Search

Many problems in industrial smart manufacturing, such as process operational optimization and decision-making, can be regarded as distributed non-convex optimization problems, whose goal is to utilize distributed nodes to cooperatively search for the minimal value of the global objective function. Considering data transmission mode, transmission condition and communication waste in industrial applications, it is meaningful to study the distributed non-convex optimization problem with an event-triggered strategy over time-varying directed networks. To solve such a problem, a distributed non-convex event-triggered algorithm is proposed in this paper. Under some assumptions on local objective functions, gradients and step-sizes, the convergence of the proposed event-triggered algorithm to the local minimum is proved. Moreover, it is obtained that the proposed distributed event-triggered algorithm has a convergence rate of O(1/ln(t)). Finally, two examples in industrial systems are provided to validate the effectiveness of the proposed algorithm.

Quality control in industry involves trained operators to manipulate and inspect metallic surfaces in order to identify, and eventually correct, manufacturing defects. These tasks are manually performed, and a poor performance (e.g., missing defects) leads to an increase of the costs and prolongation of the manufacturing time cycle. In this work, we propose a multi-agent robotic platform to autonomously perform Industry 4.0 quality control processes of metallic surfaces. The platform consists of three anthropomorphic robots with custom-made end-effectors designed to manipulate, inspect, and eventually correct a metallic frame of a motorcycle. The description of a novel multi-agent platform is followed by the presentation of the developed inspection procedure, in which a linear laser scanner is used to reconstruct the three-dimensional metallic surface of a motorcycle with a resolution of ~0.1 mm. In order to validate the platform, we perform a set of experiments to assess the performance of the robotic platform in a real Industry 4.0 scenario. Results confirmed that such a system guarantees a sub-millimetric precision to identify defects on complex-shaped metallic surfaces and effectively correct them. The proposed robotic platform can be adopted for overcoming the drawbacks of a traditional procedure that relies on visual-tactile manual defects correction (e.g., low-repeatability, high-subjectivity) and is scalable to different industrial applications. The proposed approach aims to elevate the role of operators to expert supervisors of the process, limiting the interactions with potentially-dangerous tools/procedures and thus improving the working conditions in an industrial 4.0 scenario.

Traditionally, numerical trajectory integration for shooting equation calculation and iterations for shooting with randomly guessed initial solutions deteriorate the real-time performance of indirect methods for on-board applications. In this study, the indirect method is improved to achieve real-time trajectory optimization of fuel-optimal powered planetary landings with the help of analytical shooting equation derivations and a practical homotopy technique. Specifically, the contributions of this paper are threefold. First, the analytical expressions for shooting equation calculation are derived to replace the traditional time-consuming trajectory integration. Consequently, the computational efficiency is significantly improved. Second, the original three-dimensional landing problem is connected with a simplified one-dimensional problem that only involves the vertical dynamics, and its analytical solution is obtained based on Pontryagin’s minimum principle. Third, starting with the analytical solution, the accurate solution of the original landing problem can be obtained through an adaptive homotopy process. Simulation results of Earth landing scenarios are given to substantiate the effectiveness of the proposed techniques and illustrate that the developed method can obtain a fuel-optimal landing trajectory in 5 ms with 100% success rate.

In the switching process of driving authority for man-machine cooperative driving, it should be ensured that the driver can take over the vehicle safely, stably and efficiently. This paper proposes a driving authority allocation strategy based on real-time allocation domain (RAD), which mainly includes the establishment of RAD and the dynamic optimization. The RAD is a comprehensive dynamic allocation internal, whose range is determined by the corresponding allocation values of driver's cognitive state, driver's muscle state and environment state. The dynamic optimization is to search for the optimal driving authority in RAD at any time, and its optimization objectives and constraints all contain time-varying parameters. This work proposes an improved simulated annealing algorithm to solve this problem. Simulation results show that the proposed driving authority allocation strategy can effectively allocate the driving authority according to the current condition and driver's state, and ensure the safety, stability and efficiency of the take-over process.

This paper presents and evaluates a unified cooperative perception framework that employs vehicle-to-vehicle (V2V) connectivity. At the core of the framework is a decentralized data association and fusion process that is scalable with respect to participation variances. The evaluation considers the effects of the communication losses in the ad-hoc V2V network and the random vehicle motions in traffic by adopting existing models along with a simplified algorithm for individual vehicle's on-board sensor field of view. Furthermore, a multi-target perception metric is adopted to evaluate both the errors in the estimation of the motion states of vehicles in the surrounding traffic and the cardinality of the fused estimates at each participating node/vehicle. The extensive analysis results demonstrate that the proposed approach minimizes the perception metric for a much larger percentage of the participating vehicles than a baseline approach, even at modest participation rates, and that there are diminishing returns in these benefits. The computational and data traffic trade-offs are also analyzed.

Vehicle trajectory collection usually faces challenges such as inaccurate and incomplete trajectory data, mainly due to missing trajectories caused by Global Navigation Satellite System (GNSS) outages. In this paper, a novel ensemble transfer regression framework is proposed for urban environments with transfer learning as the primary solution for constructing a fine-grained trajectory dataset during GNSS outages. First, GNSS and motion information are fused for the training process. Then, a regression-to-classification (R2C) process is employed to implement incremental training to adapt to dynamically changing environments. Third, to account for GNSS outages, transfer learning is integrated to construct a data filtering strategy that minimizes negative sample weights during the current scenario. Finally, a more accurate classification-type loss function for ensemble learning is designed to obtain the ensemble transfer regression model. We utilize real-world datasets to verify the accuracy of the comparative methods and the proposed framework in trajectory interpolation prediction. The experimental results show that our framework is significantly superior to the comparative methods.

Matrix calculations are often required for the analysis of any big-data cloud computing system. It is quite common to process big-data associated matrices possessing the sparsity and low-rank properties. In order to efficiently deal with big-data matrices, we propose a new divide-and-iterate framework, which can be invoked to solve an enormously large linear system of equations by taking advantage of factored matrices. The Kaczmarz algorithm (KA) is utilized here to design the parallel iterative algorithms which are capable of solving a large system of equations by iteratively updating the solution through the reduction into the factorized subsystems in parallel. The convergences of our proposed new iterative algorithms are justified by the rigorous proofs. Besides, the time- and memory-complexities are studied to demonstrate the resource efficiency of the proposed algorithms. Numerical experiments are also presented to illustrate the effectiveness of this proposed new framework.

