An improved Non-dominated Sorting Genetic Algorithm Ⅱ (NSGA-Ⅱ) based on the Deep Q-Network (DQN) is proposed to solve the Multi-Objective Flexible Job shop Scheduling Problem (MO-FJSP) with the goals of minimizing makespan and energy consumption. The Markov decision process and a reward function are defined in the DQN algorithm, considering the influence of selected machines on makespan and energy consumption both locally and globally. This approach enhances the quality of the initial population of the NSGA-Ⅱ. The elite retention strategy of the NSGA-Ⅱ is improved to ensure population diversity during execution and preserve high-quality individuals throughout the evolutionary process. The effectiveness of the DQN algorithm in generating initial solutions is validated by comparing its initial solutions with those generated by a greedy algorithm. Furthermore, the improved NSGA-Ⅱ based on the DQN algorithm is compared with other heuristic algorithms on standard and simulation cases, demonstrating its effectiveness in solving MO-FJSP.

Aiming at optimizing the maximum completion time for Flexible Job Shop Scheduling Problem (FJSP), it proposes an Improved Snake Optimization (ISO) algorithm. This algorithm studies the snake optimization algorithm and uses two-stage coding to replace the real coding of the original algorithm, so that the improved snake optimization algorithm can update the position of the individual snake in the discrete space. In addition, to solve the problem of low initial population quality in the original algorithm, GLR strategy is used to balance the machining load to improve the initialization quality of the algorithm. The position change and interaction mechanism of individual snake in the original algorithm are redesigned by using two operators on the basis of preserving the evolution of the original snake group. Finally, the orthogonal experiment is used to analyze the algorithm parameters, and 15 benchmark examples and 1 case in the workshop are simulated and compared to verify the effectiveness and stability of the proposed algorithm to solve the problem.

Optimized workforce configuration plays important roles in design and operation of product assembly lines. Considering the characteristics of complex product assembly lines,such as multiple and intricate serial and parallel processes,as well as shared workforces within teams,the optimal allocation of workforces is modeled as a two-dimensional bin packing problem. An improved Best-Fit Decreasing (BFD) optimization algorithm is thus proposed to achieve satisfied workforce allocation scheme through sorting,placing,updating,inspecting,repeating,and judging steps. Furthermore,a software system for rapid modeling,simulation,and optimization of production lines based on the commercial Plant Simulation package has been developed,which automatically matches both the basic information of the production line and the optimized workforce allocation results with the simulation models,thereby improving the efficiency of production line modeling and simulation. Taking an industrial product assembly line as an example,the effectiveness of the optimization algorithm is verified,ensuring high utilization of workforces and enabling rapid simulation of the assembly line.

The existing collaborative control methods for the extrusion and the movement of six-axis robotic arms are mostly focused on filament extrusion methods,while the optimization for granular material extrusion methods is limited. To address the issue of motion asynchrony between end-effector motion speed and material extrusion speed in six-axis robotic arm granular material 3D printing,a six-axis robotic arm granular material 3D printer is constructed,and a coordinated control approach for six-axis robotic arm granular material 3D printing system based on speed smoothing and look-ahead algorithm is proposed. Optimization is carried out for working conditions such as lifting points,and corner points.The results indicate that the line width of the printed parts is uniform,the material accumulation or missing can be avoided,and the feasibility of this method is verified.

The assembly process information of complex aerospace products is typically hidden in multi-source information storage media such as process documents and 3D models,leading to low efficiency in reusing historical assembly processes and high time costs for learning assembly process knowledge. To address these issues,it proposes an intelligent question and answer method for assembly process information based on large language models to improve the effective utilization of assembly process knowledge. First,a method for representing assembly process information is designed to structurally present assembly process knowledge,enabling automatic knowledge graph construction. Secondly,a key information identification method is proposed,which combines large language models with a Bidirectional Long Short-Term Memory (Bi-LSTM) network and Conditional Random Fields (CRF). Finally,key information is used to generate retrieval queries for matching answers from the assembly process knowledge graph. The proposed method is validated using the assembly process information of a complex aerospace product,and the experimental results show a significant improvement in the efficiency of assembly process designers.

In order to meet the requirement of automatic navigation of robot arm in intelligent maintenance of railway overhead contact system,a research method is proposed to comprehensively solve the problem of path planning and obstacle avoidance. This method transforms the dual objective into a single constrained optimization problem. On this basis,the standard Rapidly exploring Random Tree (RRT) algorithm is improved,the map complexity assessment strategies and Gaussian mixed distribution sampling strategies are introduced to constrain the generation direction of random sampling points. By adding the angle constraint strategy and the variable step size mechanism near the obstacle,the generation direction of the random tree is always growing towards the target point,so that the asymptotically optimal path is planned.In addition,a Recurrent Neural Network based on the Beetle Olfactory Detection (RNNBOD) algorithm is designed to configure the optimal joint angle to drive the redundant manipulator end effector to move along the planned reference path,thereby reducing its computing cost. The simulation results show that the proposed method not only improves the search efficiency,node utilization and path quality of the standard RRT algorithm,but also successfully solves the problem of tracking and controlling the redundant manipulator during operation.

