A VMD-EWT-LSTM-based multi-step prediction approach for shield tunneling machine cutterhead torque

doi:10.1016/j.knosys.2021.107213

Knowledge-Based Systems

Volume 228, 27 September 2021, 107213

https://doi.org/10.1016/j.knosys.2021.107213 Get rights and content

Highlights

•
We present a VMD-EWT-LSTM-based multi-step prediction method for cutterhead torque.
•
VMD and EWT decomposes original input sequence into relatively simple subsequences.
•
The appropriate mode quantity K of VMD could be determined adaptively.
•
It only deals with one input reducing calculation time and improving efficiency.
•
Its five-step average accuracy reaches 97.7%, 97.2%, 96.9%, 96.7% and 96.3%.

Abstract

Cutterhead torque is an important operational parameter that reflects the obstruction degree of geological environment to shield tunneling machine. Accurate multi-step prediction for cutterhead torque is of crucial significance for ensuring efficient and safe propulsion of shield tunneling machine. In this study, a novel hybrid multi-step prediction model combining variational mode decomposition (VMD), empirical wavelet transform (EWT) and long short-term memory (LSTM) network is proposed for shield tunneling machine cutterhead torque. To begin with, the VMD is employed to decompose the original cutterhead torque subsequences into some subsequences and residual sequence, and the residual sequence is further decomposed by the EWT. The combination of VMD and EWT significantly reduces the complexity of the original cutterhead torque sequence. On this basis, the LSTM neural network is employed to predict each subsequence in multiple time steps, and finally add the prediction results of each subsequence to realize the multi-step cutterhead torque prediction. To demonstrate performance of the presented VMD-EWT-LSTM-based approach, comparisons with recent prediction algorithms in six datasets is conducted. The results testify that accuracy of the VMD-EWT-LSTM-based prediction approach is better than other methods. In six datasets, from 1st step prediction to 5th step prediction, the average accuracy of presented prediction approach reaches 97.7%, 97.2%, 96.9%, 96.7% and 96.3%, respectively. Hence, the VMD-EWT-LSTM-based approach can accurately predict cutterhead torque of shield tunneling machine in multiple time steps.

Introduction

In terms of tunnel construction, compared with conventional blasting method, the shield construction method is safer, cleaner and more efficient. Therefore, shield tunneling machine is widely used in water conservancy, highway and railway tunnel construction [1], [2], [3], [4]. To ensure safe construction of shield tunneling machine, it is necessary to adjust the operating parameters of the equipment according to the geological environment. However, it is extremely difficult to predict exactly the geological conditions before excavation at present [5], [6], [7]. Compared with the prediction of geological conditions, predicting the shield machine operational parameters seems more feasible in actual tunnel construction. Cutterhead torque is an important operational parameter of shield tunneling machine, reflecting obstruction degree of geological environment to shield tunneling machine. Precisely predicting the cutterhead torque can help operators to adjust operational parameters in advance, which is beneficial to avoid the cutterhead being stuck and trapped, and ensure safe and efficient tunneling construction.

On the one hand, when the geological conditions are complex, the prediction accuracy of the current cutterhead torque prediction method is low, which cannot effectively extract the variation law of cutterhead torque. On the other hand, the current cutterhead torque prediction methods are mainly realizing the prediction of single time step. The prediction time step is short and the practical application value is still limited. If multiple time step prediction of cutterhead torque can be realized, it will be more effective to guide the shield machine driver to adjust operating parameters in advance, which has more practical significance. Since the change rule of cutterhead torque time series is very complicated, it is difficult to effectively learn its long-term change rule. Therefore, multi-step prediction of cutterhead torque is challenging at present.

