Closed-loop Data Research Articles

Optimal Strictly Stealthy Attack Design on Cyber-Physical Systems: A Data-Driven Approach.

In this article, an issue of data-driven optimal strictly stealthy attack design for the stochastic linear invariant systems is investigated, with the aim of maximizing the system performance degradation under an energy bounded constraint and bypassing the parity-space-based attack detector. Importantly, the proposed attack policy refrains from the assumption that the system knowledge is known to attackers. A novel strictly stealthy attack sequence (SSAS), coordinating the sensor and actuator signals simultaneously, is proposed with a sufficient and necessary condition for the existence of such an attack presented. Specifically, the SSAS is parameterized as a vector in the null space of a specific matrix which is constructed by a parity matrix and the system Markov parameters. For the purpose of data-driven attack realization, modified subspace identification methods are utilized to achieve an unbiased estimation of the required parameters via the closed-loop data. On this basis, the attack design is formulated as a constrained optimization problem, an explicit solution to which is given to characterize the optimal strictly stealthy attack. Finally, the vulnerability of the cyber-physical systems is analysed from the perspective of the parameter selection for the parity space-based detector. A case study on a three-tank model verifies the efficiency of the proposed approach.

Reduced-order helicopter model identification from closed-loop data

This paper presents the development of an algorithm for open-loop transfer functions identification and reduced-order modelling of a dynamical system, from closed-loop data with highly correlated inputs. Suitable for aeronautical applications (particularly for intrinsically unstable vehicles such as helicopters), it allows the identification of aircraft transfer functions from the knowledge of the time histories of arbitrary external inputs and the corresponding actuated controls and responses. After a numerical verification of the proposed approach considering a simple analytical aircraft dynamical system, it is successfully applied to a realistic engineering problem consisting of identifying the transfer functions of the AW-09 helicopter that relate pilot inputs to vehicle attitude and kinematics. It is shown that helicopter responses to arbitrary pilot inputs predicted by the reduced-order model representation of the helicopter dynamics based on the rational approximation of the identified transfer functions are in good agreement with those determined through the high-fidelity nonlinear flight dynamics solver used to evaluate the closed-loop database.

Controller Design Based on Closed-loop Data of an Encrypted Control System

Controller Design Based on Closed-loop Data of an Encrypted Control System

Closed-loop identification of stabilized models using dual input–output parameterization

This paper introduces a dual input–output parameterization (dual IOP) for the identification of linear time-invariant systems from closed-loop data. It draws inspiration from the recent input–output parameterization developed to synthesize a stabilizing controller. The controller is parameterized in terms of closed-loop transfer functions, from the external disturbances to the input and output of the system, constrained to lie in a given subspace. Analogously, the dual IOP method parameterizes the unknown plant with analogous closed-loop transfer functions, also referred to as dual parameters. In this case, these closed-loop transfer functions are constrained to lie in an affine subspace guaranteeing that the identified plant is stabilized by the known controller. Compared with existing closed-loop identification techniques guaranteeing closed-loop stability, such as the dual Youla parameterization, the dual IOP requires neither a doubly-coprime factorization of the controller nor a nominal plant that is stabilized by the controller. The dual IOP does not depend on the order and the state-space realization of the controller either, as in the dual system-level parameterization. Simulation shows that the dual IOP outperforms the existing benchmark methods.

Blockchain-based collaborative data analysis framework for distributed medical knowledge extraction

To ensure the privacy preservation and transparent use of regulated medical big data at decentralized and distributed medical institutions, this paper proposes a blockchain-based collaborative data analysis framework to realize multiparty secure data sharing and cooperative medical knowledge extraction through a transparent and regulatory machine learning approach. A smart contract is employed on the blockchain as the underlying technique to realize autonomous control and transparent regulation of closed-loop data acquisition and analysis. Considering the execution complexity of smart contracts for analysis collaboration, Petri net is adopted to formulize the workflows of smart contracts, and it acts as the underlying on-chain learning (OcL) approach. Finally, an experimental case study is conducted using real-life medical data to verify and evaluate the effectiveness and efficiency of our framework. A prototype system is established to demonstrate the real-life distributed knowledge extraction demand of our cooperating company. Four groups of experiments are designed and conducted to determine the effectiveness and efficiency of the learning process. The results show that the proposed framework significantly outperforms federated learning (FL) in terms of accuracy on small datasets, where the framework achieves an accuracy of 55.050% compared to FL. Meanwhile, the framework exhibits superior convergence in loss compared to FL, with a difference of 76.663%. In the case of big datasets, the framework achieves a faster completion of model training by 58.883%, with lower CPU utilization by 44.023% and lower memory utilization by 16.227% compared to FL.