In order to solve the inverse kinematics solving problem of the cable-driven redundant manipulator,the particle swarm algorithm was improved based on the strategies of chaotic mapping and inverse learning,and sub-population particles as well as the dynamic adjustment strategy of the population size were introduced to propose a Dynamic Reorganization Multi-swarm Particle Swarm Optimization (DRMPSO) algorithm with chaotic initialization. The kinematic model of the cable-driven redundant manipulator was established,and the fitness function with the minimum position error and the minimum energy consumption at the end of the manipulator was constructed. By comparing with other algorithms for the inverse kinematics of the manipulator,the proposed dynamic reorganization multi-swarm particle swarm algorithm has obvious advantages in both solution accuracy and stability. Finally,the dynamic reorganization multi-swarm particle swarm algorithm was used in the motion planning of the robotic arm,and the simulation results show that the dynamic reorganization multi-swarm particle swarm algorithm can ensure the continuity of the attitude while the robotic arm reaches the end position,and the robotic arm motion process is smooth.

Aiming at the slow convergence and large error issues in solving the inverse kinematics of a snake-arm robot using traditional BP neural networks,a progressive BP neural network inverse kinematics solution was proposed. Firstly,a forward kinematics model was established. Then,the principle of optimal compliance was introduced. A weighted sum of joint angle changes was used as the penalty term in the loss function,ensuring continuity and smoothness of the joint angle changes in the robot′s joint movements. Finally,based on the constraints between the joints of the snake-arm robot,a BP neural network model with a hierarchical progressive structure was constructed to improve the solving speed.Simulation results show that the progressive BP neural network achieves high accuracy and fast convergence,effectively solving the inverse kinematics problem of the snake-arm robot.

Aiming at the problems of exoskeleton robots in terms of bio-imitability and portability,as well as the lack of research on the internal and external rotation motion of the third degree of freedom of the wrist,a wearable wrist flexible exoskeleton device was designed based on Shape Memory Alloy (SMA) actuation based on analyses of the mechanism of muscle movement during internal and external rotation motion of the wrist driven by the forearm. A biomechanically compliant exoskeleton device was designed using 3D software based on the mechanism of the pronator teres,musculus pronator quadratus,and supinator muscles in the anatomical structure of the human forearm,which is driven by SMA springs to achieve the internal and external rotation movement of the wrist. The biomimetic and flexible nature of this exoskeleton device is in line with the human wrist movement characteristics,and its light weight and convenience can effectively assist the patients with hand function deficiency to complete more flexible daily life activities,which provides new perspectives and methods in the field of hand function rehabilitation medical treatment.

To improve the vibration performance of the exhaust system for a specific vehicle model, the Average Drive Degree Of Freedom Displacement method (ADDOFD) was firstly used to determine hook position of the exhaust system,followed by a vibration performance analysis. Static analysis revealed that the support reaction forces of the hangers exceeded durability design requirements,while dynamic analysis indicated that the dynamic reaction forces transmitted from the hangers to the vehicle body were excessively large. Therefore,the sum of the dynamic reaction forces of all hangers was selected as the first optimization objective,and the weighted sum of the normalized static displacements and support reaction forces at the lugs was defined as the second optimization objective.Using HyperStudy software,Hammersley sampling and Design Of Experiments (DOE) were conducted,followed by variable screening. Ultimately,the torsional stiffness in the RY and RZ directions of the flex pipe and the dynamic stiffness in the Z direction of five hangers were identified as design variables. A response surface model was constructed,and Global Response Search Method (GRSM) was employed to determine the optimal solutions for the dynamic stiffness of the flex pipe and the hangers. One set of optimal solutions was substituted into the finite element model of the exhaust system for analysis. The results showed all optimization objectives meet the vibration performance design requirements. A valuable reference was provided for the design of hanger positions,as well as the selection of stiffness for flex pipes and hangers in exhaust systems.

