Sound Recognition and Early Warning Mechanism for Liquid Aluminum Leakage Based on Improved EfficientNetV2

doi:10.12141/j.issn.1000-565X.250006

Abstract

Abstract:

Liquid aluminum leakage is the direct cause of explosion accidents in aluminum deep-well casting processes. To address the practical engineering challenges of strong lag, low accuracy, and limited monitoring range in existing leakage detection methods, this paper proposed a sound recognition method for liquid aluminum leakage based on an improved EfficientNetV2 model. This method utilizes acoustic characteristics to identify leaks, thereby expanding the monitoring range. The core enhancement involves optimizing the stacking factor and integrating an efficient channel attention mechanism into the EfficientNetV2 architecture to further improve the recognition speed and accuracy. Firstly, a sound database encompassing seven types of acoustic scenes was constructed by collecting audio data under different scenarios using pickups. Then, log-Mel spectrograms were extracted from the sound signal as the feature set and fed into the improved EfficientNetV2 model for training and validation, finally yielding the liquid aluminum leakage sound recognition model. The experimental results show that the recognition accuracy of the improved EfficientNetV2 reaches 95.48%. Compared to the original EfficientNetV2, ResNet, RegNet and DenseNet, the proposed model requires only 12.34%, 8.64%, 11.14%, and 10.80% of the floating point operations, and 11.37%, 9.55%, 15.95%, and 17.24% of the parameters, respectively. Furthermore, it processes 6.53, 6.14, 4.41, and 8.00 times more frames per second in a CPU environment, confirming its fast and accurate recognition performance. In addition, a risk early-warning mechanism for liquid aluminum leakage was established based on the proposed sound recognition method and deployed for real-time risk monitoring in a casting unit. Practical application results verify the effectiveness of both the identification method and the warning mechanism, providing a valuable technical reference for preventing explosion accidents in aluminum deep-well casting.

Key words: aluminum deep-well casting, liquid aluminum leakage, sound recognition, risk early warning, improved EfficientNetV2, logarithmic Mel-spectrogram

CLC Number:

TP29

LIANG Yanhui, WEN Chengjie, YAN Junwei, ZHOU Xuan, ZHANG Hongtao. Sound Recognition and Early Warning Mechanism for Liquid Aluminum Leakage Based on Improved EfficientNetV2[J]. Journal of South China University of Technology(Natural Science Edition), 2026, 54(2): 38-51.