NC-SIMC: Neuro-Controller Simple Internal Model Control

This study introduces a novel Neuro-Controller Simple Internal Model Control (NC-SIMC) paradigm. Our primary goal is to develop a neural network architecture that integrates the logical principles of Proportional-Integral-Derivative (PID) controller design, as outlined by the SIMC rules, into a model capable of generating control actions without the necessity for training with closed-loop data. This is achieved by employing the Simple Internal Model Control framework to effectively train the adapted neural network structure. We present a machine-learning approach that combines a specialized inductive bias node by using a custom new layer inside the neutral net structure to address this goal. Hence, it bridges traditional control theory and machine learning. A key innovation of our NC-SIMC approach is its ability to learn and internally formulate a control structure and parameters. This enables it to establish suitable control actions based on the system’s state, setpoint, and feedback without relying on a fixed control structure. Our paper demonstrates the efficacy of NC-SIMC in stabilizing control actions and its adaptability across various operating conditions without requiring retuning. This performance is shown to be favorably comparable to traditional SIMC rules. We conclude by discussing the potential for future improvements in dynamic performance and the integration of constraints directly into the inductive bias layer, opening new avenues for advanced control system design. The main advantage of our strategy is that no closed-loop data is needed to identify the NC-SIMC PID controller. Instead, the model learns from the SIMC rules.

Prediction method for closed-loop data for emulated linear time-varying system

In this study, data-driven control methods, namely VRFT, FRIT, and estimated response iterative tuning (ERIT), are proposed for controlled plants in linear systems. Notably, the ERIT method allows the prediction of input/output data in advance and the adjustment of feedforward controllers. A method is proposed in this study for predicting closed-loop data for an arbitrary controller with a linear and time-invariant system as the controlled plant.

SEMPro: A Data-Driven Pipeline To Learn Structure-Property Insights from Scanning Electron Microscopy Images.

Analyzing hydrogel microstructure through scanning electron microscopy (SEM) images is crucial in understanding hydrogel properties. However, the analysis of SEM images in hydrogel research heavily relies on the intuition of individual researchers and is constrained by the limited size of the dataset. To address this, we propose SEMPro, a data-driven solution using web-scraping and deep learning (DL) to compile and analyze the structure-property relationships of hydrogels through SEM images. It accurately predicts the elastic modulus from SEM images within the same order of magnitude and displays a learned extraction of modulus-relevant features in SEM images as seen through the nontrivial activation mapping and transfer learning. By employing Explainable AI through activation map exposure, SEMPro validates the model predictions. SEMPro represents a closed-loop data collection and analysis pipeline, providing critical insights into hydrogels and soft materials. This innovative approach has the potential to revolutionize hydrogel research, offering high-dimensional insights for further advancements.

Data driven performance monitoring and retuning using PID controllers

This paper presents a joint methodology for performance assessment and retuning of PID controllers. The methods are data-driven and based on the set-point response collected during operation or tests, using as reference input steps, ramps, and other signals. For each set-point response, the IAE (Integral of Absolute Error) is computed and a statistical test is evaluated to check if performance is below a given threshold. A data-driven design method is applied if new controller gains are required to resume the performance, using closed-loop data and a reference model obtained also from data under operation with good performance. Both monitoring and design require only closed-loop data, and statistical tests are used to provide evidence that monitoring and retuning were successful. Three applications for two different pilot plants are presented, evaluating: (i) flow, level, and pressure loops, (ii) outer and inner loops, (iii) validation of closed-loop data, (iv) performance recovery after introducing faults that reduce performance, (v) performance monitoring and retuning for three levels of performance. As illustrated, the proposed methodology can be applied directly to closed loops even in the presence of very noisy signals, with no assumptions about the design method used for the controller design.

