实现产品服役性能稳定可靠提供新的设计理论方法。本项目创新点在于:（1）通过K-均值聚类方法实现多工况划分，应用多模型建模技术构造产品性能衰退联合模型，能有效辨识出多工况下产品性能衰退严重时段以及异常实测数据；（2）依据产品性能要求和故障危害度确定功能模块重要度，挖掘异常实测参数与模块设计参数之间相关度，最终依据重要度和相关度确定需要再设计的关键设计参数；（3）引入环境因子对试验数据和运行数据进行融合，将地面环境下的试验数据转化为运行环境下的等效数据，与运行数据相融合，增加了数据样本容量，提高了再设计决策分析的准确性。

The redesign is the design improvement after the product runs for a period of time., and a major means for product development and upgrading, also the main mode for the development of aerospace products. This project intends to introduce the performance analysis to product redesign process, through big data analysis, the regulations of product time-varying performance and the intrinsic relations of performances and design parameters have been revealed, so as to identify the design weak links resulting in abnormal performance degradation, the project provides a new design theory for achieving stable and reliable product operation. The contribution of this project lies in: (1) obtain the multi-condition classification through the k-means clustering method and effectively identify the abnormal field data and the product performance degradation under multi-conditions by applying the multi-model technology to construct the combined product performance degradation-model; (2) The importance weight of functional module was determined according to performance requirements and failure hazard degree. Besides, the correlations between abnormal field data and module design parameters were mined. Then, the to-be-improved design parameters were identified according to the importance of function modules and the above correlations; (3) introduce the environmental factors to the test-operational data fusion process with respect to different distributions of performance parameters for the test data and operational data, respectively. By the appropriate environmental factors, the test data within ground environment can be converted to the equivalent data within operational environment. Through the integration of the test data and the operational data, the sample size can be effectively increased, therefor improving the precision of decision analysis.