Abstract:With the vigorous development of the manufacturing industry, the problems such as environmental pollution an resource shortage have gradually become prominent, seriously affecting the sustainable development of society. Therefore, the transformation of manufacturing energy saving and carbon reduction is an inevitable requirement for global green and lowcarbon development. As one of the most important parts of the manufacturing systems, production scheduling can realize efficient and green operation of the manufacturing systems through the reasonable allocation of resources. Under the context of green manufacturing, the research of green shop scheduling problem has become a hot spot in the field of production scheduling. Therefore, a systematic review of the existing research since 2018 from the aspects of the green parallel machine scheduling problem, the green flow shop scheduling problem, the green job shop scheduling problem, the green flexible job shop scheduling problem, and the green distributed shop scheduling problem was conducted, the shortcomings of the existing research was summed up, and the direction of future research was pointed out.

Abstract:To improve the disassembly efficiency and reduce disassembly energy consumption, the robotic parallel disassembly mode was introduced in the disassembly sequence planning problem, a robotic parallel disassembly sequence planning model was constructed, and a genetic algorithm based on reinforcement learning was designed. To verify the correctness of the model, a mixed integer linear programming model was constructed. In the algorithm, a goal-oriented encoding and decoding strategy was constructed to improve the quality of the initial solution.Q learning was used to select the best crossover and mutation strategies in the iteration process to enhance the algorithms adaptability. Finally, in an engine disassembly case with 34 tasks, the superiority of the proposed algorithm was verified by comparing with four classic multi-objective algorithms. The analysis of the disassembly schemes shows that the robotic parallel disassembly mode can effectively shorten the completion time and reduce disassembly energy consumption.

Abstract:Most current studies formulate the cloud workflow scheduling as a single-objective or multi-objective optimization problem with at most three objectives, which is unable to fully meet practical scenarios′ needs. To address the limitations above, many-objective cloud workflow scheduling model was established, where many indicators such as time, cost, reliability, resource consumption, load balancing, were taken into account. Then, an improved co-evolutionary algorithm was introduced to address this problem, where dual-stage search strategy and multi-indicator cooperation mechanism were employed to effectively balance the convergence and diversity of solution set, so as to enhance the search capability of algorithm. Experiments on seven types of real life workflow instances reveal that our proposal outperforms the existing peers and can find better scheduling schemes in most cases.

Abstract:To address the issue of the lack of generalization capability of deep reinforcement learning in flexible job shop scheduling problems, a method combining curriculum learning and deep reinforcement learning was proposed. The training instance difficulty was dynamically adjusted, with an emphasis on enhancing the training of the most difficult instances, to adapt to different data distributions and avoid the forgetting problem during the learning process. Simulation test results demonstrate that the algorithm maintained decent performance on large-scale untrained problems and benchmark datasets. It achieves better performance on four large-scale untrained problems with two artificial distributions. Compared to exact methods and metaheuristic methods, for problem instances with larger computational complexity, it could rapidly obtain solutions of decent quality. Moreover, the algorithm can adapt to flexible job shop scheduling problems with different data distributions, exhibiting a relatively fast convergence speed and good generalization capability.

Abstract:Aiming at the problem of complex structure and high fault occurrence in the attitude control system of launch vehicle, a multiple model fault detection and isolation algorithm was proposed. The small deviation attitude dynamics model of the launch vehicle was established, and the Kalman filter of the system was designed. Combined with the idea of special observer, multiple Kalman filter banks with different structures were used to generate corresponding residuals, so that a single residual was only sensitive to a fault of the sensor or actuator. The fault isolation strategy was deduced theoretically to achieve the detection and isolation of different fault types of the launch vehicle. Simulation analysis shows that when no fault occurs, the residual results do not exceed the set threshold, and the algorithm does not alarm; when the sensor or actuator fails, the proposed isolation strategy can accurately locate the fault, which verifies the effectiveness of the algorithm.

Abstract:To grasp the effect of normal overload on the airborne thermoelectric conversion system, the convective heat transfer process of working fluid in heat exchange ducts was simulated on the basis of the computational fluid dynamics software. Results show that the growth of normal overload makes the flow field structure inside the pipeline changed, which finally leads to the gradually decreasing wall temperature. Along the duct, the formation of one wall temperature peak of the heat absorption channel is closely related to the change of turbulent heat flux near the heated wall. When the normal overload increases from 0g to 2g, the flow structure at the front end of the heat absorption channel changes from 8 vortexes converts to the two main vortexes structure and the influence of secondary flow is gradually increasing, the formation of the two main vortexes makes turbulent heat flux of fluid near the heated wall increases gradually due to the thinner temperature boundary layer. Hence, local heat transfer coefficient can be increased by 80%. With the enhancement of heat transfer in duct, the wall temperature of the heat absorption channel decreases and the maximum of wall temperature drop can be up to 290 K.

