Publications

Detection of Oscillations in Process Control Loops from Visual Image Space Using Deep Convolutional Networks

Published in IEEE/CAA Journal of Automatica Sinica, 2024

Oscillation detection has been a hot research topic in industries due to the high incidence of oscillation loops and their negative impact on plant profitability. Although numerous automatic detection techniques have been proposed, most of them can only address part of the practical difficulties. An oscillation is heuristically defined as a visually apparent periodic variation. However, manual visual inspection is labor-intensive and prone to missed detection. Convolutional neural networks (CNNs), inspired by animal visual systems, have been raised with powerful feature extraction capabilities. In this work, an exploration of the typical CNN models for visual oscillation detection is performed. Specifically, we tested MobileNet-V1, ShuffleNet-V2, EfficientNet-B0, and GhostNet models, and found that such a visual framework is well-suited for oscillation detection. The feasibility and validity of this framework are verified utilizing extensive numerical and industrial cases. Compared with state-of-the-art oscillation detectors, the suggested framework is more straightforward and more robust to noise and mean-nonstationarity. In addition, this framework generalizes well and is capable of handling features that are not present in the training data, such as multiple oscillations and outliers.

Recommended citation: Tao Wang, Qiming Chen, Xun Lang, Lei Xie, Peng Li, Hongye Su. IEEE/CAA Journal of Automatica Sinica. (2024). https://ieeexplore.ieee.org/abstract/document/10482814

Federated Domain Separation for Distributed Forecasting of Non-IID Household Loads

Published in IEEE Transactions on Smart Grid, 2024

Household load forecasting is increasingly essential since it enables various demand-side management applications. The federated learning approach is becoming popular for its advantages in fully using different households’ load data with privacy preservation. However, due to the non-independent and identically distributed (non-IID) characteristic of each household’s local data, the knowledge acquired by local training may have a strong bias. It can introduce contamination and make the global model vulnerable if locally trained models are simply aggregated as traditional FL methods do. To this end, we develop a novel framework that integrates federated domain separation to alleviate the negative effects caused by non-IID data. Specifically, we divide the acquired knowledge into the useful part and potentially contaminating part. By acquiring the former and removing the latter through a well-designed algorithm, a more anti-contamination and more personalized FL model can be expected. Compared to current post-processing personalization methods, the proposed framework can avoid global knowledge forgetting, thus achieving more comprehensive knowledge utilization to give more accurate results. Extensive comparison experiments with benchmarking methods are conducted on a publicly available dataset to validate the superiority of the proposed framework, while a variety of ablation experiments prove the effectiveness of all inner components.

Recommended citation: Nan Lu, Shu Liu, Qingsong Wen, Qiming Chen, Liang Sun, Yi Wang. IEEE Transactions on Smart Grid. (2024). Federated Domain Separation for Distributed Forecasting of Non-IID Household Loads

Adaptive multi-scale TF-net for high-resolution time–frequency representations

Published in Signal Processing, 2024

A novel adaptive multi-scale time–frequency network (AMTFN) is proposed to provide high-resolution time–frequency representations for nonstationary signals. AMTFN is an end-to-end deep network, which firstly adaptively learns the comprehensive basis functions to produce time–frequency (TF) feature maps through multi-scale 1D convolutional kernels. Then, the channel attention mechanism is embedded into AMTFN to rescale the TF feature maps selectively. Thus, the subsequent residual encoder–decoder block’s energy concentration performance is greatly improved with these rescaled TF feature maps. Besides, this paper designs a new training strategy to elegantly enable the model to pay more attention to the intersections of instantaneous frequency trajectories. In the end, a series of simulations as well as real-world cases, are studied to demonstrate the effectiveness and advantages of the proposed method.

Recommended citation: Tao Chen, Qiming Chen*, Qian Zheng, Zhishan Li, Ziyi Zhang, Lei Xie*, Hongye Su. Control Engineering Practice. (2024). https://www.sciencedirect.com/science/article/pii/S0165168423003213

Detection and time–frequency analysis of multiple plant-wide oscillations using adaptive multivariate intrinsic chirp component decomposition

Published in Control Engineering Practice, 2023

Analyzing plant-wide oscillations is a challenging task owing to the presence of noise, nonstationarity, and multiple modes in a process control system. Multivariate intrinsic chirp component decomposition (MICCD) is a novel powerful tool for multivariate signal processing. Nevertheless, MICCD requires users to provide component number in advance, which restricts its adaptability. This study proposes an adaptive MICCD (AMICCD) that can adaptively determine the component number by utilizing the permutation entropy of instantaneous frequency. An AMICCD-based time–frequency analysis framework is presented to detect and characterize the multiple plant-wide oscillations. Compared to the latest methods, such as multivariate empirical mode decomposition and multivariate intrinsic time-scale decomposition, the proposed method can process not only single/multiple plant-wide oscillations, but also time-invariant/time-varying plant-wide oscillations. In particular, the proposed method can directly provide the time–frequency curves of multiple plant-wide oscillations, which have not been achieved by the state-of-the-art techniques. Finally, the effectiveness and advantages of the proposed approach are demonstrated on a wide variety of simulations and industrial cases.