In order to effectively extract the variation law of cutterhead torque and realize high-precision multi-step prediction of cutterhead torque, this paper proposes a VMD-EWT-LSTM-based multi-step prediction method for cutterhead torque of shield machines. Considering VMD and EWT have excellent characteristics of decomposing nonlinear signals, VMD is used to decompose the original torque sequence of cutterhead to obtain some subsequences and residual sequence, and EWT is employed to decompose the residual sequence. On this basis, the original cutterhead torque sequence with a relatively complicated change law could be decomposed into a series of subsequences with a relatively simple change law, which are easier to predict. Then, LSTM neural network that has excellent ability to extract the time-varying features of time series is used to make multi-step prediction for each subsequence obtained by decomposition. Eventually, the final prediction result could be obtained by adding forecast results for each subsequence. Compared with recent approaches, VMD-EWT-LSTM-based method can capture change features of cutterhead torque effectively, and has a higher multi time step prediction accuracy.

The contributions and innovations of this study are summarized as follows:

(1) We present a novel VMD-EWT-LSTM-based multi time step prediction method for shield machine cutterhead torque, in which combining VMD and EWT to decompose the original cutterhead torque sequence in advance can effectively extract the intrinsic characteristics of the original sequence, and LSTM with high prediction performance is employed to make multi-step prediction for each subsequence.

(2) The appropriate mode quantity K of VMD can be determined adaptively by setting appropriate threshold of the average decomposition relative error and K. A proper mode quantity could improve the cutterhead torque prediction accuracy and reduce unnecessary computation time.

(3) Compared with existing single time step cutterhead torque prediction methods, the presented VMD-EWT-LSTM-based multi time step prediction method has higher practical significance. Moreover, the presented VMD-EWT-LSTM-based method has higher prediction accuracy and stronger generalization ability than these methods, and overcomes the shortcomings of traditional methods that cannot effectively learn the long-term change law of cutterhead torque.

(4) Compared with existing cutterhead torque prediction approaches, the presented VMD-EWT-LSTM-based method only deals with one input parameter (i.e., cutterhead torque), which reduces the calculation time and improves the efficiency.

The rest of this study is arranged as follows. Section 2 introduces the researches related to cutterhead torque prediction. Section 3 introduces the materials and Section 4 focuses on describing the proposed VMD-EWT-LSTM-based multi-step cutterhead torque prediction method in this paper. In Section 5, the experimental design, comparative results and analysis are presented. Thereafter, in Section 6, some discussions on the method and results are provided. Eventually, Section 7 gives the conclusion of this study.

Section snippets

Related work

In this part, we introduce the current prediction methods of shield machine operation parameters and some researches on multi-step prediction of time series. At present, prediction approaches for shield machine operational parameters mainly include two types: physical model-based calculation approaches and data-driven prediction approaches. The former is mainly based on the physical principles of these operational parameters for predictive calculation. The latter is mainly to utilize historical

Materials

The dataset of this research was sampled from the Singapore T225 subway tunnel project (Fig. 1(a)), which was constructed by Shanghai Tunnel Company. This tunnel includes a total of 501 rings. In each ring, the shield tunneling machine excavated forward approximately 1 to 2 m. The overall length for this tunnel is approximately 750 m. Fig. 1(b) provides the shield machine utilized in this tunnel project. Its main machine equipment parameters are given in Table 1.

These data monitored by the

The proposed VMD-EWT-LSTM-based approach

Fig. 3 presents the overall flowchart of presented multi-step cutterhead torque prediction approach. It is mainly composed of three stages: signal decomposition, data normalization and multi-step torque prediction. The details of the proposed prediction method are described in the following Sections.

Performance evaluation index

We use three indicators to investigate prediction effect of the proposed multi-step cutterhead torque prediction method. That is, root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE). The calculation formulas for RMSE, MAE and MAPE is as follows: $RMSE = \sqrt{\frac{1}{n} \sum_{i = 1}^{n} {(x_{i} - {\hat{x}}_{i})}^{2}}$ $MAE = \frac{1}{n} \sum_{i = 1}^{n} | x_{i} - {\hat{x}}_{i} |$ $MAPE = \frac{1}{n} \sum_{i = 1}^{n} | \frac{x_{i} - {\hat{x}}_{i}}{x_{i}} | \times 100 %$ in which $x_{i}$ represents actual value of cutterhead torque at the time i, and ${\hat{x}}_{i}$ is the predicted value of cutterhead torque at the time i.