A data-driven LQG method for linear time-varying systems to monitor the controller performance in the batch process

PurposeThe flexibility of batch process enables its wide application in fine-chemical, pharmaceutical and semi-conductor industries, whilst its complexity necessitates control performance monitoring to ensure high operation efficiency. This paper proposes a data-driven approach to carry out controller performance monitoring within batch based on linear quadratic Gaussian (LQG) method.Design/methodology/approachA linear time-varying LQG method is proposed to obtain the joint covariance benchmark for the stochastic part of batch process input/output. From historical golden operation batch, linear time-varying (LTV) system and noise models are identified based on generalized observer Markov parameters realization.FindingsOpen/closed loop input and output data are applied to identify the process model as well as the disturbance model, both in Markov parameter form. Then the optimal covariance of joint input and output can be obtained by the LQG method. The Hotelling's Tˆ2 control chart can be established to monitor the controller.Originality/value(1) An observer Markov parameter approach to identify the time-varying process and noise models from both open and closed loop data, (2) a linear time-varying LQG optimal control law to obtain the optimal benchmark covariance of joint input and output and (3) a joint input and output multivariate control chart based on Hotelling's T2 statistic for controller performance monitoring.

Brain-Computer interface control of stepping from invasive electrocorticography upper-limb motor imagery in a patient with quadriplegia.

Introduction: Most spinal cord injuries (SCI) result in lower extremities paralysis, thus diminishing ambulation. Using brain-computer interfaces (BCI), patients may regain leg control using neural signals that actuate assistive devices. Here, we present a case of a subject with cervical SCI with an implanted electrocorticography (ECoG) device and determined whether the system is capable of motor-imagery-initiated walking in an assistive ambulator. Methods: A 24-year-old male subject with cervical SCI (C5 ASIA A) was implanted before the study with an ECoG sensing device over the sensorimotor hand region of the brain. The subject used motor-imagery (MI) to train decoders to classify sensorimotor rhythms. Fifteen sessions of closed-loop trials followed in which the subject ambulated for one hour on a robotic-assisted weight-supported treadmill one to three times per week. We evaluated the stability of the best-performing decoder over time to initiate walking on the treadmill by decoding upper-limb (UL) MI. Results: An online bagged trees classifier performed best with an accuracy of 84.15% averaged across 9 weeks. Decoder accuracy remained stable following throughout closed-loop data collection. Discussion: These results demonstrate that decoding UL MI is a feasible control signal for use in lower-limb motor control. Invasive BCI systems designed for upper-extremity motor control can be extended for controlling systems beyond upper extremity control alone. Importantly, the decoders used were able to use the invasive signal over several weeks to accurately classify MI from the invasive signal. More work is needed to determine the long-term consequence between UL MI and the resulting lower-limb control.

Task-Offloading and Resource Allocation Strategy in Multidomain Cooperation for IIoT

This study proposes a task-offloading and resource allocation strategy in multidomain cooperation (TARMC) for the industrial Internet of Things (IIoT) to resolve the problem of the non-uniform distribution of task computation among various cluster domain networks in the IIoT and the solidification of traditional industrial wireless network architecture, which produces low efficiency of static resource allocation and high delay in closed-loop data processing. Based on the closed-loop process of task interaction of intelligent terminals in wireless networks, the proposed strategy constructs a network model of multidomain collaborative task-offloading and resource allocation in IIoT for flexible and dynamic resource allocation among intelligent terminals, edge servers, and cluster networks. Considering the partial offloading mechanism, various tasks were segmented into multiple subtasks marked at bit-level per demand, which enabled local and edge servers to process all subtasks in parallel. Moreover, this study established a utility function for the closed-loop delay and terminal energy consumption of task processing, which transformed the process of multidomain collaborative task-offloading and resource allocation into the problem of task computing revenue. Furthermore, an improved Cuckoo Search algorithm was developed to derive the optimal offloading position and resource allocation decision through an alternating iterative method. The simulation results revealed that TARMC performed better than strategies.