Recommended citation: Qiming Chen, Qingsong Wen, Xialai Wu, Xun Lang, Yao Shi, Lei Xie, Hongye Su. Control Engineering Practice. (2023). [https://www.sciencedirect.com/science/article/abs/pii/S0967066121002094](https://www.sciencedirect.com/science/article/abs/pii/S0967066123002848)

Data-driven identification and fast model predictive control of the ORC waste heat recovery system by using Koopman operator

Published in Control Engineering Practice, 2023

Organic Rankine Cycle (ORC) has received its wide application in low-grade waste heat recovery (WHR) technology for its significant performance and easy access to its components. Given the highly fluctuating nature of the waste heat source, Model Predictive Control (MPC) is usually utilized to realize the reasonable adjustment of ORC based WHR systems and has performed satisfactorily. In order to apply MPC, a relatively precise model should be built up for the ORC system which ensures acceptable control performance. However, popular modeling methods based on the mechanism of the ORC establish the high-dimensional nonlinear model and suffer from computational costs when utilizing complicated nonlinear MPC. Even if linear MPC is employed for the purpose of reducing the calculation amount, the model obtained by linearization near the operating points usually makes it valid locally, thus bringing about suboptimal or unstable control performance. To address this problem, the Koopman operator is introduced for the data-driven identification and MPC of the ORC system. Koopman identification constructs a linear model in the lifted space in which the ORC system possesses nearly global linear evolution and enables the prediction of the nonlinear ORC dynamics from measurement data. In view of the possible online calculation burden increment caused by the rise of variable dimension through lifting action, the fast Koopman MPC (FKMPC) algorithm is thus proposed based on the invariance of the Hessian matrix of the optimization problem to shorten the computing time. Simulations on setpoint tracking and disturbance rejection are performed to verify the established model accuracy and the control effectiveness of the proposed strategy in comparison with other approaches.

Recommended citation: Yao Shi, Xiaorong Hu, Zhiming Zhang, Qiming Chen, Lei Xie, Hongye Su. Control Engineering Practice. (2023). https://www.sciencedirect.com/science/article/abs/pii/S0967066123002484

Energy forecasting with robust, flexible, and explainable machine learning algorithms

Published in AI Magazine, 2023

Energy forecasting is crucial in scheduling and planning future electric load, so as to improve the reliability and safeness of the power grid. Despite recent developments of forecasting algorithms in the machine learning community, there is a lack of general and advanced algorithms specifically considering requirements from the power industry perspective. In this paper, we present eForecaster, a unified AI platform including robust, flexible, and explainable machine learning algorithms for diversified energy forecasting applications. Since October 2021, multiple commercial bus load, system load, and renewable energy forecasting systems built upon eForecaster have been deployed in seven provinces of China. The deployed systems consistently reduce the average Mean Absolute Error (MAE) by 39.8% to 77.0%, with reduced manual work and explainable guidance. In particular, eForecaster also integrates multiple interpretation methods to uncover the working mechanism of the predictive models, which significantly improves forecasts adoption and user satisfaction.

Recommended citation: Zhaoyang Zhu, Weiqi Chen, Rui Xia, Tian Zhou, Peisong Niu, Bingqing Peng, Wenwei Wang, Hengbo Liu, Ziqing Ma, Xinyue Gu, Jin Wang, Qiming Chen, Linxiao Yang, Qingsong Wen, Liang Sun. AI Magazine. (2023). [https://www.sciencedirect.com/science/article/abs/pii/S0967066121002094](https://onlinelibrary.wiley.com/doi/full/10.1002/aaai.12130)

A Modified Multivariate Variational Mode Decomposition for Multi-channel Signal Processing

Published in International Conference on Communication, Image and Signal Processing (CCISP) 2023, 2023

The multivariate variational mode decomposition (MVMD) is an optimization-based method that allows simultaneous processing of non-stationary multi-channel signals. It has recently garnered considerable attention due to its excellent signal separation capability and robustness against the sampling frequency. Nevertheless, MVMD also encounters certain challenges. Specifically, in cases where the signal is significantly disturbed by noise, MVMD struggles to accurately decompose the correct sub-signals. Moreover, the computational time required for MVMD increases dramatically with the number of channels, thereby making it challenging to solve multivariate signals with high-density channel. An effective solution to these issues is the implementation of the proposed modified MVMD (MMVMD) technique. This method aims to convert the task of finding multiple multivariate oscillations corresponding to MVMD into finding multiple univariate oscillations in one-dimensional space. The satisfactory noise robustness and faster decomposition rate of MMVMD, especially for signals with a large number of channels, are demonstrated by its application on multivariate synthetic signals. We finally provide further convincing validation of the effectiveness of the algorithm through the analysis of a 4-channel EEG signal.