Discussion

Empirical mode decomposition (EMD), EWT, and VMD are commonly used signal decomposition methods. EMD has the problems such as mode aliasing, sensitivity to noise data, and boundary effects. The signal decomposition effect of EMD is worse than VMD and EWT. Compared with VMD, EWT can extract sub signals with a narrower bandwidth based on the Fourier spectrum of the signal. Hence, EWT requires more decomposition numbers than VMD to achieve more thorough decomposition to the signals with many

Conclusion

In this paper, we propose a novel VMD-EWT-LSTM-based multi-step prediction method for shield machine cutterhead torque. In the presented method, VMD is used to decompose the original cutterhead torque sequence, and the EWT is used to further decompose the residual sequence. The cutterhead torque sequence with complex change characteristics is decomposed into a series of subsequences with relatively simple change characteristics, which reduces the complexity of the original cutterhead torque

CRediT authorship contribution statement

Gang Shi: Conceptualization, Methodology, Software, Writing - original draft, Code review. Chengjin Qin: Conceptualization, Methodology, Software, Writing - original draft, Writing - review & editing, Resources, Data curation, Validation, Supervision. Jianfeng Tao: Conceptualization, Methodology, Software, Resources, Data curation, Visualization, Validation. Chengliang Liu: Conceptualization, Supervision, Project administration, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgments

This work was partially supported by the National Key R&D Program of China (Grant No. 2018YFB1702503), the State Key Laboratory of Mechanical System and Vibration (Grant No. MSVZD202103) and Shanghai Municipal Science and Technology Major Project (Grant No. 2021SHZDZX0102).

References (46)

YuH. et al.
A field parameters-based method for real-time wear estimation of disc cutter on TBM cutterhead
Autom. Constr.
(2021)
YuH. et al.
Rock mass type prediction for tunnel boring machine using a novel semi-supervised method
Measurement
(2021)
ZhangQ. et al.
Modeling of the thrust and torque acting on shield machines during tunneling
Autom. Constr.
(2014)
GaoX. et al.
Recurrent neural networks for real-time prediction of TBM operating parameters
Autom. Constr.
(2019)
AtesU. et al.
Estimating torque thrust and other design parameters of different type TBMs with some criticism to TBMs used in turkish tunneling projects
Tunn. Undergr. Space Technol.
(2014)
WangL. et al.
Modeling and analysis of thrust force for EPB shield tunneling machine
Autom. Constr.
(2012)
ZhouX. et al.
Estimation of the cutterhead torque for earth pressure balance TBM under mixed-face conditions
Tunn. Undergr. Space Technol.
(2018)
ZhangK. et al.
Dynamic characteristic analysis of TBM tunnelling in mixed-face conditions
Autom. Constr.
(2010)
HanM.D. et al.
Dynamic numerical simulation of cutterhead loads in TBM tunnelling
Unn. Undergr. Space Technol.
(2017)
WuL. et al.
Discrete element model for performance analysis of cutterhead excavation system of EPB machine
Unn. Undergr. Space Technol.
(2013)

SunW. et al.

Dynamic load prediction of tunnel boring machine (TBM) based on heterogeneous in-situ data

Autom. Constr.

(2018)

SongX. et al.

A new fuzzy c-means clustering-based time series segmentation approach and its application on tunnel boring machine analysis

Mech. Syst. Signal Proc.

(2019)

ZhangP. et al.

A critical evaluation of machine learning and deep learning in shield-ground interaction prediction

Tunn. Undergr. Space Technol.

(2020)

ZhouC. et al.

