摘要
由于日间干扰大容易误报,声学检漏通常在夜间进行,这与实时监测需求不太符合。为此,开展全天候声学检漏试验。通过现场增加日间音频信号采集,比较分析日间/夜间音频的信号特征、时频谱图及机器学习算法识别结果,探究日间检测的可行性。试验结果表明,噪音计日间检测准确率和AUC值达到0.890和0.959,符合全天候实时监测需求。
Due to large interference during the daytime and possibility of false alarm,acoustic leak detection is usually carried out at night.That is not consistent with the demand for real-time monitoring.Therefore,daytime and night acoustic leak detection test is carried out.The feasibility of daytime detection is explored by increasing the collection of daytime sound signals on site,comparing and analyzing the signal characteristics,time-frequency spectrum and machine learning algorithm recognition results of daytime/night sound signals.The test results show that the daytime detection accuracy and AUC value of the noise logger reach 0.890 and 0.959,respectively,and the whole day real-time monitoring is feasible.
作者
王海涛
彭浩
齐丽强
葛奇
黄清龙
徐哲
WANG Haitao;PENG Hao;QI Liqiang;GE Qi;HUANG Qinglong;XU Zhe(School of Automation,Hangzhou Dianzi University,Hangzhou 310018,China;Hangzhou Water Group Co.Ltd.,Hangzhou 310009,China;715th Research Institute of China Shipbuilding Industry Corporation,Hangzhou 310023,China)
出处
《杭州电子科技大学学报（自然科学版）》
2024年第5期47-56,共10页
Journal of Hangzhou Dianzi University：Natural Sciences
基金
国家自然科学基金项目(U1509205)
杭州市重点科技研发计划项目(20212013B06)。
关键词
噪音计
全天候检测
声学检漏
机器学习算法
noiselogger
daytimeandnightleakdetection
acousticleakdetection
machinelearning algorithm
