Aiming at the demand for screw visual inspection in the appearance inspection of a certain complex product assembly quality, a real-time multi-scale small object detection and classification model (RMSDC) is studied. RMSDC uses YOLOv8n as the base network structure. SPD-Conv consisting of a space-to-depth (SPD) layer and a non-strided convolution layer is used to improve the performance of the model in low-pixel images and small target detection. A convolution attention mixing (CAMixing) block is introduced in the middle of the backbone network and the neck network, which combines the attention mechanism with the multi-scale convolution to enhance the multi-scale feature extraction capability and improve the model accuracy. Choose an ultra-lightweight and effective dynamic upsampler (DySample), which is lightweight and can preserve the high-level feature semantic information of small targets as much as possible. The deformable convolution v4 (DCNv4) is fused to the dynamic head (Dyhead) to improve the generalization ability of the model in terms of shape transformation. Through comparative experiments, RMSDC in this paper improves the recall (R), mAP0.5 and mAP0.5:0.95 by 3.7%, 1.5%, and 1.8% compared to the baseline model on the experimental dataset, and mAP0.5 reaches 98.5%. The excellent performance of RMSDC has also been demonstrated on the ROSD dataset.

为保障液态危化品装卸过程中的生产安全,引入基于局部异常因子(LOF)与欧氏距离进行安全鹤位分配.首先通过使用LOF计算空位点局部离群程度,再计算空位点与出口间的欧氏距离,采用模糊量化法对得到的 2 个指标进行无量纲化处理,将 2 个结果加权计算后生成最终的综合评价指标.针对装卸区域建立坐标系,通过鹤位分配算法评估候选鹤位的安全程度,依照候选鹤位的综合评价得分结果,选取当前轮次得分最高的鹤位作为最终分配结果.同时根据鹤位实时作业进行迭代优化计算,实现安全鹤位分配.实验结果表明,基于LOF与欧氏距离的安全鹤位分配算法可以同时保证作业鹤位的相对较大安全间隔和相对较短避险距离.通过多轮次候选鹤位点选取,装卸安全性得到改善.

In the context of assembling and docking cylindrical parts characterized by large dimensions, susceptibility to deformation, and heavy loads, pose planning is a crucial prerequisite for effectively controlling alignment mechanisms. It holds significant importance in improving docking precision and enhancing efficiency. Current research in pose planning often focuses on single research objectives, whereas in the assembly and docking of cylindrical parts, it is essential to consider a holistic approach that encompasses efficiency, motion smoothness, and motion complexity. In this engineering context, this study investigates a pose planning method for assembling and docking cylindrical parts based on fifth-order spline curves. Subsequently, a multi-objective optimization model is established, using total alignment time, jerk, and alignment motion complexity as objective functions. To address the issue of NSGA-II algorithm's susceptibility to local optima in high-dimensional problems, a self-organizing migration strategy is proposed, enhancing its global search capabilities. Finally, experimental analysis is conducted to validate the proposed method. The obtained Pareto solution set demonstrates superior diversity and uniformity. Additionally, the effectiveness of the generated pose trajectories is verified.

现有复杂产品装配制造成熟度等级评估依赖专家凭经验确定指标权重和指标评分,存在主观性较强、工作量大、耗时长、无法传承评价实例所蕴含的知识等问题.为了提高复杂产品装配制造成熟度等级评估的效率以及客观性,利用成熟度等级评价实例数据,研究基于 BP 人工神经网络和 Ada-Boost算法的制造成熟度等级评估方法.构建复杂产品装配制造成熟度评价指标体系,给出基于模糊评价法和隶属函数的评价指标及成熟度等级达成度量化方法,建立基于 BP 神经网络的复杂产品装配制造成熟度等级评估模型,并使用AdaBoost算法优化成熟度等级评估BP神经网络模型.采用复杂产品分系统装配制造成熟度评价数据集对评估模型进行训练和实验,分析 BP-AdaBoost 的评估结果,获得最优评价模型.实验结果表明,基于BP-AdaBoost算法的复杂产品装配制造成熟度等级评估方法具有较好的可靠性与准确度.