Heteroscedastic Bayesian Optimisation for Active Power Control of Wind Farms*

Active power control of wind farms remains an open challenge due to inherent noise in wind power that arises from uncertain wind speed measurements and plant/model mismatch. To leverage the heteroscedastic nature of the wind power noise, heteroscedastic Bayesian optimisation (BO) is used for active power control of wind farms. BO utilises closed-loop performance data to tune the parameters of a stochastic model predictive controller (SMPC) in a systematic and data-efficient manner. This, in turn, allows for enhancing the closed-loop performance of the controller intended to decrease the power tracking error. A case study with 9 turbines in a 3 × 3 wind farm shows that the heteroscedastic BO approach achieves a reduced closed-loop power tracking error in terms of root-mean-square by 8.89% compared to one that relies on nominal BO and a decrease by 64.99% compared to a nominal model predictive controller (MPC) whose performance is not tuned using closed-loop data and BO.

Gradient-Enhanced Bayesian Optimization via Acquisition Ensembles with Application to Reinforcement Learning

Bayesian optimization (BO) has shown great promise as a data-efficient strategy for the global optimization of expensive, black-box functions in a plethora of control applications. Traditional BO is derivative-free, as it solely relies on observations of a performance function to find its optimum. Recently, so-called first-order BO methods have been proposed that additionally exploit gradient information of the performance function to accelerate convergence. First-order BO methods mostly utilize standard acquisition functions, while indirectly using gradient information in the kernel structure to learn more accurate probabilistic surrogates for the performance function. In this work, we present a gradient-enhanced BO method that directly exploits performance function (zeroth-order) and its corresponding gradient (first-order) evaluations in the acquisition function. To this end, a novel gradient-based acquisition function is proposed that can identify stationary points of the performance optimization problem. We then leverage ideas from multi-objective optimization to develop an effective strategy for finding query points that optimally tradeoff between a zeroth-order acquisition function and the proposed gradient-based acquisition function. We show how the proposed acquisition-ensemble gradient-enhanced BO (AEGEBO) method enables accelerating convergence of policy-based reinforcement learning by combining noisy observations of the reward function and its gradient that can be directly estimated from closed-loop data. The performance of AEGBO is compared to standard BO and the well-known REINFORCE algorithm on a benchmark LQR problem, for which we consistently observe significantly improved performance over a limited data budget.

MPC for Artificial Muscles Using FRIT based Optimized Pseudo Linearization Model

A McKibben-type artificial muscle exhibits strong asymmetric hysteresis characteristics. In a previous study, fictitious reference iterative tuning (FRIT) was applied to a 1-DOF rotational actuator applying these muscles because it was able to determine the controller parameters using only a pair of single closed-loop data. However, it is generally difficult to design a reference model properly without information on the plant, which is strongly related to the control performance, and sometimes this leads to performance degradation. To solve this problem, in this study, we propose a method combining FRIT and model predictive control (MPC). First, we define the pseudo linearization (PL) model, which is adopted as a predictor in MPC, and is used to optimization between the model and design a control system by matching the closed-loop system to the PL model using FRIT. Additionally, it can consider input constraints by combined MPC.

Big Data Health Care Platform With Multisource Heterogeneous Data Integration and Massive High-Dimensional Data Governance for Large Hospitals: Design, Development, and Application.

BackgroundWith the advent of data-intensive science, a full integration of big data science and health care will bring a cross-field revolution to the medical community in China. The concept big data represents not only a technology but also a resource and a method. Big data are regarded as an important strategic resource both at the national level and at the medical institutional level, thus great importance has been attached to the construction of a big data platform for health care.ObjectiveWe aimed to develop and implement a big data platform for a large hospital, to overcome difficulties in integrating, calculating, storing, and governing multisource heterogeneous data in a standardized way, as well as to ensure health care data security.MethodsThe project to build a big data platform at West China Hospital of Sichuan University was launched in 2017. The West China Hospital of Sichuan University big data platform has extracted, integrated, and governed data from different departments and sections of the hospital since January 2008. A master–slave mode was implemented to realize the real-time integration of multisource heterogeneous massive data, and an environment that separates heterogeneous characteristic data storage and calculation processes was built. A business-based metadata model was improved for data quality control, and a standardized health care data governance system and scientific closed-loop data security ecology were established.ResultsAfter 3 years of design, development, and testing, the West China Hospital of Sichuan University big data platform was formally brought online in November 2020. It has formed a massive multidimensional data resource database, with more than 12.49 million patients, 75.67 million visits, and 8475 data variables. Along with hospital operations data, newly generated data are entered into the platform in real time. Since its launch, the platform has supported more than 20 major projects and provided data service, storage, and computing power support to many scientific teams, facilitating a shift in the data support model—from conventional manual extraction to self-service retrieval (which has reached 8561 retrievals per month).ConclusionsThe platform can combine operation systems data from all departments and sections in a hospital to form a massive high-dimensional high-quality health care database that allows electronic medical records to be used effectively and taps into the value of data to fully support clinical services, scientific research, and operations management. The West China Hospital of Sichuan University big data platform can successfully generate multisource heterogeneous data storage and computing power. By effectively governing massive multidimensional data gathered from multiple sources, the West China Hospital of Sichuan University big data platform provides highly available data assets and thus has a high application value in the health care field. The West China Hospital of Sichuan University big data platform facilitates simpler and more efficient utilization of electronic medical record data for real-world research.