In the process industry, it is essential to establish a data-driven soft sensor to predict the key variable that is difficult to online measure directly. The accuracy performance of data-driven soft sensors relies heavily on data. Unfortunately, it is hard to acquire sufficient and informative data from the samples with limited number, which is called as the small sample problem. For handling the small sample problem, it is a good solution to generating virtual samples according to the distribution of original data. This paper proposes an enhanced method of virtual sample generation utilizing manifold features to develop soft sensors using small data. First, T-Distribution Stochastic Neighbor Embedding (t-SNE) is utilized to extract the features of input data. The main idea of generating virtual samples is to use the interpolation algorithm to obtain virtual t-SNE input features and then the random forest algorithm is utilized to get the virtual outputs using virtual t-SNE input features. Finally, virtual samples using the proposed t-SNE based virtual sample generation (t-SNE-VSG) can be achieved. For the sake of confirming the effectiveness and feasibility of the presented t-SNE-VSG, a standard data set is first used. What is more, a small data set from an actual industrial process of Purified Terephthalic Acid is used to establish a soft sensor model. The results from simulations show that the accuracy performance of the soft sensor established with small data can be effectively improved by adding the virtual samples generated by t-SNE-VSG. In addition, t-SNE-VSG achieves superior accuracy to state-of-the-art virtual sample generation methods.

Data-driven models are obtained purely from data without complex theoretical modeling and without explicit model knowledge. This results in black box models whose traceability and reliability constitute a major challenge. This contribution addressed this issue and presents a novel hybrid estimation method, which enhances the reliability of artificial neural networks. Within this method, a simple model based on physical knowledge secures a model based on an artificial neural network. An unscented Kalman filter realizes the interaction of the two individual models. Thereby, a confidence level determines which model is trusted to a greater extent or even entirely. As part of the method for adjusting this confidence level, the input variables of the artificial neural network are related to the data used in training. The more often the artificial neural network has encountered a situation in the training process, the greater the confidence level will be. Finally, the confidence level is used to set the covariances of the unscented Kalman filter. In this contribution, the method is presented using the application of roll angle estimation for passenger cars. By using the hybrid method the reliability of the estimation is increased in comparison to the artificial neural network. For this purpose, sensor malfunctions as well as a sensor failure are simulated. These disturbances are compensated by the introduced method. In addition, the hybrid state estimator increases the estimation quality compared to the individual estimators. The proposed method can be applied to any problem, where knowledge-based models are available to secure data-driven models.

In this paper, an online optimization control (OPC) scheme is proposed for the angle control of MEMS scanning micro-mirror actuated via electromagnetic method. In order to avoid complicated process of building precise model of the MEMS micro-mirror, a control structure not relying on precise model is developed. In this structure, a simplified linear dynamic model is used to describe the basic dynamic behavior of the micro-mirror. Then, a model error compensation mechanism (MECM) is introduced to compensate for the model uncertainty caused by un-modelled dynamics and nonlinear factors, such as hysteresis. Moreover, a predictive gradient descent (PGD) based optimization mechanism is proposed to compensate the effect of time-delay and speed up the convergence of on-line optimization process. Then, the stability analysis of the proposed control system is implemented. Finally, the proposed control scheme is applied to the angle control of electromagnetically driven scanning micro-mirror (EDSMM) and the corresponding experimental results are presented.

The deep learning-based satellite image analysis and retraining systems are getting emerging technologies to enhance the capability of the sophisticated analysis of terrestrial objects. In principle, to apply the explainable DNN model for the process of satellite image analysis and retraining, we consider a new acceleration scheduling mechanism. Especially, the conventional DNN acceleration schemes cause serious performance degradation due to computational complexity and costs in satellite image analysis and retraining. In this article, to overcome the performance degradation, we propose cooperative scheduling schemes for explainable DNN acceleration in analysis and retraining process. For the purpose of it, we define the latency and energy cost modeling to derive the optimized processing time and cost required for explainable DNN acceleration. Especially, we show a minimum processing cost considered in the proposed scheduling via layer-level management of the explainable DNN on FPGA-GPU acceleration system. In addition, we evaluate the performance using an adaptive unlabeled data selection scheme with confidence threshold and a semi-supervised learning driven data parallelism scheme in accelerating retraining process. The experimental results demonstrate that the proposed schemes reduce the energy cost of the conventional DNN acceleration systems by up to about 40% while guaranteeing the latency constraints.

The distribution of abrasive particles in fluids is an important research topic in the study of abrasive waterjet cutting processes. However, it is impossible to obtain the accurate distribution law and influencing factors by performing only experiments; therefore, it is necessary to study abrasive waterjet cutting processes with the help of numerical models. The existing numerical models usually adopt the form of artificial settings for the distribution of abrasive particles in fluid. This method cannot accurately simulate the random distribution of particles. In this paper, the random algorithm method is used to simulate the impact azimuth and the random distribution of abrasive particles in water. The smoothed particle hydrodynamics (SPH) method is used to simulate the distribution of abrasive particles and the process of jet impingement. The influence of the particle distribution on the simulation results is studied. Comparisons show that the dent formed by the jet impinging on the target with random abrasive particles is similar to the dent from the actual cut, and the contour distribution of the dent is more uniform than that of the cut. The simulation results obtained by the SPH method are accurate.

The fundamental communication paradigms in the next-generation mobile networks are shifting from connected things to connected intelligence. The potential result is that current communication-centric wireless systems are greatly stressed when supporting computation-centric intelligent services with distributed big data. This is one reason that makes federated learning come into being, it allows collaborative training over many edge devices while avoiding the transmission of raw data. To tackle the problem of model aggregation in federated learning systems, this paper resorts to multiple reconfigurable intelligent surfaces (RISs) to achieve efficient and reliable learning-oriented wireless connectivity. The seamless integration of communication and computation is actualized by over-the-air computation (AirComp), which can be deemed as one of uplink non-orthogonal multiple access (NOMA) techniques without individual information decoding. Since all local parameters are uploaded via noisy concurrent transmissions, the unfavorable propagation error inevitably deteriorates the accuracy of the aggregated global model. The goals of this work are to 1) alleviate the signal distortion of AirComp over shared wireless channels, and 2) speed up the convergence rate of federated learning. More specifically, both the mean-square-error (MSE) and the device set in the model uploading process are optimized by jointly designing transceivers, tuning reflection coefficients, and selecting clients. Compared to baselines, extensive simulation results show that 1) the proposed algorithms can aggregate model more accurately and accelerate convergence; 2) the training loss and inference accuracy of federated learning can be improved significantly with the aid of multiple RISs.

