基于多源试验大数据融合的装备质量缺陷诊断与溯源研究

The Study on Equipment Quality Defect Diagnosis and Traceability Based on Multi-source Test Big Data Fusion

摘要:

针对装备试验中多源数据孤岛、质量缺陷诊断准确率低、溯源定位模糊等问题，提出一套“融合-诊断-溯源”全链条技术方案；构建多源试验数据采集体系，采用改进加权融合算法实现异构数据集成，设计CNN-LSTM融合模型提取时空特征并识别缺陷类型，基于贝叶斯网络建立溯源模型定位根因等方法达到了试验数据融合准确率、缺陷诊断准确率、溯源平均误差等指标均有明显向目标指数变化，较传统方法综合性能明显提升，为装备试验质量管控提供技术支撑。

Abstract:

To address issues in equipment testing such as multi-source data silos, low accuracy in quality defect diagnosis, and unclear traceability positioning, a full-chain technical solution of "Fusion-Diagnosis-Traceability" is proposed. Measures including constructing a multi-source test data collection system, adopting an improved weighted fusion algorithm to achieve heterogeneous data integration, designing a CNN-LSTM fusion model to extract spatiotemporal features and identify defect types, and establishing a traceability model based on Bayesian networks for root cause localization have led to significant improvements in indicators like test data fusion accuracy, defect diagnosis accuracy, and average traceability error toward the target values. Compared with traditional methods, the comprehensive performance has been remarkably enhanced, providing technical support for quality control in equipment testing.