Blast Induced Vibration Estimation by Using Different Machine Learning Methods in a Tunnel Excavation

doi:10.21203/rs.3.rs-336167/v1

Download PDF

Research Article

Blast Induced Vibration Estimation by Using Different Machine Learning Methods in a Tunnel Excavation

https://doi.org/10.21203/rs.3.rs-336167/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

This study aimed to develop a new model in which rock characteristics, blasting design parameters and excavation planning were also considered by using various machine learning methods in a funicular line excavation where blast-induced vibrations could not estimate with a high correlation by using the commonly and successfully used PPV-SD (Peak Particle Velocity-Scaled Distance) estimation formula. In addition to developing a new model, another aim was to reveal the effect of rock characteristics, blasting and excavation planning parameters on PPV estimation numerically in the form of weights.

For this purpose, 225 events in 57 shots were recorded in the funicular line excavation. Each blasted cross-section's rock characteristics were obtained from on-site inspection and geological reports. At first, recorded blasting vibration data were evaluated using the well-known PPV–SD equation, and it was seen that the relationship between PPV and SD was not able to represent the site-specific vibration attenuation. Therefore, the obtained data were evaluated with Random Forest and other Machine Learning Methods. In these evaluations, RQD, UCS, unit of advance, the maximum charge per delay, the cross-sectional area of tunnel face, total charge, and distance between shot point and vibration measurement station were used as inputs, and peak particle velocity was used as output. The results showed that the random forest model's prediction accuracy was more acceptable than the well-known PPV–SD equation and other machine learning methods. Another significant finding of the study is that parameters not considered in PPV estimation, such as UCS, RQD, and cross-sectional area of tunnel face, may be more effective than the commonly used scaled distance.

Environmental Engineering

Environmental Policy

Tunnel blasting

blast-induced ground vibrations

rock characteristics

Random forest

machine learning

Table 1

A representative set of obtained data

Shot No	Vibration Measurement Device	PPV (mm/s)	F (Hz)	RQD	UCS (MPa)	Cross Sectional Area (m²)	Unit of Advance (m)	Total Charge (kg)	Max. Charge per Delay (kg)	Distance (m)	SD
1	BE10704	31.62	100	60	100	110	1.5	52.98	0.654	16.59	19.11
	BE10708	26.29	64	60	100	110	1.5	52.98	0.654	21.84	25.16
	BE10705	26.03	100	60	100	110	1.5	52.98	0.654	20.73	23.88
	BE10706	6.858	100	60	100	110	1.5	52.98	0.654	30.91	35.61
	BE10771	10.29	100	60	100	110	1.5	52.98	0.654	40.83	47.04
3	BE10705	14.48	100	50	70	110	1.5	34.44	0.436	28.36	37.4
	BE10706	8.382	100	50	70	110	1.5	34.44	0.436	36.09	47.59
	BE10771	9.525	100	50	70	110	1.5	34.44	0.436	28.66	37.8
5	BE10705	8.001	100	55	60	110	1.4	21.255	0.545	47.61	58.29
	BE10706	6.223	100	55	60	110	1.4	21.255	0.545	56.76	69.49
	BE10771	3.937	85.33	55	60	110	1.4	21.255	0.545	25.12	30.75
7	BE10705	2.032	34.13	35	55	110	2.4	32.7	0.545	26.3	32.2
	BE10706	6.477	100	35	55	110	2.4	32.7	0.545	30.31	37.11
	BE10771	3.683	100	35	55	110	2.4	32.7	0.545	36.81	45.06
12	BE10705	1.27	56.89	55	70	110	3	68	1.36	45.58	41.14
	BE10706	9.652	100	55	70	110	3	68	1.36	28.66	25.87
	BE12923	3.175	85.33	55	70	110	3	68	1.36	49.62	44.79
16	BE10705	1.397	100	55	70	110	2.5	81.75	1.64	47.82	40.55
	BE10706	19.18	100	55	70	110	2.5	81.75	1.64	30.18	25.59
	BE12923	3.429	85.33	55	70	110	2.5	81.75	1.64	51.52	43.69
19	BE10708	6.477	100	60	50	36.5	1.5	54.38	0.872	34.29	35.89
19	BE10771	2.54	100	60	50	36.5	1.5	54.38	0.872	34.23	35.83
21	BE10706	9.779	73.14	35	50	36.5	5	106	1.28	42.8	39.42
21	BE10771	3.429	100	35	50	36.5	5	106	1.28	41.02	37.78
25	BE10706	2.159	85.33	50	40	36.5	5	104.97	1.64	392.33	332.69
25	BE10771	0.889	51.2	50	40	36.5	5	104.97	1.64	430.59	365.13
30	BE10706	1.651	73.14	26	35	36.5	5.5	32.7	1.09	616.09	598.64
30	BE10771	0.889	42.67	26	35	36.5	5.5	32.7	1.09	654.46	635.93
31	BE10706	0.762	64	26	50	36.5	9	30.52	1.09	608.23	591.01
31	BE10771	0.635	56.89	26	50	36.5	9	30.52	1.09	646.58	628.27
33	BE10706	1.016	64	40	70	36.5	4.5	35.36	1.36	601.71	543.09
33	BE10771	0.635	39.38	40	70	36.5	4.5	35.36	1.36	640.03	577.68
35	BE10706	1.143	64	40	40	36.5	4	64.31	1.09	598.24	581.3
35	BE10771	0.635	42.67	40	40	36.5	4	64.31	1.09	636.55	618.52
39	BE10706	1.27	51.2	45	60	36.5	2.5	32.7	1.36	575.43	519.37
39	BE10771	0.635	42.67	45	60	36.5	2.5	32.7	1.36	613.67	553.89