Dynamic prediction for attitude and position in shield tunneling: A deep learning method

Autom. Constr.

(2019)

QinC. et al.

Precise cutterhead torque prediction for shield tunneling machines using a novel hybrid deep neural network

Mech. Syst. Signal Proc.

(2021)

KaoI. et al.

Exploring a long short-term memory based encoder-decoder framework for multi-step-ahead flood forecasting

J. Hydrol.

(2020)

LiuH. et al.

Wind speed forecasting method based on deep learning strategy using empirical wavelet transform, long short term memory neural network and Elman neural network

Energy Convers Manage.

(2018)

UpadhyayA. et al.

Instantaneous voiced/non-voiced detection in speech signals based on variational mode decomposition

J. Franklin Inst. B

(2015)

WangY. et al.

Research on variational mode decomposition and its application in detecting rubimpact fault of the rotor system

Mech. Syst. Signal Process.

(2015)

LiH. et al.

An optimized VMD method and its applications in bearing fault diagnosis

Measurement

(2020)

XuZ. et al.

Fault diagnosis of rolling bearing of wind turbines based on the variational mode decomposition and deep convolutional neural networks

Appl. Soft Comput.

(2020)

LahmiriS.

A variational mode decomposition approach for analysis and forecasting of economic and financial time series

Expert Syst. Appl.

(2016)

LiuY. et al.

Non-ferrous metals price forecasting based on variational mode decomposition and LSTM network

Knowl.-Based Syst.

(2020)

Cited by (71)