Abstract:Aiming at the station keeping control problem of stratospheric aerostat in dynamic wind field, a station keeping controller designed based on deep reinforcement learning D3QN algorithm for different control channels of aerostat operated with ambient wind, studied the impact of different reward functions on the performance of regional resident controllers. Station keeping control simulation was carried out under the task constraint of a station keeping duration of three days and a station keeping radius of 50 km. Results show that: compared with the station keeping controller designed by DDQN method, the performance of the controller designed by D3QN method is significantly improved. When the control trajectory of aerosat is only adjusted by altitude, the average station keeping radius can reach 25.26 km, and the station keeping ratio is 96-25%. With the aid of horizontal propulsion, the average station keeping radius can be significantly reduced and the station keeping time ratio can be significantly increased. At the same time, the strong robustness of the station keeping controller based on deep reinforcement learning was verified, and the controller can be designed with different reward functions to meet the requirements of different station keeping tasks.

Abstract:In order to solve the problem that the super-pressure balloon in the near space tandem balloon system being subjected to special boundary and load conditions, a method for optimizing the local shape of the two poles of spherical super-pressure balloons under bidirectional tension based on the natural shape equations was proposed. Through the smooth connection of the numerical solution generatrix segment and the arc generatrix segment, the complete generatrix of super-pressure balloon with a certain angle at two poles was obtained, which was symmetrical up and down. The stress distributions of the super-pressure balloons before and after the local shape optimization were analyzed by the finite element method, and the effectiveness of the local balloon shape optimization method was verified. On this basis, the key factors affecting the effectiveness of local balloon shape optimization were analyzed, including the arc length of the numerical solution generatrix segment, circumferential stress input condition, local shape, and the radius of the arc generatrix segment, providing an important reference for the selection of key parameters in the local shape optimization of spherical super-pressure balloon.

Abstract:Satellite monocular pose estimation usually extracts the keypoints of the satellite in the images, and then solves the PnP (perspective-n-points) problem to obtain the relative position and attitude between the camera and the satellite, in which the accuracy of satellite keypoints detection is the key to determine the accuracy of monocular pose estimation. To solve this problem, a high-precision satellite keypoints detection method was proposed, which predicted the image coordinates of the keypoints and gave the uncertainty of keypoints coordinate prediction synchronously. Then, a weighted PnP constraint equation was constructed to solve the relative position and attitude on this basis, which significantly improved the accuracy of satellite monocular pose estimation. Relevant experiments were carried out on the public satellite monocular pose estimation dataset named SPEED. The experimental results show that the proposed keypoints detection method for synchronously predicting keypoints coordinates and their uncertainty can significantly improve the accuracy of satellite keypoints detection,and by solving the weighted monocular pose estimation problem, the accuracy of satellite monocular pose estimation has been significantly improved.

Abstract:With regard to accurately predicting thermodynamic environment for air-to-air missile, fluid-thermal-solid coupling analysis should be adopted, and coupling effects on flow field and structure-temperature field should be investigated. A fluid-thermal-solid multi-field coupling simulation model for air-to-air missile was established via using partition algorithm, and the coupling relationships among missile structural deformation, temperature and pressure were analyzed and the coupling effects on temperature and pressure simulation results were examined as well. Results show that, missile bending deformation is caused by aerodynamic and aerothermal, and the structure deformation could lead to the variation of missile temperature field and pressure field. The main reason for the variation is induced by the structure deformation from aerodynamic effect. The predicting accuracy of missile thermodynamic environment can be affected by the coupling effects. If the missile slenderness ratio or the angle of attack is small, or the flight speed is low, the coupling effects have a little influence on thermodynamic environment for supersonic missile. If the missile slenderness ratio and the angle of attack are large, and the flight speed is high, high predicting accuracy can be obtained by considering coupling effects.

Abstract:The boost-glide missile is a kind of precision guided weapons, which flies in complex and changeable flight conditions with high requirements for the guidance law. The flight state variation model of the missile was established, the improved optimal guidance law with the impact angle constraint was derived by the optimal control method, and the longitudinal guidance coefficients and lateral guidance coefficients were introduced into the guidance law. The influence of single guidance coefficient on guidance precision was analyzed and the selection method of guidance coefficients was determined. The influence of fixed impact angle constraint and dynamic impact angle constraint on guidance precision were analyzed according to the requirements of different flight missions. The numerical simulation and the hardware-in-the-loop simulation were finished with a small solid propellant boost-glide test vehicle. Simulated results show that the improved optimal guidance law is reasonable with high guidance precision.