Recommended citation: Jiayi Wang, Qiming Chen, Xun Lang, Songhua Liu, Yanjiang Liu, Hongsheng Su. International Conference on Communication, Image and Signal Processing (CCISP). (2023). https://ieeexplore.ieee.org/abstract/document/10355770/

A Hybrid Deep Neural Network for Nonlinear Causality Analysis in Complex Industrial Control System

Published in IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2023, 2023

It is important to efficiently and accurately locate the fault root cause to maintain the control performance, when the industrial control system fails. However, this task is very challenging because the industrial control system is large in scale and complex in connection. This paper proposes a novel neural causality analysis network with directed acyclic graph to locate the root cause for complex industrial systems. This network fits the temporal nonlinearity and intervariable non-linearity to mine the causal graph. The proposed method is data-driven, which acts without process knowledge. Compared with the state-of-the-art, this method can effectively output accurate root cause from nonlinear and highly coupled data. The effectiveness and advantages are demonstrated by industrial cases.

Recommended citation: Tian Feng, Qiming Chen*, Yao Shi, Xun Lang, Lei Xie*, Hongye Su. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). (2023). https://ieeexplore.ieee.org/abstract/document/10095701

一元及多元信号分解发展历程与展望

Published in 自动化学报 ACTA AUTOMATICA SINICA, 2023

现实世界中所获得的信号大部分都是非平稳和非线性的, 将此类复杂信号分解为多个简单的子信号是重要的信号处理方法. 自1998年Huang提出希尔伯特–黄变换(Hilbert-Huang transform, HHT), 历经20余年的发展, 信号分解已经成为信号处理领域相对独立又具有创新性的重要内容. 特别是近10年来, 多元/多变量/多通道信号分解理论方法方兴未艾, 在诸多领域得到了成功应用. 但目前尚未见到相关综述报道, 因此为了填补这个空缺, 从单变量和多变量两个方面系统综述了国内外学者对主要信号分解方法的研究现状, 对这些方法的时频表达性能进行了分析和比较, 指出了这些分解方法的优势和存在的问题. 最后对信号分解研究进行了总结展望.

Recommended citation: 陈启明，文青松，郎恂，谢磊，苏宏业. 自动化学报 ACTA AUTOMATICA SINICA. (2023). http://www.aas.net.cn/cn/article/doi/10.16383/j.aas.c220632

中值互补集合经验模态分解

Published in 自动化学报 ACTA AUTOMATICA SINICA, 2023

针对经验模态分解 (Empirical mode decomposition, EMD) 系列方法存在的模态分裂 (Mode Splitting, MS) 问题, 本文提出中值互补集合经验模态分解 (Median complementary ensemble EMD, MCEEMD) 算法. 通过概率模型量化互补集合经验模态分解 (Complementary ensemble EMD, CEEMD) 的 MS 问题, 证明了使用中值算子替代算术平均算子对抑制 MS 的有效性. MCEEMD 算法首先添加 N 对互补的白噪声至原信号中, 并经过 EMD 分解得到 2N 组固有模态函数 (Intrinsic mode functions, IMFs), 然后分别对其中互补相关的 IMFs 两两取平均得到 N 组 IMFs, 最后使用中值算子处理上述 N 组 IMFs 得到输出结果. 对仿真信号与实测信号的分析结果表明, 本文提出的 MCEEMD 方法不仅有效抑制了 CEEMD 的 MS 问题, 而且避免了单一使用中值算子的两个缺点, 即: 1) 分解完备性差和 2) IMFs 中存在毛刺现象

Recommended citation: 刘淞华，何冰冰，郎恂， 陈启明，张榆锋，苏宏业. 自动化学报 ACTA AUTOMATICA SINICA. (2021). http://www.aas.net.cn/article/doi/10.16383/j.aas.c201031

On demodulation of time-varying oscillations in process plant

Published in Journal of Process Control, 2022

Due to the presence of process fluctuation, transient impacts and noise, most of the oscillations in industrial control-loops evolve with time-varying properties. If significant variation is observed under a limited monitoring window, the existing regularity-based oscillation detectors may be invalid. To tackle this problem, this paper proposes to describe the oscillations as a kind of amplitude modulated and frequency modulated (AM–FM) signals, and designs corresponding scheme to extract and detect them. More specifically, our proposed method is featured by following steps: (i) Decompose the process output into several AM–FM modes using ensemble empirical mode decomposition; (ii) Demodulate each mode with Teager energy operator for extracting the corresponding instantaneous amplitude and frequency functions; (iii) Identify the oscillatory variables by developing a new statistical measure based on the deviation of the frequency variation. The applicability of the proposed method is validated through a set of simulations and representative chemical applications. Comparative studies also demonstrate that it outperforms existing time–frequency tools and typical regularity-based detectors.

Recommended citation: Xun Lang, Qiming Chen, Shan Lu, Alexander Horch, Yufeng Zhang. Journal of Process Control. (2022). https://www.sciencedirect.com/science/article/pii/S0959152422001901

Controlling Pressure of Gas Pipeline Network Based on Mixed Proximal Policy Optimization

Published in Chinese Automation Congress (CAC) 2022, 2022

The gas pipeline network plays a significant role in transportation due to its low cost and safe operation. However, the traditional gas pipeline network control scheme has been criticized for its inefficiency. Reinforcement learning (RL) provides an emerging model-free alternative, but its poor generalization performance becomes a practical obstacle in diverse equipment scenarios. In this paper, a Mixed Proximal Policy Optimization (PPO) approach, Mixed-PPO, is proposed to simultaneously control the continuous and discrete equipment. Compared with the original PPO, the proposed method exhibits better performance in the following aspects: (i) the ability to realize the control of both continuous and discrete equipment, making gas pipeline network pressure more stable; (ii) quicker convergence while obtaining higher profits. The effectiveness of our method is illustrated by a case study. The results show that Mixed-PPO outperforms the original PPO in terms of both profit increase and pressure fluctuation decrease, by 76.86% and 38.65%, respectively.