Figures/Tables 12

Fig.1

Fig.2

Fig.3

Fig.4

Table 1

Table 2

Fig.5

Fig.6

Fig.7

Table 3

Fig.8

Fig.9

References 33

[1]	佘欣未，蒋显全，谭小东，等．中国铝产业的发展现状及展望［J］．中国有色金属学报，2020，30（4）：709-718.
	SHE Xin-wei， JIANG Xian-quan， TAN Xiao-dong，et al ．Status and prospect for aluminum industrial development in China［J］．The Chinese Journal of Nonferrous Me-tals，2020，30（4）：709-718.
[2]	YI X J， LU Y L， HE G Z，et al ．Global carbon transfer and emissions of aluminum production and consumption［J］．Journal of Cleaner Production，2022，362：132513/1-9.
[3]	YUE L， LI H， KATGERMAN L，et al ．Recent advances in hot tearing during casting of aluminum alloys［J］．Progress in Materials Science，2020，117：100741/1-77.
[4]	沈正祥，陈虎，王杜，等．铸造工艺中蒸汽爆炸事故形成机制分析［J］．中国安全生产科学技术，2018，14（5）：150-154.
	SHEN Zhengxiang， CHEN Hu， WANG Du，et al ．Analysis on formation mechanism of steam explosion accident in casting process［J］．Journal of Safety Science and Technology，2018，14（5）：150-154.
[5]	沈正祥，陈虎，陈定岳，等．半封闭空间蒸汽爆炸的影响因素分析［J］．中国安全科学学报，2018，28（7）：52-57.
	SHEN Zhengxiang， CHEN Hu， CHEN Dingyue，et al ．Analysis of factors affecting steam explosion in semi-confined space［J］．China Safety Science Journal，2018，28（7）：52-57.
[6]	王铭琳．铝加工深井铸造安全生产风险动态评价与分级报警方法研究［D］．广州：华南理工大学，2023.
[7]	刘杰，王壬．深井铸造熔融铝泄漏遇水爆炸破坏影响研究［J］．中国安全科学学报，2023，33（S1）：105-111.
	LIU Jie， WANG Ren ．Study on damage influence of explosion caused by molten aluminum leakage in deep well casting with water［J］．China Safety Science Journal，2023，33（S1）：105-111.
[8]	国务院安全生产委员会办公室．国务院安委会办公室关于今年以来铝加工（深井铸造）企业生产安全事故有关情况的通报［EB/OL］．（2024-09-22）［2024-12-04］．.
[9]	RICHTER R T， EKENES J M ．Cause and prevention of explosions involving DC casting of aluminum sheet ingot［M］．Essential readings in light metals．Cham：Springer，2016：1085-1090.
[10]	LI L， XU K L， YAO X W，et al ．Probabilistic analysis of aluminium production explosion accidents based on a fuzzy Bayesian network［J］．Journal of Loss Prevention in the Process Industries，2021，73：104618/1-8.
[11]	LI L， XU K L， YAO X W，et al ．A method for the core accident chain based on fuzzy-DEMATEL-ISM：an application to aluminum production explosion［J］．Journal of Loss Prevention in the Process Industries，2024，92：105414/1-9.
[12]	LI X， CHEN B， ZHOU N，et al ．Experiment research on explosion process of high temperature molten aluminum in contact with water［J］．Heat Transfer Research，2022，53（12）：75-90.
[13]	刘子健，沈致和，陈兵，等．高温熔融铝液柱遇水蒸汽爆炸的压力及固体产物分布实验研究［J］．实验力学，2022，37（1）：88-98.
	LIU Zijian， SHEN Zhihe， CHEN Bing，et al ．Experimental study on steam explosion pressure and solid pro-duct distribution of high-temperature molten aluminum liquid column in contact with cooling water［J］．Journal of Experimental Mechanics，2022，37（1）：88-98.
[14]	应急管理部．中华人民共和国应急管理部令（第10号）工贸企业重大事故隐患判定标准［EB/OL］．（2023-04-17）［2024-12-04］．.
[15]	广州远正智能科技股份有限公司．一种用于铝加工分流盘铝液泄露的监测报警系统及方法：ZL 202211116411.8［P］．2022-12-09.
[16]	YAN J W， LI X， ZHOU X ．Intelligent identification of liquid aluminum leakage in deep well casting production based on image segmentation［J］．Applied Sciences-Basel，2024，14（13）：5470/1-18.
[17]	张永明，王克威，张启兴，等．一种基于红外图像特征融合的高温铝液模拟泄漏监测算法［J］．安全与环境学报，2020，20（2）：518-523.
	ZHANG Yong-ming， WANG Ke-wei， ZHANG Qi-xing，et al ．Simulated leakage monitoring algorithm for high-temperature molten aluminum based on the infrared image feature fusion［J］．Journal of Safety and Environment，2020，20（2）：518-523.
[18]	KONG L Y， LIU X Z， SHI Z N，et al ．Numerical simulation of liquid aluminum leakage in casting process［J］．Transactions of Nonferrous Metals Society of China，2021，31（1）：297-305.
[19]	LEE J， KIM T Y， BEAK S，et al ．Real-time pose estimation based on ResNet-50 for rapid safety prevention and accident detection for field workers［J］．Electro-nics，2023，12（16）：3513/1-22.
[20]	YANG G T， HU S L， WANG L T ．Enhanced anomaly detection in compressor components using deep learning and an attribute updating model［J］．Industrial & Engineering Chemistry Research，2024，63：18027-18042.
[21]	辛景舟，刘倩茹，唐启智，等．融合密集卷积网络和注意力机制的拱桥损伤识别［J］．振动与冲击，2024，43（14）：18-28，36.
	XIN Jingzhou， LIU Qianru， TANG Qizhi，et al ．Damage identification of arch bridges based on dense convolutional networks and attention mechanisms［J］．Journal of Vibration and Shock，2024，43（14）：18-28，36.
[22]	TAN M X， LE Q V ．EfficientNetV2：smaller models and faster training［C］∥ Proceedings of the 38th International Conference on Machine Learning．［S.l.］：OpenReview，2021：10096-10106.
[23]	ANVAR A A T， MOHAMMADI H ．A novel application of deep transfer learning with audio pre-trained models in pump audio fault detection［J］．Computers in Industry，2023，47：103872/1-12.
[24]	YU X C， LI X W ．Sound recognition method of coal mine gas and coal dust explosion based on GoogLeNet［J］．Entropy， 2023，25（3）：412/1-12.
[25]	GLOWACZ A， TADEUSIEWICZ R， LEGUTKO S，et al ．Fault diagnosis of angle grinders and electric impact drills using acoustic signals［J］．Applied Acoustics，2021，179：108070/ 1-14.
[26]	RUSTAM F， ISHAQ A， HASHMI M S A，et al ．Railway track fault detection using selective MFCC features from acoustic data［J］．Sensors，2023，23（16）：7018/1-21.
[27]	LUO L Y， GUO S T， WANG M，et al ．Adaptive noise reduction algorithm based on SPP and NMF for environmental sound event recognition under low-SNR conditions［J］．Wireless Communications and Mobile Computing，2023，2023：6582296/1-11.
[28]	ER M B， ILHAN N， GUNER K，et al ．A novel approach for autocorrelation-based classification of environmental audio signals using recurrent network and metaheuristic algorithms［J］．Signal Image and Video Processing，2025，19（8）：600/1-13.
[29]	SINGH V K， SHARMA K， SUR S N ．A survey on preprocessing and classification techniques for acoustic scene［J］．Expert Systems with Application，2023，229：120520/1-35.
[30]	SAHIDULLAH M D， SAHA G ．A novel windowing technique for efficient computation of MFCC for speaker recognition［J］．IEEE Signal Processing Letters，2013，20（2）：149-152.
[31]	ABDUL Z K， AL-TALABAni A K ．Mel frequency cepstral coefficient and its applications：a review［J］．IEEE Access，2022，10：122136-122158.
[32]	WANG Q， WU B， ZHU P，et al ．ECA-Net：efficient channel attention for deep convolutional neural networks［C］∥ Proceedings of 2020 IEEE/CVF Confe-rence on Computer Vision and Pattern Recognition．Seattle：IEEE，2021：11531-11539.
[33]	TANG W J， DAI Q， HAO F ．An efficient knowledge distillation-based detection method for infrared small targets［J］．Remote Sensing，2024，16（17）：3173/1-15.