Abstract:In order to meet the requirements of automatic assembly of equipment in the spacecraft cabin and obtain an assembly robot with small structure size, large workspace, high load capacity and high flexibility, a lightweight, high-load 8-DOF hybrid attitude adjustment robot cabin assembly system based on PRR/PR(PRR) R mechanism was proposed. By analyzing the position mapping relationship, velocity mapping relationship, Jacobian matrix and acceleration mapping relationship of the hybrid assembly robot, the dynamic model of the hybrid assembly robot was established, and the mapping relationship between the driving force, driving torque and joint speed was obtained. Furthermore, the stiffness model of the hybrid robot was established to solve the deformation degree of the mechanism after six dimensional force was applied to the end of the mechanism. ADAMS and ANSYS simulation models verify the kinematic, dynamic and theoretical stiffness models of the mechanism. It provides a feasible scheme and theoretical basis for the realization of large equipment assembly automation in a narrow and long space.

Abstract:To solve the inter-chiplet deadlock and network connectivity problems caused by link failures in multi-chiplet network, an optimized packet retransmission mechanism for multi-chiplet network was proposed. By using the “message merging” function in the retransmission mechanism, the number of control packets and the network load was reduced. By using the “message forwarding” function and adopting the forwarding to neighbor strategy, the fault-tolerant cost of the inter-chiplet network link failure was reduced. And more balanced load of the intra-chiplet network was realized. The simulation results show that the proposed method can increase the saturation bandwidth by 12-5%~25% with similar latency compared with the turn restriction strategy. Furthermore, it can increase the saturation bandwidth by up to 50% in case of link failures. “Message merging” can effectively reduce the number of control packets, thus reduce the overall load of the network. “Packet forwarding” has lower fault-tolerance cost and can achieve more balanced load of intra-chiplet network.

Abstract:Particle transport simulation is one of the main applications of high performance computers. But facing to its fast growing compute requirements, the general-purpose microprocessors cannot adapt to the particle transport program features, owing to the complexity architecture of its single core, and then it is difficult to obtain high ratio of performance and power. Therefore, the program features of the particle transport non-deterministic numerical simulation were extracted and analyzed. Based on the characteristics of the algorithm, the architecture of open-source microprocessor core was designed, including pipeline structure, branch prediction unit, multi-level Cache hierarchy and main memory design. A specific accelerator architecture was designed in accordance to the particle transport program features. The simulation results of running the particle transport program on the general architecture simulator show that, as compared with ARM Cortex-A15, the proposed specific accelerator can achieve 4.6 times performance improvement under the same power consumption, and 3.2 times under the same area.

Abstract:In the process of geometric model processing by mesh generation software, a filling method based on B-spline surface was proposed to solve the hole problem caused by missing geometric surface information. The hole boundaries were extracted from the given set of B-spline curves based on their topological relationship, and curve approximation fitting and combination techniques were employed to process the curves within individual holes to obtain compatible curves. Hole repair was achieved by first constructing unidirectional interpolating ruled surfaces and tensor product surfaces from the curves, these surfaces were then combined through interpolation and Boolean sum operations to generate bilinear difference B-spline surfaces for filling the holes. In addition, the ruled surface was applied as an alternate method in complex special holes to ensure the overall robustness of the method. Experimental results show that the method is highly general and can be applied to the dirty geometry repair of various types of morphological holes in real-world industrial geometric models, providing clean and closed geometric models for subsequent mesh generation.

Abstract:Image anomaly detection aims to identify and locate the abnormal region in an image. To address the issue on the insufficient utilization of different-level feature information in the existing methods, an image anomaly detection method based on multi-level feature fusion network was put forward. By using the pseudo anomaly data generation algorithm incorporated with the anomaly prior knowledge, the anomaly data of the training set were augmented, and then the anomaly detection task was transformed into a supervised learning task. A multi-level feature fusion network was constructed to enriches the low-level texture information and high-level semantic information of features by fusing the different levels of features in the neural network, which could make the features used for anomaly detection more discriminative. In the training phase, the score constraint loss and the consistency constraint loss were designed and combined with the feature constraint loss to train the whole network model. Experimental results on the MVTec dataset showed that the proposed model could achieve 98.7% AUROC (area under the receiver operating characteristic) mean value in the detection task, 97.9% AUROC mean value in the pixel-wise localization task and 94.2% mean value of per-region-overlap in the localization task, which outperformed several existing anomaly detection approaches.