A hierarchical control strategy for lateral stability based on nonsingular fast terminal sliding mode control is proposed to address the lateral stability problem of distributed-drive electric vehicles under complex working conditions. Firstly, the two-degree-of-freedom reference model of the vehicle is established to obtain the desired yaw rate and the desired side slip angle. Secondly, the upper controller designs a direct yaw-moment control strategy based on nonsingular fast terminal sliding mode control with the desired yaw rate and desired side slip angle as the control objectives, and uses grey wolf optimization algorithm to solve the global optimum of the controller parameters in order to improve the dynamic performance of the system and to ensure strong robustness. Then, based on multiple constraint conditions, a lower-level controller is designed using the quadratic programming method to optimize the torque distribution to each wheel. Finally, in order to verify the effectiveness of the control strategy, simulation experiments on the lateral stability hierarchical control strategy are conducted on the Simulink simulation platform under the sinusoidal rotation angle condition and the double shift line condition. The results indicate that the control strategy proposed in this study can achieve precise tracking of the vehicle′s yaw rate and sideslip angle. Under both sinusoidal steering and double lane change conditions, the overshoot of the yaw rate is optimized by 19 % and 22 %, respectively, while the overshoot of the sideslip angle is optimized by 52 % and 67 %, respectively. This demonstrates that the strategy can effectively control the lateral stability of the vehicle under complex driving conditions, meeting the requirements for vehicle stability control.

In order to improve the steering stability of four wheel motor driven vehicles when cornering at high speed,and accurately coordinate the differential control among the driving wheels,a differential steering control strategy based on driving torque distribution is designed. The strategy employs a layered control architecture. The upper controller calculates the total driving torque required by the vehicle based on sliding mode variable structure control algorithm and optimization of fuzzy global fast terminal sliding mode control for calculating additional yaw moment of vehicle differential steering based on improved particle swarm optimization algorithm. The lower controller optimizes the calculated total driving torque and additional yaw moment based on quadratic programming algorithm to obtain the driving torque of each wheel. Finally,the designed control strategy is verified by Carsim/Simulink software co-simulation,and the results show that compared to traditional control strategies, the differential steering control strategy can effectively reduce the peak response of the vehicle′s yaw rate and the side slip angle during high-speed cornering.

CFRP composite materials and aluminum alloy bolted structures were taken as the research objects to analyze the effects of different materials of the connected parts on the preload degradation.Based on ABAQUS/Standard software,a parametric finite element model of a real threaded bolt node was constructed to analyze the effects of different factors on the preload degradation. As the thickness of the connected parts increases,the degree of preload degradation decreases gradually. The stresses are mainly concentrated in the threaded part,and the increase in the thickness of the connected part slows down the stress degradation. Under static loading condition,the preload decayed by 6 538,5 542 and 5 437 N after 200 hours for 4.0,4.5 and 5.0 mm thicknesses of the connected parts,respectively.

Aiming at the requirement of constant force characteristics of proportional electromagnet for external solenoid valve,a parametric model of proportional electromagnet was established in the 2D module of Maxwell software. The correctness of the simulation was verified by the proportional electromagnet test bench. The correctness of the material and the magnetic circuit segmentation method was verified from the simulation results and the armature working stroke was determined. On this basis,the single factor parameter analysis method was used to analyze the influence of the structural parameters of the magnetic isolation ring height L₂,the radial clearance δ₄,the diameter of the limit piece D₀ and the radial clearance δ₂ on the constant force characteristics of the electromagnet in the working stroke of the armature. According to the analysis results,the orthogonal test method was used to optimize the selected parameters to obtain a better parameter combination. The simulation shows that the proportional electromagnet has better constant force characteristics when L₂ =4.1 mm,δ₄ = 0.30 mm,and δ₂ =0.05 mm. The optimal combination parameters are verified by simulation. The electromagnetic force variance is 12.96 % lower than the original structure,and the constant force characteristics are significantly improved.

The rapid product iteration in the electronics manufacturing industry demands increased assembly efficiency. Three-Dimensional (3D) PCBA models offer intuitive guidance for assembly processes,yet most PCBA design files in manufacturing still exist as 2D drawings,which limits their ability to convey component layout,connectivity,and assembly details effectively. Current automated methods for constructing 3D PCBA models lack sufficient automation,making it challenging to transform numerous 2D design files into 3D models. This study introduced a parametric modeling-based technique for rapid 3D PCBA model construction to address the need for 3D visualization in electronic assembly. A structured data model was defined to meet the 3D assembly process requirements,facilitating the extraction of geometric and positional information from PCBA files for effective parsing. Subsequently,parametric feature modeling for components and PCB was proposed,followed by a coordinate-matching algorithm that enables automatic component positioning and PCB assembly. A case study validated the efficiency and accuracy of this method,showing its potential to construct precise 3D PCBA models efficiently,aiding assembly process guidance and enhancing manufacturing efficiency.