Recommended citation: Haiying Chang, Qiming Chen, Runze Lin, Yao Shi, Lei Xie, Hongye Su. Chinese Automation Congress (CAC). (2022). https://ieeexplore.ieee.org/abstract/document/10055122

Multivariate time-varying complex signal processing framework and its application in rotating machinery rotor-bearing system

Published in Measurement Science and Technology, 2022

Affected by multi-field coupling factors, the vibration response of rotating machinery similar to the hydro-generator unit often exhibits strong time-varying frequency components, which makes rotor fault detection more challenging. The fusion analysis of the vibration signals of multiple bearing sections of the rotor has been proved to be a very effective method for rotor vibration fault diagnosis. However, how to more accurately and synchronously extract the instantaneous features of rotor non-stationary vibration signals associated with multiple sections has been unresolved. To this end, a framework for multivariate time-varying complex signal decomposition of the rotor-bearing system (RBS) is proposed, namely multivariate complex nonlinear chirp mode decomposition. First, the decomposition of multivariate time-varying complex signals is realized by two-stage processing. Second, instantaneous orbit features (IOFs) are obtained through the proposed framework. Finally, a three-dimensional instantaneous orbit map reflecting the time-varying process is constructed through the IOFs. The framework not only realizes the decomposition of the multi-channel time-varying complex signals of the rotor but also simultaneously extracts the instantaneous features of the multi-channel signals. In addition, it also realizes the description of the instantaneous vibration state of the RBS in the non-stationary process (such as startup and shutdown). Simulation experiments show that the framework is superior to other multi-channel signal processing methods in processing time-varying complex signals. The results based on field-measured signals show that the framework can guide the real-time analysis of the signals generated by rotating machinery, which improves the intuition of condition monitoring.

Recommended citation: Jie Huang, Xiaolong Cui, Chaoshun Li, Zhihuai Xiao, Qiming Chen. Measurement Science and Technology. (2022). https://iopscience.iop.org/article/10.1088/1361-6501/ac919b/meta

Quality-Relevant Process Monitoring with Concurrent Locality-Preserving Dynamic Latent Variable Method

Published in ACS omega, 2022

A concurrent locality-preserving dynamic latent variable (CLDLV) method is proposed to extract the correlation between process variables and quality variables for quality-related dynamic process monitoring. Given that dynamic process data can easily be contaminated by noise and outliers and conventional dynamic latent variable models lack robustness, a low-rank autoregressive model is developed to deal with autocorrelation and cross-correlation properties among the data. Then neighborhood structure information is integrated into the partial least squares model, which can better reveal the essential structure of the data. The final concurrent projection of the latent structures is employed to monitor output-related faults and input-related process faults that affect quality. The Tennessee Eastman process and hot strip mill process are used to demonstrate the effectiveness of CLDLV-based detection and diagnostic methods.

Recommended citation: Qi Zhang, Shan Lu, Lei Xie, Qiming Chen, Hongye Su. ACS omega. (2022). https://pubs.acs.org/doi/full/10.1021/acsomega.2c02118

SVD-Based Robust Distributed MPC for Tracking Systems Coupled in Dynamics With Global Constraints

Published in IEEE Transactions on Cybernetics, 2022

This article presents a novel singular value decomposition (SVD)-based robust distributed model predictive control (SVD-RDMPC) strategy for linear systems with additive uncertainties. The system is globally constrained and consists of multiple interrelated subsystems with bounded disturbances, each of whom has local constraints on states and inputs. First, we integrate the steady-state target optimizer into the MPC problem through the offset cost function to formulate a modified single optimization problem for tracking changing targets from real-time optimization. Then, the concept of constraint tightening is utilized to enhance the robustness and ensure robust constraint satisfaction in the presence of interferences. On this basis, the SVD method is introduced to decompose the new optimization problem into several independent subsystems on the orthogonal projection space, and a distributed dual gradient algorithm with convergence proved is implemented to obtain the control of each nominal subsystem. The recursive feasibility is then ensured and the tracking ability of the strategy is analyzed. It is verified that for a target, the system can be steered to a neighborhood of the closest possible steady setpoint. At last, the effectiveness of the raised SVD-RDMPC strategy is established in two simulations on building temperature control and load frequency control.