Monitoring air quality in cities as well as indoors has become an important topic in recent years. Being easily and densely deployable and rather low-cost, chemiresistive gas sensors indicate a feasible technology for this problem, especially if they are assembled as e-noses tracking multiple gases in one device. Nonetheless, the long-term stability of such sensors poses a severe problem for their measurement accuracy, which substantiates the need for drift compensation procedures as well as robust algorithms in the gas prediction process. In order to test such drift compensation methods and to generate synthetic data for algorithm training, a simulation model would be highly useful. In this paper, we present our investigations towards a simulation framework aiming at generating typical sensor responses of a graphene-based gas sensor with an emphasis on its drift behavior. Different drift models are studied, implemented and compared to real measurement data from a lab setup. The evaluation shows promising results when compared to experiments from a real chemiresistive sensor to which the model was parametrized. By using a second setup with fundamentally different heating characteristics we observed that recalibration is a necessary step to generalize to different sensing modes and to ensure the overall quality of the simulation data. Overall, these investigations provide a proper basis towards tackling the challenges in drift simulation.

This paper considers an unmanned aerial vehicles (UAV) communication network, where UAV operate as an aerial mobile relay between a source and a destination of the network. To capture the “timeliness” of the received information at the destination node, a new performance metric named age of information (AoI) is considered. In addition, a short packet communication scheme maintains low latency in the proposed UAV wireless communication system. The finite block-length theory investigates the performances of short packet communications scheme in the UAV-assisted wireless communications system. In this paper, the Average Age of Information (AAoI) is estimated by applying the Stochastic Hybrid Systems (SHS) model. The SHS model comprises discrete states represents events that reset the AoI process and continuously dynamics that represent linearly growing age processes. Finally, a closed-form expression for the AAoI of the proposed UAV wireless communication system is derived and it can be used to estimate the optimal altitude, block length and other parameters.

Environmental context prediction is important for wearable robotic applications, such as terrain-adaptive control. System efficiency is critical for wearable robots, in which system resources (e.g., processors and memory) are highly constrained. This article aims to address the system efficiency of real-time environmental context prediction for lower limb prostheses. First, we develop an uncertainty-aware frame selection strategy that can dynamically select frames according to lower limb motion and uncertainty captured by Bayesian neural networks (BNNs) for environment prediction. We further propose a dynamic Bayesian gated recurrent unit (D-BGRU) network to address the inconsistent frame rate which is a side effect of the dynamic frame selection. Second, we investigate the effects on the tradeoff between computational complexity and environment prediction accuracy of adding additional sensing modalities (e.g., GPS and an on-glasses camera) into the system. Finally, we implement and optimize our framework for embedded hardware, and evaluate the real-time inference accuracy and efficiency of classifying six types of terrains. The experiments show that our proposed frame selection strategy can reduce more than 90% of the computations without sacrificing environment prediction accuracy, and can be easily extended to the situation of multimodality fusion. We achieve around 93% prediction accuracy with less than one frame to be processed per second. Our model has 6.4 million 16-bit float numbers and takes 44 ms to process each frame on a lightweight embedded platform (NVIDIA Jetson TX2).

The high di/dt before zero-crossings easily leads to delayed or failed operations for the active current injection mechanical DC circuit breaker, especially for small direct current interruptions. The pre-charging process also makes this type of circuit breaker hard to realize repeat operations economically in the high voltage DC system. This paper proposes a novel DC circuit breaker and its design method that deeply coordinating a superconducting fault current limiter with the active current injection DC circuit breaker. This method significantly reduces di/dt before current zero-crossings and enables repeat operations economically. Besides limiting fault currents, the superconducting fault current limiter also charges the resonance capacitor flexibly in the circuit breaker, realizing the injected reverse current keep a constant proportion to the arbitrary fault current. Validations of the flexible capacitor charge and repeat operations are made by simulation and low voltage experiments. The charging times are within 2.5 ms and the reverse current keeps 1.2 to 1.7 times over the fault current under arbitrary conditions. Di/dt at current zero is reduced by 80% under 30% of rated interruption capacity. Flexible capacitor charge experiments are consistent with the simulation design. The superconducting tapes can recover within 300 ms to support repeat operations.

Rubber tire container gantry cranes (RTG) are the most-commonly-used equipments to stack containers in ports. With four sets of ropes hoisting the cargo, the system dynamics become much more complicated/unpredictable compared with traditional gantry cranes, which imposes higher requirements to the skills of operators. During the delivering process, large sway motion of the container is often excited by inappropriate manual operations, which badly impairs efficiency and safety. This issue is made even worse when considering the elasticity of rubber tire, wind disturbances, and many other practical disturbances. To improve the working efficiency of ports, an automatic antiswing trajectory planning method is proposed for RTG in this paper. Specifically, reference trajectories are first generated to ensure that the trolley/rope accurately reach the desired position. After that, appropriate feedback terms are incorporated to online optimize the trajectories, so that the container swing motion is sufficiently suppressed even under unfavorable conditions. Lyapunov-based analysis is presented to guarantee the correctness of our conclusion. Furthermore, extensive hardware experiments on a practical RTG are performed to demonstrate the efficiency of the proposed method. To our best knowledge, the proposed method is one of the very few methods which can be successfully applied to the practical RTG.

To model threshold weighted voting systems with multifailure modes, we analyze system performance in terms of system reliability by focusing on indecisive effect. The measurement of system reliability is generated by the decision rule where we apply dynamic process on two major aspects including the indecisive parameter θ and voters' errors resulting from various situations such as weights replacement and inputs discrepancy. The motivation is to make the existing model more applicable via the dynamic process. In this article, two major contributions are addressed as follows. First, the continuous distributed indecisive parameter gives an intuitive idea of how the model adapts from the decision rule, which provides the flexibility of adjustment to satisfy different applications. Second, the time delay degradation process on voters' errors illuminates the opposite system performance for two types of inputs by considering different indecisive effects, which leads to further investigation of model sensitivity of input types.