Performance-oriented model learning for control via multi-objective Bayesian optimization

The closed-loop performance of model-based controllers, such as model predictive control, largely depends on the quality of their underlying model of system dynamics. Inspired by the notion of identification for control, this paper presents a strategy for performance-oriented, data-driven model adaptation for control. The fundamental idea is to mitigate plant-model mismatch via improving the model’s control-oriented predictive quality for optimizing closed-loop performance measures of interest, as opposed to enhancing its general predictive accuracy. To this end, we leverage a composite model structure that consists of a prior system model (physics-based or data-driven) and a data-driven model that can be efficiently adapted in a performance-oriented manner towards maximization of the closed-loop control performance. To solve the performance-oriented model adaptation problem, we use multi-objective Bayesian optimization (MOBO) that can directly handle black-box and expensive-to-evaluate functions that are computed from noisy observations of closed-loop performance measures. The MOBO approach is demonstrated on a benchmark bioreactor case study. Simulation results indicate that, given a fixed budget of process runs, performance-oriented model adaptation can yield control-oriented models that result in a significant improvement in closed-loop performance, compared to model identification using closed-loop data, while systematically accounting for multiple performance measures. The proposed approach can be particularly useful for multi-objective model learning and auto-tuning of model-based controllers for processes with finite-time control objectives, where each process run is associated with a high monetary value.

Neural Network Training Using Closed-Loop Data: Hazards and an Instrumental Variable (IVNN) Solution

An increasing trend in the use of neural networks in control systems is being observed. The aim of this paper is to reveal that the straightforward application of learning neural network feedforward controllers with closed-loop data may introduce parameter inconsistency that degrades control performance, and to provide a solution. The proposed method employs instrumental variables to ensure consistent parameter estimates. A nonlinear system example reveals that the developed instrumental variable neural network (IVNN) approach asymptotically recovers the optimal solution, while pre-existing approaches are shown to lead to inconsistent estimates.

Reinforcing Hurst Exponent with Oscillation Detection for Control Performance Analysis: An Industrial Application

Control performance assessment is one of the most important components of control assets monitoring. In the past decades, many different data-based approaches have been presented to assess control performance. Due to the high number of control loops for production industries, the methods requiring less parameters and less priori information became popular in industrial use. Detrended Fluctuation Analysis (DFA) is one of these methods, requiring only closed loop data and a pair of segment length to estimate Hurst exponent based control performance. However, the selection of segment lengths plays an important role in estimation, especially for oscillating control loops. In this paper, we have proposed an additional step (i.e. a step zero) to DFA to estimate control performance. This step uses an autocorrelation based and robust oscillation detection and characterization method to identify the oscillation period. Then, the calculated period is used to select the minimum segment length which leads more robust and precise estimation of control performance based on DFA. The proposed method has been applied and tested on several industrial control loops. The comparisons of the results between two approaches are given.

Control Optimization via A Practical Closed-loop Identification Method for A Low-pressure Heater in Power Plant

During the operation of a power plant, open-loop experiments are prohibited and this results in great difficulties for proportional-integral (PI) controller optimization. To overcome this question, a practical closed-loop identification method with feedforward control is proposed in this paper. The theoretical analyses and the implementations are carried out based on closed-loop data, and simulation results have validated the effectiveness of the proposed method. Finally, the identified system built by the proposed method is applied to the control optimization for a low-pressure heater in a 330MW power plant. Based on the identified system, the PI parameters are retuned and optimized, and the field running data validate the obvious improvement of the disturbance rejection performance. The successful application shows a promising future of the proposed closed-loop identification in practical industrial processes.