Recommended citation: Yao Shi, Zhiming Zhang, Xiaorong Hu, Pei Sun, Lei Xie, Qiming Chen, Hongye Su. IEEE Transactions on Cybernetics. (2022). https://ieeexplore.ieee.org/abstract/document/9781343

Detecting multiple plant-wide oscillations in process control systems based on multivariate intrinsic chirp component decomposition

Published in CAA Symposium on Fault Detection, Supervision, and Safety for Technical Processes (SAFEPROCESS), 2021

An oscillation originating at one place tends to propagate to other parts of the plant due to underlying interactions and connected process flows, thus causing plant-wide oscillations. The plant-wide oscillations can result in unstable product quality, equipment wear, security degradation. Therefore, it is of importance to detect the plant-wide oscillations to maintain the control system performance. This paper proposes an MICCD-based (multivariate intrinsic chirp component decomposition) detector, which is able to detect and analyze the plant-wide oscillations. Firstly, we present the MICCD algorithm, which is an extension of intrinsic chirp component decomposition (ICCD). Then, the plant-wide oscillations is decomposed into a series of multivariate nonlinear chirp modes by MICCD. Following, the normalized correlation index, regualrity index, and sparseness index are used to identify the oscillation modes. Compared with the existing methods, the proposed method can process both time-invariant and time-varying multiple plant-wide oscillations and provide corresponding time-frequency information. The effectiveness and superiority of the MICCD-based detector are demonstrated via simulations as well as industrial cases.

Recommended citation: Qiming Chen, Xun Lang, Yi Pan, Yao Shi, Lei Xie, Hongye Su. SAFEPROCESS. (2021). https://ieeexplore.ieee.org/abstract/document/9693592

Two-layer structure strategy for large-scale systems integrating online adaptive constraints adjustment method and cooperative distributed DMC algorithm

Published in Control Engineering Practice, 2021

In this paper, a distributed two-layer structure strategy for large-scale systems is described. In the upper level of the structure, usually known as the Steady State Target Optimizer (SSTO) layer, we fully consider the possible constraints in the industrial process as well as their priority order and put forward an online adaptive constraints adjustment (OACA) scheme to ensure the stable operation of the production device. Based on Pareto optimal, a new cooperative distributed dynamic matrix control (CDDMC) algorithm is proposed for the lower MPC layer. The algorithm makes use of the finite step response model and Dynamic Matrix Control (DMC) method which are commonly applied in the process industry, decomposes the large-scale system into multiple interconnected subsystems for a significantly reduced computational burden, and works over a cooperative way based on Jacobi-type iteration to achieve the global optimal solution. Then the convergence of the CDDMC algorithm is investigated and the offset-free control of the strategy is discussed. Finally, the two-layer structure distributed strategy is applied to two examples to analyze its applicability and effectiveness in comparison with the centralized one.

Recommended citation: Yao Shi, Zhiming Zhang, Pei Sun, Lei Xie, Qiming Chen, Hongye Su, Xiaoqiang Chen. Control Engineering Practice. (2021). https://www.sciencedirect.com/science/article/abs/pii/S0967066121002094

Self-tuning variational mode decomposition

Published in Journal of the Franklin Institute, 2021

Variational mode decomposition (VMD) has attracted a lot of attention recently owing to its robustness to sampling frequency and its high-frequency resolution. However, its performance highly depends on two key preset parameters (the mode number and the penalty parameter ), both of which tightly limit its adaptability and applications. In this study, a self-tuning VMD (SVMD) is proposed to tackle this problem. Within the proposed method, and update themselves respectively and adaptively via the energy ratio and orthogonality between modes in the frequency domain. The proposed SVMD is similar to a matching pursuit method and it shows a VMD-like equivalent filter bank structure but with much less mode-mixing probability. We show that SVMD is more robust to both changes in and noise level than the original VMD; also, it has better convergence and reduces mode-mixing and end-effect. The experiments on SVMD indicate that SVMD outmatches several classic signal decomposition algorithms. In the end, real-world applications in three fields, namely, length of day variation analysis in geophysics, climate cycle study in meteorology, and oscillation detection in process control, are provided to demonstrate the effectiveness and advantages of the proposed SVMD.

Recommended citation: Qiming Chen, Junghui Chen, Xun Lang, Lei Xie, Naveed ur Rehman, Hongye Su. Journal of the Franklin Institute. (2021). https://www.sciencedirect.com/science/article/abs/pii/S0016003221004233

Adaptive clutter filtering for ultrafast Doppler imaging of blood flow using fast multivariate empirical mode decomposition

Published in IEEE International Ultrasonics Symposium (IUS), 2021

The rapid development of ultrafast ultrasound imaging based on the unfocused transmission of plane-wave has led to the widespread attention of clutter filtering technology, since the discrimination between tissue and blood motion is critical for non-contrast ultrafast Doppler imaging of blood flow. The increasingly used clutter filtering method, i.e., the singular value decomposition (SVD) is essentially a black-box technique, and the direct rejection of any singular vector corresponding to the first singular values will result in the potential loss of blood flow or residue of tissue activities. To cater for a better filtering performance, an adaptive clutter rejection method based on the fast multivariate empirical mode decomposition (FMEMD) is proposed, and compared to the high-pass filter (HPF) and SVD in the task of performing blood flow imaging and velocity profile estimation on signals collected from carotid arteries of 10 healthy human subjects. The results demonstrate the superiority of the proposed method over the state-of-the-art techniques, especially for discriminating between blood flow and tissue signals near the vessel walls.