Compressors play an important role in day-to-day operation in most oil and gas platforms, especially in the case for maintaining gas pressure in transportation pipe. Its complex problem to detect the sensors health and abnormality as the sensor reading would reflect the various states of the compressor. In ideal situation, sensor readings offer vast amounts of information on compressor health and could possibly indicate early fault of machines. Furthermore, due to harsh site and process operating conditions, sensors are often found to have drifted or failed, and there is no standard methodology to predict abnormality apart from applying emerging industrial smart sensor technologies. In this paper, we investigate a minimalist approach for detecting abnormality of compressor's shaft's RPM sensor. As the sensors in the compressor are correlated, we first use the outputs of other sensors to predict the shaft's RPM using regression-based models (neural networks and multiple linear regression). Second, we calculate the histogram of residuals by taking the difference between the predicted sensor value and the actual sensor value plus the abnormality in terms of bias/miscalibration and noise. The histogram of residuals can be used for sensor abnormality monitoring. In general, sensor states can be monitored by observing the shifting of the mean in the histogram of residuals. The sensor readings contaminated with noise can be seen by a shifted mean whose value is between the normal condition mean and the biased condition mean. This method is compact and would be relevant to monitor irregularity of the sensors.

One of the bottlenecks in flexible flank polishing is the non-uniform over-cuts problem caused by the difficulties of the polishing tool to conform to variable curvatures of the freeform surface along the polishing path. Consequently, it will significantly reduce the quality of finished surfaces and the service life of polishing tools. To solve the problem, a novel path generation method is proposed in this article, which not only gets the uniform over-cut at a single cutter location, but also keeps its uniformity over an entire polishing path. First, a tool orientation model is built on the freeform surface by analyzing the process of flexible flank polishing. Then, a novel indicator of over-cut uniformity is proposed with the developed Dupin indicatrix, and the tool orientation is determined by maximizing the indicator. Next, based on Taylor expansion, a polishing path generation method is presented with the consideration of over-cut uniformity. Finally, the proposed method is verified and analyzed relative to a comparison method by both virtual simulations and physical experiments. These results present a good consistency with each other and together provide strong support for the correctness, feasibility, and efficiency of the proposed method, which further contributes to surface quality and polishing tools service life.

This paper aims to handle the optimal attitude tracking control tasks for rigid bodies via a reinforcement learning-based control scheme, in which a constrained parameter estimator is designed to compensate system uncertainties accurately. This estimator guarantees the exponential convergence of estimation errors and can strictly keep all instant estimates always within pre-determined bounds. Based on it, a critic-only adaptive dynamic programming (ADP) control strategy is proposed to learn the optimal control policy with respect to a user-defined cost function. The matching condition on reference control signals, which is commonly employed in relevant ADP design, is not required in the proposed control scheme. We prove the uniform ultimate boundedness of the tracking errors and critic weight's estimation errors under finite excitation conditions by Lyapunov-based analysis. Moreover, an easy-to-implement initial control policy is designed to trigger the real-time learning process. The effectiveness and advantages of the proposed method are verified by both numerical simulations and hardware-in-loop experimental tests.

In the docking process of aeroengine rotor parts, docking accuracy that indicates the gaps between the end faces is strictly required. A key issue is improving docking accuracy using automated docking equipment. In this paper, a systematic study is carried out on the error modeling and compensation of a novel six-degrees-of-freedom (6-DOF) docking equipment for aeroengine rotors. First, a new model for indicating the main indexes of docking accuracy is proposed. Then, the error model of a specially designed 6-DOF docking equipment is established based on a modified Denavit Hartenberg method with five parameters. Subsequently, two error compensation methods are proposed. Based on the above models, a docking accuracy simulation algorithm is proposed using the Monte Carlo method. Finally, verification experiments are conducted. The results show that, for the maximum values and standard deviations of the gaps between the rotor end-faces in the actual and target positions and attitudes, i.e., main indexes that represent docking accuracy, the deviation rates between the simulation and experimental results are less than 20%. The modeling methods have referential significance. The decline rates of these values are 50–65% when using the two proposed compensation methods. The compensation methods significantly improve the docking accuracy.

Near-threshold computing is essential for energy-efficient operation of VLSI systems, but wide performance variation and nonlinearity to process variations block the proliferation. To cope with this, in this paper, we propose a holistic hardware performance monitoring methodology for accurate timing prediction in a near-threshold voltage regime. Precisely, (1) we formulate the problem of finding an efficient configuration of monitoring circuits into an instance of optimal experiment design problem and (2) propose a new timing prediction flow, consisting of statistical estimation of FEOL and BEOL process variations and a neural network based timing inference model. For accurate control of timing margin and overcoming simulation-silicon discrepancies, (3) we introduce uncertainty learning in the prediction model construction and calibrate it through transfer learning. Furthermore, (4) we avoid time-consuming SPICE simulations in our methodology by employing efficient but accurate surrogate models. Through simulations using a 28nm industry PDK and DK characterized at 0.6V operation, it is shown that our methodology is highly effective, reducing the average prediction pessimism of maximum delay by 77.9% over conventional signoff results while respecting target prediction yield. Besides, for test chips fabricated using a 10nm process, we demonstrated that our holistic approach from design to post-silicon phase in conjunction with adaptive voltage scaling reduces dynamic power consumption by 28.2% 28.8% on average, in comparison with typical supply voltage operation.

To ensure the structural safety and normal operation of a cable-stayed bridge in China (Rainbow Bridge), cables were replaced and construction monitoring and control were conducted. In the process of cable replacements, the cable force and girder elevation were monitored and the distribution of cable force was adjusted according to the construction monitoring results to improve the structural force and increase the safety reserve of the structure. In the process of monitoring the cable force, the difference between the measured cable force and the design cable force was 0.4–4.4%. After adjusting the cable force, the difference between the measured cable force and the design cable force was less than 2%, and the change in the cable force was uniform, meeting the original design requirements. In the monitoring process of the main girder elevation, the elevation was increased by a maximum of 0.054 m. After completing cable adjustment, the main girder elevation was increased by 0–5 mm. This elevation was more reasonable and met the design requirements.