A neural network model based on convolutional attention mechanism (CBAM) multi-channel processing was studied for the fault diagnosis of rolling bearings under noisy actual working conditions in the attitude adjustment mechanism of a certain missile assembly and docking equipment. The model firstly carries out feature extraction through the first layer wide convolutional kernel, which enhances the robustness of the model and maintains the characteristics of the temporal sequence signal; the two channels adopt the improved first layer wide convolutional kernel deep convolutional network (IWDCNN) and the long-short-term memory neural network (LSTM), which combines the advantages of both of them in order to maximize the effect of the convolutional network in extracting the features; using the attention mechanism to adaptively assign weights to each of the features of the two channels, in order to enhance the effectiveness of the feature extraction. The attention mechanism is used to adaptively assign the weights to each feature of the two channels, which enhances the effective features while suppressing the influence of the invalid features on the prediction results of the model. Through comparative experiments, the results show that the model has higher classification accuracy and better training efficiency under noisy data.

The key to the motion stability control of heavy-duty mobile robots for transshipment of large products is the accurate acquisition of the sideslip angle, and model-driven based methods are difficult to estimate the state quantities due to the high cost of direct measurements, dynamical nonlinearity and uncertainty. To this end, a new data-driven state estimation method is proposed, which combines the feature extraction capability of convolutional neural network (CNN) and the data memory property of long-short-term memory network (LSTM) to generate a time-lag nonlinear state estimation model through the noise reduction of conventional sensor measurements by the wavelet denoising method. Simulation software is used to simulate different working conditions to collect motion state quantities to form a dataset for training and simulation experiments on the estimation model, and the simulation estimation results of the estimation model are analysed to obtain the optimal estimation model. The simulation experiments show that the estimation accuracy of the sideslip angle estimation model based on wavelet threshold denoising and CNN-LSTM network is better than that of the single LSTM network estimation model.

Software testing is an important means to ensure software quality. The quality and efficiency of software testing can be greatly improved by modeled software testing. Software test maturity model (TMM) is a reference model to guide software organizations to improve test maturity. However, there is a lack of guidance on software testing objectives and process improvement, which leads to poor enforceability and low execution efficiency. To solve the above problems, based on the maturity objectives and content of the five test levels of the TMM model, an improved software testing V model (RAD) is proposed, and a software quality evaluation method is proposed for the improved RAD model.

The complexity and difficulty of dynamic obstacle avoidance for AGVs are increased by the uncertainty in a dynamic environment. The adaptive speed obstacle method allows the size of the collision cone to be dynamically changed to solve this problem, but this method may cause the AGV to turn too much when it is close to obstacles, as the collision cone expands too fast, which may lead to unstable operations or even collision. In order to address these problems, we propose an improved speed obstacle algorithm. The proposed algorithm uses Kalman filtering to estimate the positions of dynamic obstacles and adopts the idea of forward simulation to build a speed obstacle buffer according to the estimated positions of obstacles, such that the AGV can use the predicted positions of obstacles in the next moment, instead of the current positions, to build a speed obstacle model. Finally, an objective function that balances efficiency and safety was established to score all the candidate speeds, such that the highest-rated speed could be selected as the candidate speed for the next moment.

Memory corruption is a root cause of software attacks. Existing defense mechanisms (e.g., DEP, ASLR, CFI, CPI/CPS, and DFI) either offer limited security guarantees or incur high performance overhead. In this paper, we designed and developed a fast out-of-band (OOB) integrity monitor dubbed FastDIM to protect both applications and kernels against memory corruption attacks with less overhead. With FastDIM, a program in question is statically hardened by a compiler module. After that, the integrity of sensitive program data such as control-flow transfers (e.g., code pointers) and security relevant non-control data (e.g., encryption keys) are automatically protected by a monitor at run time. The key differences between FastDIM and related work are in the following aspects: 1) FastDIM offers an OOB monitor that protects the programs independently rather than letting the protected programs verify themselves using inlined reference monitor (IRM); 2) FastDIM extends the concept of shadow stacks originally proposed in CFI to protect not only return addresses but also other sensitive data such as function pointers, vtable pointers, and user-annotated sensitive non-control data. Thus, the protection of FastDIM is beyond control-flow data; 3) FastDIM provides a fast communication mechanism between programs and the monitor, so that the integrity checks are performed efficiently without context switch; and 4) for a better scalability and compatibility, FastDIM does not rely on LTO and Cross-DSO to support applications with dynamically linked libraries. We implemented a Kernel version and a TrustZone version of FastDIM to protect both user programs and Linux/Android kernels. The evaluation results show that the average overhead of FastDIM is 4.4% on SPEC CPU2017 C/C++ benchmarks and around 3% on AnTuTu benchmarks.