声音类型	录音时段	距离/m	是否遮挡	是否可分辨	声音类型
风机声	铸造前	7.5	是	否	背景噪声
叉车声	铸造前	11.0	是	否	背景噪声
电机声	铸造中	4.0	是	否	背景噪声
水流声	铸造中	0.5~1.0	否	是	环境声音
风扇声	铸造中	3.0	否	否	背景噪声
说话声	铸造中	1.0~2.0	否	否	背景噪声
蜂鸣声	铸造中	4.0	否	否	背景噪声
锯棒声	铸造后	8.0	否	是	环境声音
天车声	铸造后	8.0	否	否	背景噪声

类型	声音场景定义	示例场景
A类	铝液泄漏声音	非铸造期间，熔炼炉出口未完全封堵导致铝液流入水井的场景。
B类	铝液泄漏声音与水流声音混叠	铸造期间，铝液泄漏至水井的场景。
C类	铝液泄漏声音与锯棒声音混叠	铸造后铝棒切割期间，熔炼炉出口未完全封堵导致铝液流入水井的场景。
D类	铝液泄漏声音、水流声音与锯棒声音混叠	某个铸造单元铸造后铝棒切割期间，同时另一个铸造单元铸造期间铝液泄漏至水井的场景。
E类	水流声音	铸造前水系统测试时冷却水流入水井或者铸造期间没有铝液泄漏的场景。
F类	锯棒声音	铸造后铝棒切割的场景。
G类	水流声音与锯棒声音混叠	某个铸造单元铸造后铝棒切割期间，同时另一个铸造单元铸造前水系统测试时冷却水流入水井或者铸造期间没有铝液泄漏的场景。

分类模型	准确率/%	声音类别	精确率/%	召回率/%	浮点运算次数/10⁶	参数量/10⁶	FPS/（f‧s^-1）
原始 EfficientNetV2	96.00	A	100.00	100.00	6 775	21.46	5.03
		B	96.61	95.00
		C	99.34	100.00
		D	95.17	85.33
		E	95.08	96.67
		F	100.00	99.33
		G	86.71	95.67
简化的 EfficienNetV2	94.43	A	100.00	100.00	843	2.73	31.05
		B	93.88	92.00
		C	97.35	98.00
		D	96.86	82.33
		E	92.16	94.00
		F	98.01	98.67
		G	84.46	96.00
改进的 EfficientNetV2	95.48	A	100.00	100.00	836	2.44	32.86
		B	95.58	93.67
		C	100.00	99.00
		D	97.62	82.00
		E	93.79	95.67
		F	99.01	100.00
		G	84.48	98.00
ResNet	95.81	A	100.00	100.00	9 673	25.56	5.35
		B	96.60	94.67
		C	100.00	99.67
		D	95.09	84.00
		E	94.77	96.67
		F	99.67	100.00
		G	85.67	95.67
RegNet	95.24	A	100.00	100.00	7 507	15.30	7.45
		B	96.19	92.67
		C	99.33	98.33
		D	97.24	82.33
		E	92.93	96.33
		F	98.35	99.33
		G	84.68	97.67
DenseNet	95.71	A	100.00	100.00	7 741	14.15	4.11
		B	96.28	95.00
		C	100.00	99.33
		D	96.85	82.00
		E	95.07	96.33
		F	99.34	100.00
		G	84.39	97.33