This paper introduces a newly developed multi-sensor data fusion for the milling chatter detection with a cheap and easy implementation compared with traditional chatter detection schemes. The proposed multi-sensor data fusion utilizes microphone and accelerometer sensors to measure the occurrence of chatter during the milling process. It has the advantageous over the dynamometer in terms of easy installation and low cost. In this paper, the wavelet packet decomposition is adopted to analyze both measured sound and vibration signals. However, the parameters of the wavelet packet decomposition require fine-tuning to provide good performance. Hence the result of the developed scheme has been improved by optimizing the selection of the wavelet packet decomposition parameters including the mother wavelet and the decomposition level based on the kurtosis and crest factors. Furthermore, the important chatter features are selected using the recursive feature elimination method, and its performance is compared with metaheuristic algorithms. Finally, several machine learning techniques have been adopted to classify the cutting stabilities based on the selected features. The results confirm that the proposed multi-sensor data fusion scheme can provide an effective chatter detection under industrial conditions, and it has higher accuracy than the traditional schemes.

One of the most challenging problems in wireless sensor networks (WSNs) is detecting and recovering faulty data. Due to resource limitations, WSNs are subject to many failures, including hardware (permanent faults), software, and communication failures. To ensure data reliability, the data collected by WSNs must be clean of faults. This article presents a centralized faulty data detection and recovery approach, which can detect, recover, and recognize different types of faults (offset, gain, stuck-at, out of bound, and random faults). It operates in two phases: Faulty data detection and recovery and fault type identification (FTI). After detecting the fault, the faulty data will be discarded and replaced by an estimated value based on the Kalman filter. FTI provides a unique solution by determining the types of faults and report them to the end-user for an application-specific decision-making process. The effectiveness of the proposed approach is demonstrated through experimental results, with comparison to state-of-the-art techniques of the same application.

This paper addresses the issue of reliability in Industrial Internet of Things (IIoT) in case of missing sensors measurements due to network or hardware problems. We propose to support the fault detection and classification modules, which are the two critical components of a monitoring system for IIoT, with a generative model. The latter is responsible of imputing missing sensor measurements so that the monitoring system performance is robust to missing data. In particular, we adopt Generative Adversarial Networks (GANs) to generate missing sensor measurements and we propose to fine-tune the training of the GAN based on the impact that the generated data have on the fault detection and classification modules. We conduct a thorough evaluation of the proposed approach using the extended Tennessee Eastman Process dataset. Results show that the GAN-imputed data mitigate the impact on the fault detection and classification even in the case of persistently missing measurements from sensors that are critical for the correct functioning of the monitoring system.

Iterative learning control (ILC) aims to improve a system's performance by learning from previous iterations in repetitive operations. Traditional ILC requires the system to operate in a strictly repetitive way and the learning process is within the confines of one single system. This paper proposes an iterative learning method that extends traditional ILC from a temporal learning process within one single system to a spatial learning process between two heterogeneous systems. This learning mechanism is made viable by designing learning filters which bridge the dynamics of different systems. To design such learning filters, firstly, a dynamic system that consists of two heterogeneous systems and two learning filters is constructed; secondly, the two learning filters are decoupled from the known dynamics step by step. Each filter is successfully placed in the feedback loop of a reconstructed system consecutively, which enables us to transfer the filter design problem into a controller design problem that can be solved by leveraging the H-infinity optimal control theory. Meanwhile, the solvability of the learning filters, the stability of the system, and the learning convergence are rigorously analyzed. The proposed new algorithm is applied to the trajectory tracking problem of quadrotor drones. Both numerical and experimental studies have been conducted to validate the effectiveness of the proposed iterative learning method for multiple heterogeneous systems.

A dynamic mechanical model is proposed to describe the complexing actions of all the rolls on the ring during the ultra-large radial-axial ring rolling (RARR) process with four guide rolls. Based on the model, the calculation models for bending moment and normal stress at any section of the ring are deduced by force method. If the maximum section bending normal stress exceeds the yield stress of the ring materials, the ring will be distorted thus leading to the instability of the RARR process. According to this, a plastic instability criterion for the ultra-large RARR process with four guide rolls is developed, based on which a mathematical model to calculate the critical guide force for avoiding plastic instability of ring is obtained. The influence rule of the position of guide roll on the dangerous ring section of plastic instability is revealed, from which it is found the dangerous ring section mainly appears at the radial and axial deformation regions and the contact positions of the guide rolls and ring. The optimized layout of guide roll around the ring in favor of stability is determined to be about α1=61° and α2=119°. The plastic instability criterion is proven to be reliable from the aspects of the critical guide force, the section bending moment and normal stress and the dangerous ring section of plastic instability. Intelligent simulation case studies for the RARR process of ultra-large aluminum alloy ring indicate that the stable forming of the process can be effectively realized by regulating the guide force based on the plastic instability criterion. This work could provide a valuable guidance for the control of guide rolls and the optimization of the ultra-large RARR process with four guide rolls.

Developing machine learning enabled smart manufacturing is promising for composite structures assembly process. To improve production quality and efficiency of the assembly process, accurate predictive analysis on dimensional deviations and residual stress of the composite structures is required. The novel composite structures assembly involves two challenges: (i) the highly nonlinear and anisotropic properties of composite materials; and (ii) inevitable uncertainty in the assembly process. To overcome those problems, we propose a neural network Gaussian process model considering input uncertainty for composite structures assembly. Deep architecture of our model allows us to approximate a complex process better, and consideration of input uncertainty enables robust modeling with complete incorporation of the process uncertainty. Based on simulation and case study, the NNGPIU can outperform other benchmark methods when the response function is nonsmooth and nonlinear. Although we use composite structure assembly as an example, the proposed methodology can be applied to other engineering systems with intrinsic uncertainties.