With the continuous development of advanced production technology, Auto Guided Vehicle (AGV) is regarded as an important part of advanced industrial production line, and the types of AGV are increasing. However,the traditional differential steering AGV cannot achieve more flexible movement and turning in narrow warehouse space, so this paper selects omnidirectional AGV as the research object.The omnidirectional AGV is free to change the attitude of the vehicle according to the demand, this brings a lot of convenience in the process of use. However, due to vehicle attitude restrictions in some stations or routes during the vehicle driving, the vehicle is only allowed to pass in a certain attitude. For example, the narrow passage only allows the vehicle to pass parallel to the route on the narrowest side of the vehicle, which brings a lot of inconvenience to the omnidirectional AGV driving.In view of the above problems,this paper proposes a comprehensive planning method for omnidirectional AGV path and attitude, which can plan the attitude adjustment stations of AGVs in the driving process according to the limitation of the AGV's attitude at the station or route in the path planning, so as to help the AGV adjust its attitude and reach the target location smoothly.

In order to manage the aerospace product manufacturing, a quantified evaluation of product manufacturing readiness based on BP neural network is proposed. In view of the unique problems of Chinese aerospace product manufacturing, the risk factors of aerospace product manufacturing are analyzed, and each manufacturing factor is decomposed hierarchically to establish a three-level indicator hierarchy for the aerospace product manufacturing readiness evaluation. According to the evaluation indicator of manufacturing readiness, the qualitative indicators and the quantitative indicators are quantified according to the demand satisfaction and fuzzy mathematics membership function. Based on the BP neural network, a quantitative evaluation of Chinese aerospace product manufacturing readiness is modeled. The comprehensive scores of manufacturing readiness is calculated by BP neural network according to the indicator evaluation scores, then the manufacturing readiness level is evaluated quantitatively and objectively. In order to optimize the evaluation model for the aerospace product manufacturing readiness, trainrp and trainlm are selected as the training function respectively for training. The error analysis experiments show that the average relative error of the manufacturing readiness evaluation model using trainlm as the training function is small, which can provide a scientific method for the objective evaluation of the aerospace product manufacturing readiness.

Optimization of assembly scheduling is the key to Just-in-time production. To optimize assembly scheduling for reducing the assembly time, an improved adaptive genetic algorithm based on neighborhood search is proposed. Two-layer encoding is used to represent the assembly scheduling. Optimizing the selection of parent chromosomes by calculating the gap between two chromosomes with Inertial Mass and Euclidean Distance in gravitational search algorithm. At the same time, the genetic algorithm and the variable neighborhood algorithm based on N6 neighborhood structure are integrated to achieve the effective optimization of the total assembly scheduling. The algorithm is tested and analyzed based on the benchmark example, and the optimization results are compared with an actual assembly scheduling case to verify the effectiveness and feasibility of the algorithm, which validates the effect of optimizing assembly scheduling.

A zonal gas concentration estimation method, proposed in this paper, based on the Gaussian diffusion model and support vector regression is to improve the prediction accuracy of the gas concentration in the area where liquid hazardous chemicals volatilized. It Aims at the difficulty for a single sensor to estimate the gas diffusion in the whole space. It uses the Gaussian diffusion model as the basis. Meanwhile, lots of sampling data from a sensor array had been used to adjust its parameters, to improve the accuracy of the diffusion estimation. Then, using the support vector regression to correct the result of the Gaussian diffusion model can further improve the accuracy and retain good generalization capability. The experiment shows that the overall average error of this method in zonal concentration prediction is less than 19%, which is about 13.8% higher than the traditional estimation method like the Gaussian diffusion model. At the same time, it has a good generalization capability and improves the prediction effect of the overall situation of gas volatilization in the region by replacing the site monitoring method with zonal concentration estimation. This method provides an effective way for leakage identification and risks early warning in petrochemical enterprises.