A novel decomposition and hybrid transfer learning-based method for multi-step cutterhead torque prediction of shield machine
2024, Mechanical Systems and Signal Processing
As a core operating parameter of shield machine, the high-precision multi-step prediction of cutterhead torque can effectively help the driver adjust the operating parameters in advance to ensure the safe operation of shield machine. Due to data distribution discrepancy, it is challenging to accurately predict the cutterhead torque under different geological conditions. Currently, the existing methods cannot precisely predict cutterhead torque in cross domain. In order to accurately predict the cutterhead torque under different geological environments, we propose a novel decomposition and hybrid transfer learning-based method for multi-step cutterhead torque transfer prediction. First, original cutterhead torque signal is decomposed into some sub-signals by variational mode decomposition (VMD), which can effectively reduce the complexity. Then hybrid transfer learning-based method (HTLM) is established to predict each sub-signal and add the prediction results to realize multi-step cutterhead torque prediction in cross domain. HTLM has three network branches. The first one is that multi-layer GRU network is used to extract time-varying characteristics of source domain data. The second is that the automatic encoder is used to extract the deep common features between source domain and target domain. The third is the adversarial network and the maximum mean discrepancy (MMD) metric for domain adaptation. Then, the extracted features are integrated to predict the cutterhead torque in cross domain. To verify the proposed VMD-HTLM, several transfer prediction experiments were carried out. From the first step to fifth step transfer prediction, the proposed method achieved higher prediction accuracy than existing methods. Especially in the fifth step, the accuracy of proposed method is on average 7.09 % higher than existing methods, the RMSE and the MAE decreased on average 40.00 % and 39.92 %, respectively. Furthermore, the average accuracy of proposed method is 93.79 %, 93.58 %, 93.55 %, 92.25 % and 91.79 % from first step to fifth step, respectively. Hence, the proposed VMD-HTLM can accurately predict the cutterhead torque under different geological conditions.
A hybrid model with combined feature selection based on optimized VMD and improved multi-objective coati optimization algorithm for short-term wind power prediction
2024, Energy
With the continuous global increase in installed wind power capacity and subsequent surge in power generation, the contradiction between the safe operation of the grid and the efficient consumption of new energy after large-scale grid connection has become increasingly prominent. This paper presents a hybrid model prediction method to further improve the stability and accuracy of wind power prediction. Firstly, variational modal decomposition (VMD) optimized by the coati optimization algorithm (COA) is employed to decompose original wind power, mitigating the non-stationary characteristics of the power sequence. Subsequently, permutation entropy (PE) is used to recombine the decomposed components, and the combined feature selection method is achieved by integrating the Spearman correlation coefficient (SCC) and the autocorrelation function (ACF). Then, the multivariate combined model is constructed, and the improved multi-objective coati optimization algorithm (IMOCOA) determines the weight coefficients of each model to enhance the performance of the hybrid model. Finally, research and analysis are conducted from multiple scenarios and time scales using historical operating data from a wind farm in Xinjiang. The experimental results show that the proposed prediction model effectively improves the accuracy and stability of the wind power prediction compared with other popular prediction models.
Prediction of cutterhead torque change trend of shield machine based on partial state visible HMM and LSTM
2024, Journal of the Franklin Institute
Cutterhead torque is a key parameter to determine the normal operation of shield machine. Accurately predicting the change trend of cutterhead torque can provide decision support for shield operators to control operating parameters. Therefore, a prediction method of cutterhead torque change trend based on partial state visible Hidden Markov Model (HMM) and Long Short-Term Memory (LSTM) network is proposed in this paper. First, the Spearman correlation coefficient is introduced to analyze the correlation of the tunneling parameters, and the historical tunneling data of the cutter torque, pitch angle and related parameters are obtained to establish a database. Secondly, the pitch angle prediction model is established based on LSTM network, and the predicted pitch angle is used as the observation sequence, and the implicit relationship between cutterhead torque and pitch angle is mined through HMM to obtain the probability matrix. Finally, the cutterhead torque prediction model is established, and the pitch angle sequence is used as input to output the corresponding cutterhead torque change trend. The results show that the method can accurately predict the dynamic change trend of the cutterhead torque in the case of less known parameter states, and the effectiveness of the method are verified in this paper.
A novel empirical reconstruction Gauss decomposition method and its application in gear fault diagnosis
2024, Mechanical Systems and Signal Processing