Recommended citation: Xun Lang, Bingbing He, Yufeng Zhang, Qiming Chen, Lei Xie. IEEE International Ultrasonics Symposium (IUS). (2021). https://ieeexplore.ieee.org/abstract/document/9593919

Median Complementary Ensemble Empirical Mode Decomposition and its application to time-frequency analysis of industrial oscillations

Published in Chinese Control Conference (CCC), 2021

Median ensemble empirical mode decomposition (MEEMD) represents a remarkable improvement based on the ensemble empirical mode decomposition (EEMD) method for alleviating the mode splitting and mode mixing problem. However, the single use of the median operator generates tough problems including the higher reconstruction error, and the presence of burr in decomposition products. Aiming at addressing these problems while catering to a better time-frequency representation of the industrial oscillations, a median complementary EEMD (MCEEMD) method is proposed in this paper. In this work, the median operator and the mean operator are skillfully combined during the ensemble process. Through the study on simulation and typical industrial oscillation case, the effectiveness of MCEEMD is verified compared with existing methods, including EEMD, CEEMD and MEEMD.

Recommended citation: Songhua Liu, Bingbing He, Qiming Chen, Xun Lang, Yufeng Zhang. Chinese Control Conference (CCC). (2021). https://ieeexplore.ieee.org/abstract/document/9550346

Multivariate intrinsic chirp mode decomposition

Published in Signal Processing, 2021

A multivariate intrinsic chirp mode decomposition (MICMD) algorithm is proposed to process multivariate/multichannel signals. In contrast to most existing multivariate time-frequency decomposition techniques, the proposed MICMD can efficiently extract time-varying signals by solving a multivariate linear system. In this paper, we first define a multivariate intrinsic chirp mode (MICM) by assuming the presence of a joint or common instantaneous frequency (IF) among all channels. Then the IFs and instantaneous amplitudes (IAs) are modeled as Fourier series. IFs can be estimated using the framework of the general parameterized time-frequency transform and then the corresponding MICMs are reconstructed by solving multivariate linear equations through an extended least square method. MICMD can characterize a set of multivariate modes without requiring more user-defined parameters than the original ICMD. Its properties and advantages, including mode-alignment, computational complexity, filter bank structure, quasi-orthogonality, channel number and noise robustness, are investigated successively. MICMD outperforms both multivariate empirical mode decomposition (MEMD) and multivariate variational mode decomposition (MVMD) in extracting time-varying components. The computational complexity of the proposed MICMD is proven to be thus much faster than MNCMD, which is of complexity. In the end, we highlight the utility and superiority of MICMD in three real-world cases, including the periodicity analysis in meteorology (three-channel), the -rhythm separation in electroencephalogram (EEG) (four-channel), and the plant-wide oscillation detection in industrial control system (eleven-channel).

Recommended citation: Qiming Chen, Xun Lang, Lei Xie, Hongye Su. Signal Processing. (2021). https://www.sciencedirect.com/science/article/abs/pii/S0165168421000487

Detection and root cause analysis of multiple plant-wide oscillations using multivariate nonlinear chirp mode decomposition and multivariate granger causality

Published in Computers & Chemical Engineering, 2021

Plant-wide oscillation detection and root cause diagnosis are important for maintaining control performance. Existing methods are mainly limited to detecting single and time-invariant plant-wide oscillations. In this paper, a data-driven model combining multivariate nonlinear chirp mode decomposition (MNCMD) with multivariate Granger causality (MGC) is proposed to detect and analyze root causes for multiple plant-wide oscillations in process control system. First, an MNCMD-based detector is developed to capture the multiple plant-wide oscillations, where oscillating variables caused by different sources are automatically clustered into various groups. Then, MGC is applied to each group to obtain the root causes of multiple plant-wide oscillations. Compared with state-of-the-art detection methods, the proposed approach shows better performance in the following aspects: (i) ability to extract both single/multiple plant-wide oscillations; (ii) capability to process both time-invariant/time-varying oscillations and provide accurate time-frequency information. This work also outperforms original Granger causality and nonlinearity index-based method in providing clearer causal network. The effectiveness and advantages of the proposed approach are demonstrated with the help of both simulation and industrial case studies.

Recommended citation: Qiming Chen, Xun Lang, Shan Lu, Naveed ur Rehman, Lei Xie, Hongye Su. Computers & Chemical Engineering. (2021). https://www.sciencedirect.com/science/article/abs/pii/S0098135421000090

Causality analysis in process control based on denoising and periodicity-removing CCM

Published in Journal of Intelligent Manufacturing and Special Equipment, 2020

A novel causality analysis framework is proposed based on denoising and periodicity-removing TD-CCM (time-delayed convergent cross mapping). We first point out that noise and periodicity have adverse effects on causality detection. Then, the empirical mode decomposition (EMD) and detrended fluctuation analysis (FDA) are combined to achieve denoising. The periodicities are effectively removed through singular spectrum analysis (SSA). Following, the TD-CCM can accurately capture the causalities and locate the root cause by analyzing the filtered signals.