Table 2

Statistical characteristics of data set that reveals Minimum (Min), Maximum (Max), Mean and Standard Deviation.

Influence Parameters	Description	Min	Max	Mean	Standard Deviation
RQD	Input	26	65	52.373	8.495
UCS	Input	27	100	69.031	17.143
Unit of advance (m)	Input	0	9	2.202	1.232
Charge per delay (kg)	Input	0.436	1.74	0.806	0.408
Cross-sectional area (m²)	Input	36.53	110	89.428	33.061
Distance (m)	Input	16.5	654.46	113.314	191.677
PPV	Output	0.508	31.62	5.988	5.242

Table 3

Comparisons of prediction performances by using various approaches (Training results).

Methods	RMSE	MAE	R
Neural Network	4.6658	3.1076	0.6087
SVM	4.6527	2.9808	0.5077
Linear Regression	4.3715	3.1552	0.5491
Random Forest	1.1349	0.8428	0.9783

Table 4

Comparisons of prediction performances by using various approaches (CV results).

Methods	RMSE	MAE	R
Neural Network	5.3375	4.2316	0.3635
SVM	4.6744	3.0212	0.4979
Linear Regression	4.4336	3.2095	0.5312
Random Forest	3.0356	2.2297	0.8145

Table 5

Comparisons of prediction performances by using various approaches (Holdout results).

Methods	RMSE	MAE	R
Neural Network	5.195	4.2827	0.5587
SVM	4.558	3.0383	0.5493
Linear Regression	4.6062	3.4342	0.5537
Random Forest	2.7441	2.1248	0.8668

Table 6

Input’s importance level on PPV Estimation and number of nodes in RF

Input	Importance Level on PPV Estimation	Number of Nodes
Unit of Advance (m)	24,35	2589
Max. Charge per Delay (kg)	44,66	1281
Distance (m)	59,77	6715
RQD	63,94	659
Cross-Sectional Area (m²)	78,24	80
UCS (MPA)	105,38	634

Download PDF

Version 1

posted

You are reading this latest preprint version

Blast Induced Vibration Estimation by Using Different Machine Learning Methods in a Tunnel Excavation

Status:

Version 1

Abstract

Figures

Full Text

Tables

Status:

Version 1