The control objective of the vertical plane underactuated manipulator (VPUM) is generally to swing up the end point (EP) from the vertical down equilibrium point (VDEP) to the vertical up equilibrium point (VUEP), and finally, to stabilize it at the VUEP. The zone control methods are usually used to achieve this control objective, but the problem of these methods is that they cannot find a unified zone range, and it is easy to appear singularity in the motion process. In this article, for two-link VPUMs, we propose a control strategy based on trajectory planning and optimization, which does not need to divide the zone range, avoids the singularity problem, and can achieve the control objective quickly and smoothly. We design the trajectory by using the time scale method for the active link according to the initial and target states. Then, we obtain the trajectory parameters by employing differential evolution algorithm to ensure that the passive link can also move from the initial states to the target states simultaneously. We design a sliding mode tracking controller to make the active link move along the designed trajectory, so the EP is moved to VUEP. Meantime, we design a sliding mode stabilization controller to ensure that the system is stable at VUEP. Finally, the universality and robustness of the proposed method are proved by simulations.

In breast mass detection, there are many different sizes of masses in the image. However, when the existing target detection model is directly used to detect the breast mass, it is easy to appear the phenomenon of misdetection and missed detection. Therefore, in order to improve the detection accuracy of breast masses, this paper proposed a target detection model D-Mask R-CNN based on Mask R-CNN, which is suitable for breast masses detection. Firstly, this paper improved the internal structure of FPN, and modified the lateral connection mode in the original FPN structure to dense connection. Secondly, modified the size of the anchor of RPN to improve the location accuracy of breast masses. Finally, Soft-NMS was used to replace the NMS in the original model to reduce the possibility that the correct prediction results may be eliminated during the NMS process. This paper used the CBIS-DDSM dataset for all experiments. The results showed that the mAP value of the improved model for detecting breast masses reached 0.66 in the test set, which was 0.05 higher than that of the original Mask R-CNN.

Multiobjective multifactorial optimization (MO-MFO), rooted in a multitasking environment, is an emerging paradigm wherein multiple distinct multiobjective optimization problems are solved together. This article proposes an evolutionary multitasking algorithm with learning task relationships (LTR) for MO-MFO. In the proposed algorithm, a procedure of LTR is well designed. The decision space of each task is treated as a manifold, and all decision spaces of different tasks are jointly modeled as a joint manifold. Then, through solving a generalized eigenvalue decomposition problem, the joint manifold is projected to a latent space while keeping the necessary features for all tasks and the topology of each manifold. Finally, the task relationships are represented as the joint mapping matrix, which is composed of multiple mapping functions, and they are utilized for information transfer across different decision spaces during the evolutionary process. In the empirical experiments, the performance of the proposed algorithm is verified and compared with several state-of-the-art solvers for MO-MFO on three suites of MO-MFO test problems. Empirical results demonstrate that the proposed algorithm surpasses other competitors on most test instances, and can well tackle complicated MO-MFO problems which involve distinct optimization tasks with heterogeneous decision spaces.

There is a growing interest in neuromorphic hardware since it offers a more intuitive way to achieve bio-inspired algorithms. This paper presents a neuromorphic model for intelligently processing continuous electrocardiogram (ECG) signal. This model aims to develop a hardware-based signal processing model and avoid employing digitally intensive operations, such as signal segmentation and feature extraction, which are not desired in an analogue neuromorphic system. We apply delay-based reservoir computing as the information processing core, along with a novel training and labelling method. Different from the conventional ECG classification techniques, this computation model is a end-to-end dynamic system that mimics the real-time signal flow in neuromorphic hardware. The input is the raw ECG stream, while the amplitude of the output represents the risk factor of a ventricular ectopic heartbeat. The intrinsic memristive property of the reservoir empowers the system to retain the historical ECG information for high-dimensional mapping. This model was evaluated with the MIT-BIH database under the inter-patient paradigm and yields 81% sensitivity and 98% accuracy. Under this architecture, the minimum size of memory required in the inference process can be as low as 3.1 MegaByte(MB) because the majority of the computation takes place in the analogue domain. Such computational modelling boosts memory efficiency by simplifying the computing procedure and minimizing the required memory for future wearable devices.

The paper presents a stability analysis of a closed-loop system with an active disturbance rejection control (ADRC) algorithm and a reduced-order extended state observer (RESO). The controller is designed for first-order plus delay time processes, and one of the input signals to the RESO is delayed to compensate for the effect of the system delay. If the process delay is known, perfect synchronization between the observer input signals can be achieved. In a more realistic case, the process delay is not precisely known, and the resulting imperfect synchronization may lead to instability of the closed-loop system. The perfect synchronization case shows that the closed-loop system can be stabilized for any delay. For the imperfect synchronization case, delay-dependent stability conditions are derived for a simple tuning method.

This article proposes an adaptive state of health (SOH) estimation method for lithium-ion batteries using machine learning. Practical problems with feature extraction, cell inconsistency, and online implementability are specifically solved using a proposed individualized estimation scheme blending offline model migration with online ensemble learning. First, based on the data of pseudo-open-circuit voltage measured over the battery lifespan, a systematic comparison of different incremental capacity features is conducted to identify a suitable SOH indicator. Next, a pool of candidate models, composed of slope-bias correction (SBC) and radial basis function neural networks (RBFNNs), are trained offline. For online operation, the prediction errors due to cell inconsistency in the target new cell are then mitigated by a proposed modified random forest regression (mRFR) based ensemble learning process with high adaptability. The results show that compared to prevailing methods, the proposed SBC-RBFNN-mRFR-based scheme can achieve considerably improved SOH estimation accuracy (15%) with only a small amount of early-age data and online measurements are needed for practical operation. Furthermore, the applicability of the proposed SBC-RBFNN-mRFR algorithms to real-world operation is validated using measured data from electric vehicles, and it is shown that a 38% improvement in estimation accuracy can be achieved.