Recommended citation: Qiming Chen, Xinyi Fei, Lie Xie, Dongliu Li, Qibing Wang. Journal of Intelligent Manufacturing and Special Equipment. (2020). https://www.emerald.com/insight/content/doi/10.1108/JIMSE-06-2020-0003/full/html

Batch-normalization-based soft filter pruning for deep convolutional neural networks

Published in 2020 16th International Conference on Control, Automation, Robotics and Vision, 2020

As convolutional neural network contains many redundant parameters, a lot of methods have been developed to compress the network for accelerating inference. Among these, network pruning, which is a kind of widely used approaches, can effectively decrease the memory capacity and reduce the computation cost. Herein, we propose a competitive pruning approach based on Soft Filter Pruning (SFP) by taking account of the scaling factors y of Batch Normalization (BN) layers as the criterion of filter selection strategy. During the soft pruning procedure, in each epoch only y values of BN layers less than threshold are set to zero instead of setting the weights of selected filters in convolutional layers to zero. Compared to the existing approaches, the proposed method can obtain a highly increased accuracy on image recognition. Notably, on CIFAR-10, the proposed method reduces the same 40.8% FLOPs as SFP on ResNet-110 with even 0.87% top-1 accuracy improvement.

Recommended citation: Xiaozhou Xu, Qiming Chen, Lei Xie, Hongye Su. International Conference on Control, Automation, Robotics and Vision. (2020). https://ieeexplore.ieee.org/abstract/document/9305481

Extracting fetal heart rate from abdominal ECGs based on fast multivariate empirical mode decomposition

Published in 2020 16th International Conference on Control, Automation, Robotics and Vision, 2020

Abdominal electrocardiogram is an important means to obtain fetal health condition during high-risk pregnancy. In this paper, a novel method for extracting fetal heart rate from multi-channel mother abdomen electrocardiograms is proposed using fast multivariate empirical mode decomposition technique (FMEMD). Firstly, FMEMD decomposes the multichannel ECG signals into a set of modes. Two significant channels are selected according to the standard deviation of the fifth layer. Then the continuous wavelet transform technique (CWT) is applied to these two channels to denoise. The baseline is removed by zero-crossing rate. Following, the interference of the mother QRS complexes and non-overlapped fetal R-peaks can be eliminated and detected by CWT coefficient. The overlapped fetal R-peaks are obtained by combining the dynamic pattern matching program and creative algorithm. The proposed method achieves an accuracy of 99.9% on the existing data set, and the calculating time is only 1/6.39 of the MEMD-based method.

Recommended citation: Jiayue Zhang, Xiaozhou Xu, Qiming Chen, Lei Xie, Hongye Su. International Conference on Control, Automation, Robotics and Vision. (2020). https://ieeexplore.ieee.org/abstract/document/9305481

MNCMD-based causality analysis of plant-wide oscillations for industrial process control system

Published in Chinese Automation Congress (CAC) 2020, 2020

In this paper, a data-driven model combining multivariate nonlinear chirp mode decomposition (MNCMD) with multivariate Granger causality (MGC) is proposed to analyze root causes for multiple plant-wide oscillations in process control system. Firstly, an MNCMD-based detector is developed to capture the multiple plant-wide oscillations, where oscillating variables caused by different sources are automatically clustered into various groups. Then, MGC is applied to each group to obtain the root causes of multiple plant-wide oscillations. Compared with state-of-the-art detection methods, the proposed approach shows better performance in the following aspects: (i) ability to extract both single/multiple plant-wide oscillations; (ii) capability to process both time-invariant/time-varying oscillations and provide accurate time-frequency information. The effectiveness and advantages of the proposed approach are demonstrated with the help of both simulation and industrial case studies.

Recommended citation: Qiming Chen, Xiaozhou Xu, Yao Shi, Xun Lang, Lei Xie, Hongye Su. Chinese Automation Congress (CAC). (2020). https://ieeexplore.ieee.org/abstract/document/9327085

Multivariate nonlinear chirp mode decomposition

Published in Signal Processing, 2020

In this paper, a novel Multivariate Nonlinear Chirp Mode Decomposition (MNCMD) is proposed. In contrast to most existing multivariate time-frequency decomposition approaches, the proposed MNCMD is capable of handling time-varying signal efficiently in an elegant variational optimization framework. The multivariate nonlinear chirp mode is defined based on the presence of a joint or common instantaneous frequency component among all channels of input signal. Then the objective function of MNCMD is defined as the sum of mode bandwidths across all signal channels. The alternate direction method of multipliers (ADMM) algorithm is employed to optimize the MNCMD problem. MNCMD can extract an optimal set of multivariate modes and their corresponding instantaneous frequencies without requiring more user-defined parameters than the original NCMD. The effectiveness and advantages of the proposed MNCMD are demonstrated by studying its mode-alignment, filter bank structure, quasi-orthogonality, the influence of channel number, noise robustness, and convergence. Specifically, we highlight the utility and superiority of the proposed method in three real-world applications, including the analysis of an oceanographic float position record (two-channel), the separation of α-rhythms in electroencephalogram (EEG) data (four-channel), and the detection of plant-wide oscillations in industrial control systems (nine-channel).