Abstract:To solve the problem of lack of early reliability evaluation in complex system design process, a MBSE (model-based system engineering) method was proposed to support the design and reliability evaluation of complex systems. With the characteristics of complex system development, a MBSE modeling method based on MOFLP-R(mission, operation, function, logic, physics and reliability) was proposed to support design and reliability assessment of complex system. KARMA, a system modeling language based on GOPPRR(graph, object, property, point, relationship and role), was used to express the MOFLP-R process. Then, in order to evaluate the reliability of the complex system, the code generation was used to map the diagram to the numerical model. Finally, the method was applied to a hydraulic system case and the results showed that the proposed method is effective for complex system design and evaluation.

Abstract:In order to overcome the problems of high computational complexity and long simulation cycle caused by the characteristics of strong dynamics, high timeliness, multiple constraints, and strong coupling during the three-dimensional spatial deployment of UAV-BS(unmanned aerial vehicle base station), an EGO(efficient global optimization) algorithm was proposed to determine the three-dimensional spatial deployment location of UAV-BS. Considering that the EGO algorithm mainly obtains new sampling points by optimizing the EI(expectation improvement) function, the improved DE(differential evolution) algorithm was proposed to optimize the EI function. The improved DE algorithm improves the optimization ability and convergence speed by adopting the successful parent selecting framework and the offspring generation strategy self-adaptive selection framework. Three typical engineering problems were selected to test the performance of the improved EGO algorithm. The results show that the optimization ability, optimization speed, and stability of the improved EGO algorithm are significantly improved. On this basis, an application example of using the improved EGO algorithm to deploy a UAV base station in three-dimensional space was given.

Abstract:The proliferation of “low, slow and small” UAVs (unmanned aerial vehicles) poses a serious threat to flight safety in airspace. Accurate analysis of the characteristics of UAV echo signals is of great significance for the detection of non-cooperative UAVs. Based on the time-domain integral echo model of rotor UAV target and the principle of cepstrum algorithm, the frequency-domain expression and cepstrum expression of echo signal were derived, the corresponding relationship between echo signal parameters and frequency-domain and cepstrum characteristics was analyzed, and a parameter estimation method for UAV echo signal was proposed, and the effectiveness of this method was verified by simulation and measured data. The results show that, it can estimate the bandwidth and rotation frequency of UAV echo signal more accurately and provide an important reference for target detection and recognition of UAV.

Abstract:Aiming at the problem that the traditional subspace-like direction finding algorithm fails in underdetermined scenarios and requires the number of signal sources as a priori information, a GNSS spoofing source direction finding method based on coprime array was proposed to improve the application security of satellite navigation receivers in spoofing environment. The cyclic correlation matrix was constructed to reduce the impact of noise on the performance of the coprime array signal processing, and the virtual domain equivalent array signal was obtained by vectoring the cyclic correlation matrix. On this basis, an optimization problem based on sparse signal reconstruction in virtual domain was designed to achieve high-precision, multi-degree of freedom direction finding for sources by minimizing the fitting error. Simulation results show that compared with traditional subspace algorithm, the proposed algorithm has higher estimation accuracy, and the direction finding results are still reliable under the case of underdetermined.

Abstract:In order to guarantee the long-term stable communication of underwater devices, the IFM-SBL(sparse Bayesian learning based on improved fast marginal likelihood maximization) algorithm was proposed to estimate underwater acoustic channels with low complexity and high performance. Especially in the case of low SNR(signal-to-noise ratio), the performance of proposed algorithm can be further improved by threshold denoising and discrete Fourier transform denoising. Simulation and sea trial results show the output bite error rate after channel estimation of IFM-SBL is similar to that of EM-SBL(sparse Bayesian learning based on expectation maximization), and it has good robustness in both low SNR and fast or slow time-varying channels. The running speed of FM-SBL and IFM-SBL algorithm is 90% better than that of EM-SBL algorithm, which greatly reduces the estimation time.

Abstract:The corrosion electric field signal of ships has characteristics such as low frequency and difficulty in elimination, and it is a kind of physical field feature of ships with obvious line spectrum characteristics. Ships with different coating damage areas have distinct electric field distribution characteristics, and the corrosion electric field signal can be utilized to detect the coating damage location of ships. Therefore, a detection method combining RCHFRDE(refined composite hierarchical fluctuation revise dispersion entropy) and IHHO-KELM(improved Harris Hawk optimization-kernel based extreme learning machine) was proposed. RCHFRDE was used to extract the feature information of the corrosion electric field signal, which was then input into IHHO-KELM for training to detect the coating damage area. The effectiveness and reliability of the proposed method were verified through simulation experiments and scale model experiments of ships. The experimental results show that this method can effectively predict the single damage area of the ships coating. The detection accuracy rates of simulation data and measurement data reach 94-67% and 89-00% respectively. It can be used as an effective supplement to non-contact detection methods in cases with less prior environmental information.