Real-world networks, such as phone-call networks and social networks, are often not static but temporal. Mining cohesive subgraphs from static graphs is a fundamental task in network analysis and has been widely investigated in the past decades. However, the concepts of cohesive subgraphs shift from static to temporal graphs raise many important problems. For instance, how to detect stable cohesive subgraphs on temporal networks such that the nodes in the subgraph are densely and stably connected over time. To address this problem, we resort to the conventional quasi-clique and propose a new model, called maximal ρ-stable (δ, ɣ)-quasi-clique, to capture both the cohesiveness and the stability of a subgraph. We show that the problem of enumerating all maximal ρ-stable (δ, ɣ)-quasi-cliques is NP-hard. To efficiently tackle our problem, we first devise a novel temporal graph reduction algorithm to significantly reduce the temporal graph without losing any maximal ρ-stable (δ, ɣ)-quasi-clique. Then, on the reduced temporal graph, we propose an effective branch and bound enumeration algorithm, named BB&SCM, with four carefully designed pruning techniques to accomplish the enumeration process. Finally, we conduct extensive experiments on seven real-world temporal graphs, and the results demonstrate that the temporal graph reduction algorithm can safely reduce 98% nodes of the temporal graph (with millions of nodes and edges) and BB&SCM is at least two orders of magnitude faster than the baseline algorithms. Moreover, we also evaluate the effectiveness of our model against other baseline models.

Transfer learning algorithm can provide a framework to utilize the previous knowledge to train fuzzy neural network (FNN). However, the performance of TL-based FNN will be destroyed by the knowledge over-fitting problem in the learning process. To solve this problem, an interactive transfer learning (ITL) algorithm, which can alleviate the negative transfer among different domains to improve the learning performance of FNN, is designed and analyzed in this paper. This ITL-assisted FNN (ITL-FNN) contains the following advantages. Frist, a knowledge filter algorithm is developed to reconstruct the knowledge in source scene by balancing the matching accuracy and diversity. Then, the knowledge from source scene can fit the instance of target scene with suitable accuracy. Second, a self-balancing mechanism is designed to balance the driven information between the source and target scenes. Then, the knowledge can be refitted to reduce the useless information. Third, a structural competition algorithm is proposed to adjust the knowledge of FNN. Then, the proposed ITL-FNN can achieve compact structure to improve the generalization performance. Finally, some benchmark problems and industrial applications are provided to demonstrate the merits of ITL-FNN.

Fault diagnosis which identifies the root of the abnormal status is of great importance to eliminate faults in the complex chemical processes. Many data-driven fault diagnosis models ignore different faults which occurred with varied frequencies in chemical plants, and they need a complete retraining process with the arrival of new fault modes. In this paper, a novel incremental imbalance modified convolutional neural network (II-CNN) is proposed to solve the aforementioned issues. The proposed method employs an imbalance modified method to extract the valuable information from the imbalance data, and generate new samples. After that, a local hyperplane-based dynamic Relief is designed to reduce the dimension of the chemical data and simplify the complex learning process. Finally, for the arrival of new fault modes, the proposed method is prompted in an incremental hierarchical way. Unlike the traditional models that are trained on static data, the proposed method inherits the existing knowledge and updates itself to include new coming fault classes. The proposed method is utilized in a simulated process and a real industrial process. Experimental results illustrate that the proposed method is better than the existing methods and has significant robustness and reliability in chemical fault diagnosis.

This paper deals with the modeling and estimation of the sociological phenomena called echo chambers and segregation in social networks. Specifically, we present a novel community-based graph model that represents the emergence of segregated echo chambers as a Markov bridge process. A Markov bridge is a one-dimensional Markov random field that facilitates modeling the formation and disassociation of communities at deterministic times which is important in social networks with known timed events. We justify the proposed model with six real world examples and examine its performance on a recent Twitter dataset. We provide model parameter estimation algorithm based on maximum likelihood and, a Bayesian filtering algorithm for recursively estimating the level of segregation using noisy samples obtained from the network. Numerical results indicate that the proposed filtering algorithm outperforms the conventional hidden Markov modeling in terms of the mean-squared error. The proposed filtering method is useful in computational social science where data-driven estimation of the level of segregation from noisy data is required.

Information granulation and degranulation play a fundamental role in granular computing (GrC). Given a collection of information granules (referred to as reference information granules), the essence of the granulation process (encoding) is to represent each data (either numeric or granular) in terms of these reference information granules. The degranulation process (decoding) that realizes the reconstruction of original data is associated with a certain level of reconstruction error. An important issue is how to reduce the reconstruction error such that the data could be reconstructed more accurately. In this study, the granulation process is realized by involving fuzzy clustering. A novel neural network is leveraged in the consecutive degranulation process, which could help significantly reduce the reconstruction error. We show that the proposed degranulation architecture exhibits improved capabilities in reconstructing original data in comparison with other methods. A series of experiments with the use of synthetic data and publicly available datasets coming from the machine-learning repository demonstrates the superiority of the proposed method over some existing alternatives.

Deep neural networks designed for point clouds, also called point cloud neural networks (PCNNs), are attracting increasing attention in recent years. In this work, we propose the first model compression framework called JointPruning (JP) that is specifically designed for compressing PCNNs. Observing that the redundancies in PCNNs are largely affected by certain parameters like the number of points, we first propose a new search space specifically designed for PCNNs. By searching the optimal pruning policy in our newly proposed search space, our JP framework can prune the PCNNs at different levels and simultaneously reduce the redundancies along multiple dimensions. As the newly proposed search space consists of multiple levels and the policy value at each level is continuous in our JP framework, it is hard to directly search for the best pruning policy in such a large search space. To this end, we further propose two strategies called search space refinement and validation set extension to progressively refine the granularity of our searching process in a coarse-to-fine and easy-to-hard fashion, which can help us gradually find better pruning policies. Comprehensive experiments on three benchmark datasets demonstrate the effectiveness of our JP framework for compressing PCNNs.

How to extract useful knowledge from a high-dimensional and sparse (HiDS) matrix efficiently is critical for many big data-related applications. A latent factor (LF) model has been widely adopted to address this problem. It commonly relies on an iterative learning algorithm like stochastic gradient descent. However, an algorithm of this kind commonly consumes many iterations to converge, resulting in considerable time cost on large-scale datasets. How to accelerate an LF model's training process without accuracy loss becomes a vital issue. To address it, this study innovatively proposes a multilayered-and-randomized latent factor (MLF) model. Its main idea is two-fold: a) adopting randomized-learning to train LFs for implementing a ‘one-iteration’ training process for saving time; and 2) adopting the principle of a generally multilayered structure as in a deep forest or multilayered extreme learning machine to structure its LFs, thereby enhancing its representative learning ability. Empirical studies on six HiDS matrices from real applications demonstrate that compared with state-of-the-art LF models, an MLF model achieves significantly higher computational efficiency with satisfactory prediction accuracy. It has the potential to handle LF analysis on a large scale HiDS matrix with real-time requirements.

