论文部分内容阅读
本文选取1974年以来我国华北、西南地区15个发生过6级以上地震的地区时段和20个没有发生过6级以上地震的地区时段作为模式识别研究的对象,用地震空区、条带、b值、集中度、密集、平静、安全域等地震活动性方法构成问题征询表。采用修改的CORA-3算法进行模式识别。消除等效特征与弱特征,进行广泛的K阈值、对象及参数变化的控制试验。计算结果表明,识别较稳定,地震条带、b值、集中度、安全域等组成的地震活动性图象特征具有重要的前兆意义。综合模式识别能力明显高于任一单项方法。本文改进了一般的识别准则,采用综合决策方案,提高了模式识别结果的可信度。
In this paper, we selected 15 regional earthquake periods of magnitude 6 and above and 20 regional periods of no magnitude 6 earthquake occurred in North China and Southwest China since 1974 as the object of pattern recognition. Value, concentration, dense, calm, security domain and other seismic activity constitute a questionnaire. The modified CORA-3 algorithm is used for pattern recognition. Eliminate equivalent features and weak features, and conduct extensive control tests on K thresholds, objects and parameter changes. The calculation results show that the seismicity image features of relatively stable, seismic band, b value, concentration, and security domain have important foreshadow significance. Comprehensive pattern recognition ability is significantly higher than any single method. In this paper, the general recognition criteria are improved and the comprehensive decision-making scheme is adopted to improve the credibility of the pattern recognition results.