To address the limitations of existing studies on silicon rod surface defect detection,as well as the issues of misdetection and imbalanced detection performance among different defect types under a unified confidence threshold, a multi-confidence-based surface defect detection system for silicon rods is proposed. Firstly,the mechanical structure of the system is designed with a diffuse reflective structured light illumination strategy. By integrating a strip light source with a white fiber-optic curtain,the system effectively resolves illumination challenges caused by the strong reflectivity and large dimensions of silicon rods. Secondly,a precise evaluation metric framework is constructed,centered on Accurate Detection Rate (ADR),Missed Detection Rate (MDR),and Over Detection Rate (ODR). Based on multi-view image acquisition of single rods and a dual-matrix discrimination method,the system establishes a regulation framework linking defect characteristics,evaluation metrics,and optimal confidence thresholds,enabling the identification of detection accuracy and false detection patterns. The proposed dual-matrix approach enables a comprehensive quantitative assessment of the system′s effectiveness across all perspectives. Finally,a defect-type-specific confidence segmentation method is introduced.According to variations in imaging features among defect types,optimal confidence thresholds are assigned to 8 types of defects,including pits and edge breakage,thereby constructing a highly efficient multi-confidence detection model. Experimental results demonstrate that the proposed model significantly improves the ADR across various defect types. In particular,the ADR for pits,bright spots,watermark,and smudges increased respectively by 8.1 %,4.8 %,2.2 % and 3.5 %. Additionally,most defects' exhibit decreased ODR and MDR,with notable reductions in the ODR for pits,bright spots,watermark,and smudges,and in the MDR for crystal cracks and bright spots. The results confirm that the multi-confidence detection system′s ability to accurately locate and identify defects,effectively overcoming prior research limitations and providing a robust and efficient solution for silicon rod quality control.

With the development of autonomous driving technology,object detection has become a prerequisite for safe driving in autonomous driving due to its precise perception of the environment,real-time processing,and robustness.In crowded traffic scenarios where vehicles frequently obstruct one another,occlusion is still a significant challenge. Existing improved algorithms had exhibited certain deficiencies in occlusion perception.To address these issues. Firstly,the study proposed a new backbone network module,C2fCIBG,to deal with poor information extraction. Then,the GAM attention mechanism and CIB deep convolution were added to the new module to better capture key features and improve extraction. Finally,since CIOU was not robust enough,the loss function was changed to MPDIOU. Changing the loss function helped the model adjust the prediction box′s position and size more accurately, improving feature extraction and classification in the target area. Thus,it met the needs of vehicle occlusion detection in autonomous driving,the KITTI dataset was chosen for the experiment. The improved algorithm was compared with the traditional YOLOv8n algorithm.Evaluation was based on three key metrics:detection accuracy,recall rate,and mean average precision.The results showed increases of 0.87 %,2.0 %,and 2.7 % respectively,which proved that the improved algorithm was viable for vehicle occlusion perception.

Additively manufactured metal components have important applications in the field of high-end equipment,but their manufacturing process is prone to porosity,cracks and other defects. The conventional ultrasonic C-scan method is difficult to stably detect defects due to the complex curved surfaces of additively manufactured components.An ultrasonic C-scan detection method for additively manufactured curved metal components based on the STereoLithography (STL) model was proposed. The normal vector parameters of the detection points on the workpiece surface required by ultrasonic scanning were extracted from the STL model of the additively manufactured components,and the scanning point coordinates were determined by the equal-arc length discrete method to realize the position finding of ultrasonic probes. A laser sintering additive manufacturing process was used to manufacture the titanium alloy test block with the aero blade profile,and C-scan experiments were conducted using a multi degree of freedom ultrasonic scanning system. The experimental results show that the STL model driven ultrasonic C-scan method effectively detects several typical additive manufacturing defects with sub millimeter sizes. However,there are deviations in the quantitative results of defect size based on C-scan images.

The built-in rotary encoder is pivotal for motor positioning and speed feedback,providing Instantaneous Angular Speed (IAS) signals that reflect the torsional vibration of the transmission system,thereby revealing mechanical faults and enabling the development of self-aware and self-diagnostic servo control systems. To address the challenge of extracting fault characteristics accurately from IAS signals of low-precision encoders using conventional signal processing methods,an IAS signal enhancement method was proposed. It commenced with acquiring the servo motor′s instantaneous angular position signals via an EtherCAT bus servo drive system,the IAS signals were subsequently obtained using timestamp extraction and local polynomial fitting differentiation techniques. Finally the acquired IAS signals were processed by the angular domain synchronized averaging technique. Validation through simulations and gear fault experiments confirmed the method′s efficacy,highlighting its potential in cost-effective health monitoring for drive systems.