The capability of deep learning (DL) techniques for dealing with non-linear, dynamic and correlated data has paved the way for DL-based fault detection and diagnosis (FDD). Among them, autoencoders (AEs) have shown their potential to serve as the fault detection network. However, misclassifying faulty samples that share similar patterns to normal samples is a common drawback of AEs. In this work, a source-aware autoencoder (SAAE) is proposed as an extension of AEs to incorporate faulty samples in the training stage. In SAAE, flexibility in tuning recall and precision trade-off, ability to detect unseen faults and applicability in imbalanced data sets are achieved. Bidirectional long short-term memory (BiLSTM) with skip connections SAAE is designed as the structure of the fault detection network. Further, a deep network with BiLSTM and residual neural network (ResNet) is proposed for the subsequent fault diagnosis step to avoid randomness imposed by the order of the input features. A framework for combining fault detection and fault diagnosis networks is also presented without the assumption of having a perfect fault detection network. A comprehensive comparison among relevant existing techniques in the literature and SAAE-ResNet is also conducted on the Tennessee-Eastman process, which shows the superiority of the proposed FDD method.

A model-free optimal tracking controller is designed for discrete-time nonlinear systems through policy gradient adaptive critic designs (PGACDs) with experience replay (ER). By using system transformation, optimal tracking control problems are converted into optimal regulation problems. An off-policy PGACD algorithm is developed to minimize the iterative Q-function and improve the tracking control performance. The proposed method is realized based on the critic network and the actor network (AN), which are applied to approximate the iterative Q-function and the iterative control policy, respectively. Then, the policy gradient technique is introduced to derive a novel weight updating law of the AN explicitly by using measured system data only. The convergence of the iteration is established through theoretical analysis, and the uniform ultimate boundedness is demonstrated for the closed-loop system under the PGACD-based controller by using Lyapunov's direct method. To guarantee the stability and increase the data usage efficiency of the learning process, an ER-based learning framework is designed to improve the realizability of the proposed method. Finally, simulation results of two examples are provided to demonstrate the performance of the off-policy PGACD algorithm.

Tubing pressure at gas injection depth in intermittent wells is one of the most critical parameters for production engineers to evaluate the performance of the system. However, monitoring of the tubing pressure is not usually carried out in real time. It has been realized that the generally used correlations are not effective enough due to complexity of the intermittent process which involve many parameters and assumptions to develop such equations. The focus of this study is to utilize machine learning (ML) algorithms to develop a model that can accurately predict tubing pressure in artificial intermittent gas lift wells. intelligent algorithms built on the field data provide a solution that is easy to use and universally applicable to the complex problems. Various non-linear regression ML methods are employed in this study, namely, Decision Tree-regression (DT), Random Forest-regression (RF) and K Nearest Neighbors-regression (KNN). All the tubing pressures obtained from ML models were compared with the actual values to ensure the effectiveness of the work. The developed models show that it can predict the pressure with more than 99.9% accuracy. This is an interesting result, as such outcome accuracy has not been reported usually in the open literature.

In general, vehicle images have varying resolutions due to vehicles’ movements and different camera settings. However, most existing vehicle re-identification models are single-resolution deep networks trained with pre-uniformly resizing vehicle images, which underestimates adverse effects of varying resolutions and leads to unsatisfactory performance. A straightforward solution for dealing with varying resolutions is to train multiple vehicle re-identification models. Each model is independently trained with images of a specific resolution. However, this straightforward solution requires significant overhead and ignores intrinsic associations among different resolution images. For that, an efficient multi-resolution network (EMRN) is proposed for vehicle re-identification in this paper. Firstly, EMRN embeds a newly-designed multi-resolution feature dimension uniform module (MR-FDUM) behind a traditional backbone network (i.e., ResNet-50). As a result, the whole model can extract fixed dimensional features from different resolution images so that it can be trained with one loss function of fixed dimensional parameters rather than training multiple models. Secondly, a multi-resolution image randomly feeding strategy is designed to train EMRN, making each mini-batch data of a random resolution during the training process. Consequently, EMRN can implicitly learn collaborative multi-resolution features via only a unitary deep network. Experiments on three large scale datasets, i.e., VeRi776, VehicleID, and VRIC, demonstrate that EMRN is superior to state-of-the-art vehicle re-identification methods.

Early warning for transformer faults is valuable for maintenance decision-making. However, limited work has been done in this area due to the difficulty of the model establishment. This paper proposes a two-stage feature extraction method for early warning of power transformer faults. By combining feature ranking and genetic programming (GP), a novel feature extraction process is presented. In the first stage, the data is labeled as normal and fault states and the feature extraction is evaluated on the data. Then, extracted key features and their growth rates are relabeled as normal and warning states. The feature extraction process is evaluated again on relabeled data. Obtained features and logic expression can be used for early warning. The proposed framework can implement an early warning with about 100 days in advance for transformer faults and is verified through 8 sequences of data. The comparisons with two recently reported methods show the superiority of the proposed method.

This paper proposes a cooperative optimal power split (COPS) method for a group of intelligent electric vehicles with battery/supercapacitor hybrid energy storage systems (HESSs). To achieve good performance, the proposed COPS method is made up of the upper and lower layers: the upper layer aims at obtaining the optimal power demand sequence and sending it to the lower layer; the lower layer attempts to optimize the power split for HESS. Firstly, to ensure the performance of the proposed method in practice, the resistance of the battery and supercapacitor packs as well as the DC/DC converter efficiency are not assumed to be constants, but modeled well in the paper. Secondly, to achieve the optimal power demand sequence, an upper power demand sequence optimization problem is formulated based on distributed model predictive control (DMPC), in which energy demand is incorporated into the cost function. Thirdly, after receiving the optimal power demand sequence, a hybrid power split optimization strategy is presented to obtain the optimal power split, in which the lower HESS power split optimization problem is formulated based on DMPC to decrease battery aging process and energy consumption. Finally, an improved particle swarm optimization algorithm with multiple dynamic populations is used to solve the formulated upper and lower optimization problems. Simulation results demonstrate that, compared with the benchmark, the proposed COPS method can significantly extend battery lifespan and slightly decrease energy consumption.