This paper proposes a novel signal decomposition method, empirical reconstruction Gaussian decomposition (ERGD). ERGD comprises two key steps: spectrum segmentation based on the unimodal symmetry hypothesis and reconstruction decomposition strategy. ERGD puts forward the unimodal symmetry hypothesis along with the corresponding unimodal symmetry estimation method. Building upon this, ERGD presents a novel spectrum segmentation approach that adaptively divides the spectrum into energy-concentrated and unimodal symmetry segments, thereby circumventing issues of over-segmentation. The reconstruction decomposition strategy defines a series of $α$ factor generalized Gaussian filters and then reconstructs them into ideal filter banks in order to accomplish signal decomposition. The ideal filter banks can effectively eliminate noise in the segment. It is worth mentioning that ERGD has the ability to adaptively adjust the overlap and attenuation properties of the ideal filter, which can effectively suppress noise interference. As a new signal decomposition method, ERGD is capable of decomposing signals into a series of Gaussian intrinsic mode functions (GIMFs) possessing orthogonality. Applying ERGD to gear fault diagnosis, both simulation and experimental signal analysis results demonstrate the excellent performance of the proposed method in signal decomposition. It significantly contributes to the attainment of precise gear fault diagnosis and holds practical value.
A causal-temporal graphic convolutional network (CT-GCN) approach for TBM load prediction in tunnel excavation
2024, Expert Systems with Applications
This research proposes a novel deep learning approach named causal-temporal graphic convolutional network (CT-GCN) which aims to provide accurate predictions on tunnel boring machine’s (TBM) load parameters. The causal associations among the key TBM operational parameters are detected and quantified with causal effects through the Peter and Clark momentary conditional independence plus (PCMCI+) method. The discovered causality can be further integrated with a deep learning model that consists of graph convolutional network (GCN) and long short-term memory (LSTM) layers for accurate prediction of TBM’s torque and thrust. Data collected from a realistic tunnel project in Singapore is utilized to demonstrate the effectiveness of the proposed approach. The load parameters are predicted with their historical values and another 7 key operational parameters. The results indicate that (1) The proposed CT-GCN approach can achieve a low mean absolute error (MAE) and root mean squared error (RMSE) of $40.90 k N ∙ m$ and $64.53 k N ∙ m$ for thrust force ( $y_{1}$ ) and that of $635.76 k N$ and $1168.59 k N$ for cutterhead torque ( $y_{2}$ ), respectively. (2) The proposed CT-GCN method achieves 12.86 % and 24.38 % average improvements in the coefficient of determination ( $R^{2}$ ) for $y_{1}$ and $y_{2}$ , respectively when compared with the long-short term memory (LSTM) method and that of 5.62 % and 8.60 % improvements when compared with the GCN-LSTM method. (3) Compared with correlation-based GCN models, the proposed approach exerts an average improvement of 43.10 %, 41.97 %, and 22.72 % in terms of MAE, RMSE, and $R^{2}$ for torque and much better performance for thrust estimation. This study contributes to improving the understanding of TBM operation by quantifying the causality among the key operational parameters. It also contributes to developing a novel deep-learning method that estimates TBM’s load parameters with high accuracy.
Multisource information fusion for real-time optimization of shield construction parameters
2024, Knowledge-Based Systems
This paper introduces a hybrid intelligent framework that combines Bayesian optimization (BO), a random forest (RF) model, and the nondominated sorting genetic algorithm-III (NSGA-III) for the optimization and control of tunnel shield construction parameters. The BO-RF method establishes a nonlinear mapping function between the input variables and three targets, surface settlement, cutter wear, and advance speed, serving as the fitness function for NSGA-III. Model interpretability analysis is conducted using Shapley Additive ExPlanations (SHAP). A multiobjective intelligent optimization model is formulated with NSGA-III, targeting surface settlement, cutter wear, and advance speed. A case study validates the applicability and effectiveness of this approach, leading to the following conclusions: (1) The BO-RF algorithm yields highly accurate prediction results, with R² values ranging from 0.930 to 0.938, RMSE ranging from 0.138 to 0.172, and MAE ranging from 0.112 to 0.138 for the three targets. (2) The optimization results for surface settlement, cutter wear, and advance speed are outstanding, with an average improvement of 12.56 %. The simultaneous adjustment of the three shield construction parameters leads to the best optimization results, with an average improvement of 19.67 %. (3) The energy consumption of the shield drive system decreases by an average of 10.70 %, and the optimization improvement for the first three objectives decreases by an average of 1.82 %, 1.46 %, and 2.23 %, respectively. By introducing the integrated BO-RF-NSGA-III algorithm, this study contributes to the field of tunnel engineering optimization management.

View all citing articles on Scopus

View full text

A VMD-EWT-LSTM-based multi-step prediction approach for shield tunneling machine cutterhead torque

Highlights

Abstract

Introduction

Section snippets

Related work

Materials

The proposed VMD-EWT-LSTM-based approach

Performance evaluation index

Discussion

Conclusion

CRediT authorship contribution statement

Declaration of Competing Interest

Acknowledgments

Autom. Constr.

Measurement

Autom. Constr.

Autom. Constr.

Tunn. Undergr. Space Technol.

Autom. Constr.

Tunn. Undergr. Space Technol.

Autom. Constr.

Unn. Undergr. Space Technol.

Unn. Undergr. Space Technol.

Autom. Constr.

Mech. Syst. Signal Proc.

Tunn. Undergr. Space Technol.

Autom. Constr.

Mech. Syst. Signal Proc.

J. Hydrol.

Energy Convers Manage.

J. Franklin Inst. B

Mech. Syst. Signal Process.

Measurement

Appl. Soft Comput.

Expert Syst. Appl.

Knowl.-Based Syst.