Anomaly Detection for Tower Crane Alarm Surge Based on Copula Function

Konferenz: ICMLCA 2021 - 2nd International Conference on Machine Learning and Computer Application
17.12.2021 - 19.12.2021 in Shenyang, China

Tagungsband: ICMLCA 2021

Seiten: 6Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Autoren:
Shi, Hongyang; Wang, Hongjun (Shenyang Institute of Computing Technology, Chinese Academy of Sciences, Shenyang, China & University of Chinese Academy of Sciences, Beijing, China)
Lu, Ming (Shenyang Institute of Computing Technology, Chinese Academy of Sciences, Shenyang, China & Infrastructure & Cloud Service, SSG, Lenovo, Beijing, China)
Zhang, Yonghong (R.C of Big Data and Artificial Intelligence Technology, Shandong University, Jinan, China & Beijing Zhongke Zhihe Digital Technology Co., Ltd, Beijing, China)
Dong, Zengshou (School of Electronic Information Engineering, Taiyuan University of Science and Technology, Shanxi, China)

Inhalt:
In recent years, tower crane safety accidents have occurred frequently, resulting in casualties and significant losses of construction site property. In order to reduce the probability of tower crane accidents, quality supervisor needs a method that can cause anomaly detection of alarm surges in a certain time period of the tower crane. Due to the lack of generalization capability of existing methods for new equipment and the large calculation overhead and weak interpretability, quality supervisor is often unable to effectively detect the risks brought about by the improper operation of special operators. This paper adopts the unsupervised anomaly detection method based on copula function, estimates the empirical copula function according to the empirical cumulative distribution of input data through nonparametric method, and then estimates the tail probability of joint distribution in all dimensions to evaluate the anomaly. In order to evaluate the effect of the model, this paper calculates the sudden increase point by comparing with each time series value and historical value, and uses it as an anomaly label. By comparing other different classes of anomaly detection algorithms horizontally, the copula function-based anomaly detection method for the scenario in this paper has better generalization and generalization ability, and the overall performance is better than other algorithms. In addition, compared with other methods, this method has the characteristics of low calculation consumption, scalability and high interpretability. For the anomaly detection of the total amount of tower crane alarm, this method has high application value.