Recommended citation: Qiming Chen, Lei Xie, Hongye Su. Signal Processing. (2020). http://chen-qiming.github.io/files/SP2020.pdf

Diagnosis of nonlinearity-induced oscillations in process control loops based on adaptive chirp mode decomposition

Published in American Control Conference (ACC) 2020, 2020

Nonlinearity-induced oscillation detection is of great significance for the control loop performance assessment. A novel nonlinearity-induced oscillation detector based on ACMD (adaptive chirp mode decomposition) is proposed in this work. ACMD is a powerful signal processing tool and can decompose the process variable into several sub-signals, called as chirp mode. Then, two common oscillation indexes, namely, the normalized correlation coefficient and the sparseness index, are adopted to identify the oscillations contained in these modes. In this way, only significant oscillatory modes are retained and can be further analyzed for nonlinearity diagnosis by investigating the relationships among different frequencies. Simulation and industrial cases highlight the effectiveness and advantages of our methodology in various cases.

Recommended citation: Qiming Chen, Junghui Chen, Xun Lang, Lei Xie, Chenglong Jiang, Hongye Su. American Control Conference (ACC). (2020). https://ieeexplore.ieee.org/abstract/document/9147951

Detection and diagnosis of oscillations in process control by fast adaptive chirp mode decomposition

Published in Control Engineering Practice, 2020

Even though several algorithms have been proposed in the literature for oscillation detection and diagnosis, they can work reliably only for a specific type of oscillation and there is a lack of a common framework that accommodates the detection and diagnosis for various types of oscillations. To tackle this problem, an FACMD-based (fast adaptive chirp mode decomposition) detection and diagnosis framework is established in this study. It consists of two common oscillation detection indices and a novel strategy for diagnosing nonlinear and linear oscillations. Apart from detecting and diagnosing various single/multiple oscillations in single-input single-output (SISO) loop, FACMD can also distinguish the combination of linear or nonlinear oscillations and contribute to the root cause analysis for plant-wide oscillations. Finally, a series of simulations and industrial cases are used for testing. Compared with the existing work, the proposed methodology has better detection and diagnosis accuracy and a higher level of automation, especially in processing complex multiple oscillations.

Recommended citation: Qiming Chen, Junghui Chen, Xun Lang, Lei Xie, Shan Lu, Hongye Su. Control Engineering Practice. (2020). http://chen-qiming.github.io/files/CEP2020.pdf

Detecting oscillations via adaptive chirp mode decomposition

Published in CAA Symposium on Fault Detection, Supervision and Safety for Technical Processes (SAFEPROCESS) 2019, 2019

High performance is a fundamental requirement for the maintenance of an industrial plant in face of increasing competitive pressure. Oscillation is one of the most common abnormal phenomena encountered in process industries, thus it is of great importance to detect oscillations before implementing the performance-improvement methods. This paper proposed a novel method based on the adaptive chirp mode decomposition (ACMD) to detect the oscillations in control loops. Firstly, the process variable is decomposed by the ACMD into several chirp modes, called as intrinsic mode functions (IMF). Then, the energy ratio, normalized correlation coefficient, consistency function and sparseness index are combined to identify oscillations contained in these IMFs. The proposed method is automatic and data-driven without requiring any prior knowledge about the underlying process dynamics. Simulation studies demonstrated the detection ability of our approach in five cases, i.e. normal condition, external disturbance, external disturbance and poor tuning, stiction, stiction with external disturbance. Results obtained from this approach on industrial data show that it can be readily implemented in industrial environment.

Recommended citation: Qiming Chen, Xun Lang, Lei Xie, Hongye Su. CAA Symposium on Fault Detection, Supervision and Safety for Technical Processes (SAFEPROCESS). (2019). https://ieeexplore.ieee.org/document/9213346

Detecting Nonlinear Oscillations in Process Control Loop Based on an Improved VMD

Published in IEEE Access, 2019

A novel detector based on improved variational mode decomposition (VMD) is proposed to detect the nonlinearity-induced oscillations. Despite its high adaptivity and frequency resolution, the effectiveness of VMD highly depends on parameters, including mode number K , initial center frequency ωinit , and the penalty coefficient α . To tackle this problem, an improved VMD is proposed, which involves: 1) the spectrum of phase-rectified signal averaging (PRSA) to determine optimal K,ωinit and 2) the summation of permutation entropy (SPE) to optimize α , respectively. The presence of nonlinearity can be monitored by investigating the relationships among different frequencies of the process variable (PV) in the control loops. In addition, the oscillation detector based on the improved VMD is capable of distinguishing multiple oscillations, even when both nonlinear and linear oscillations from different sources occur. The proposed method is completely adaptive and data driven, which acts without a priori knowledge. The validity of the raised approach is verified by a set of simulations as well as industrial applications.

Recommended citation: Qiming Chen, Xun Lang, Lei Xie, Hongye Su. IEEE Access. (2019). http://chen-qiming.github.io/files/IEEE-ACCESS-2019.pdf

Dr. Chen Qiming

Publications