It is very important to predict press-assembly quality by analyzing the massive quality data collected during high-precision servo mechanism assembly. Based on the relationship between the press-assembly force and quality of the high-precision servo mechanism, a method of predicting the press-assembly quality based on gray Markov model is proposed in this paper, and the press-assembly dataset of high-precision servo mechanism assembly is dealt with. Firstly, the system cloud gray prediction SCGM(1,1)c model is established for initial prediction; then, the Markov chain model is applied to upgrade the initial fitting accuracy to improve the prediction accuracy; after that, the prediction effects of gray model, system cloud gray model and gray Markov model are compared and analyzed through experiments. Experiments show that the gray Markov model has higher prediction accuracy and stability; finally, the outlier detection method of press-assembly force is studied to find out the control boundary value and define the control range of press-assembly force. The press-assembly quality prediction based on the grey Markov model provides a new method for that of the high-precision servo mechanism.

Focusing on controlling the press?assembly quality of high?precision servo mechanism,an intelligent early warning method based on outlier data detection and linear regression is proposed. Linear regression is used to deal with the relationship between assembly quality and press?assembly process, then the mathematical model of displacement?force in press?assembly process is established and a qualified press?assembly force range is defined for assembly quality control. To preprocess the raw dataset of displacement?force in the press?assembly process,an improved local outlier factor based on area density and P weight( LAOPW)is designed to eliminate the outliers which will result in inaccuracy of the mathematical model. A weighted distance based on information entropy is used to measure distance,and the reachable distance is replaced with P weight. Experiments show that the detection efficiency of the algorithm is improved by 5.6 ms compared with the traditional local outlier factor(LOF)algorithm, and the detection accuracy is improved by about 2％ compared with the local outlier factor based on area density (LAOF) algorithm. The application of LAOPW algorithm and the linear regression model shows that it can effectively carry out intelligent early warning of press?assembly quality of high precision servo mechanism.

Aiming at the high computational complexity of classification prediction algorithms for high-dimensional data with large scale and high dimensionality, an effective solution is to select a small number of feature subsets with high correlations among the many candidate features of high-dimensional data, and remove the irrelevant and redundant features. In this paper, based on the correlation of sparse scores and category features, the feature selection (ISSFS) algorithm based on sparse score and correlation analysis is studied to select the input features of the learning algorithm. The algorithm calculates the optimal feature subset by comprehensively analyzing the sparse score of each feature in the dataset and the degree of correlation between the feature and the category, so as to achieve the purpose of dimension reduction of high-dimensional data features. Simulation experiments show that the algorithm achieves better feature selection on UCI dataset and ice hockey game dataset, and the classification effect is good.

针对现有的离群数据检测算法时间复杂度过高,且检测质量不佳的不足,提出一种新的基于改进的OPTICS聚类和LOPW的离群数据检测算法。首先,使用改进的OPTICS聚类算法对原始数据集进行预处理,筛选由聚类形成的可达图得到初步离群数据集;然后,利用新定义的基于P权值的局部离群因子LOPW计算初步离群数据集中对象的离群程度,计算距离时引入去一划分信息熵增量确定属性的权重,提高离群检测准确性。实验结果表明,改进后的算法不仅提高了运算效率,而且提高了对离群数据检测的精确度。

In some wireless sensor network(WSN) security monitoring systems,the nodes transfer large amounts of data in a long time,which causes the phenomenon of power decreasing and power amplifier(PA) being burned in the wireless data transceiver unit,but this kind of fault diagnosis method is generally complex and inefficient.In order to solve these problems,based on the analysis of WSN cell-level fault diagnosis,this paper proposed a fault diagnosis method based on fuzzy neural network by using the current model of wireless data transceiver unit.Firstly,according to the relationship between the emission current,the temperature and the supply voltage,the current model is established.Then,the fuzzy neural network model structure is determined by the clustering algorithm,and the hybrid leaming algorithm is used to optimize the front and rear parameters of fuzzy rules.Finally,the fuzzy neural network parameters are extracted to establish the WSN node fault diagnosis model.The experimental results show that the presented fault diagnosis method of wireless data transceiver unit possesses low computational complexity and high diagnostic accuracy.Compared with Gaussian process regression model,the computational complexity of this method is reduced by 22.4 ％,and the diagnostic accuracy is increased by 17.5％.