Low-field NMR(Nuclear Magnetic Resonance) technology has been widely used in the detection of physical properties of substances due to its fast and non-destructive characteristics. To solve the problem that the imaging accuracy of T₂ spectral is low, which affects the accuracy of detection results. Therefore the imaging method of NMR T₂ spectral based on Bayesian is studied. Firstly, the basic characteristics of NMR signals are showen. Based on the Bayesian principle, the likelihood function of NMR signals is deduced, and the T₂ spectral imaging framework is constructed. Secondly, the T₂ spectrum and its uncertainty are obtained by using an improved Markov chain Monte Carlo strategy. Finally, the effectiveness of the Bayesian-based NMR T₂  spectral imaging method is verified by randomly constructing a T₂ spectral model that obeys a multimodal mixture Gaussian probability density function. This method can be used as a comprehensive experimental content of communication principle, and can also be used as an innovative training experiment.

Seismic denoising and signal recovery is a key step to improve the quality and accuracy of seismic exploration. By combining the sparsity of the convolution framelet transform and the flexibility of the low dimensional manifold learning, a CFR-LDMM ( Convolutional Framelet Regularization based Low Dimensional Manifold Model) is proposed for seismic signal recovery by using the convolution framelet coefficient energy as the low dimensional constrains. The seismic signals are then jointly represented on the low dimensional manifold in a certain embedded space by the data-driven local and nonlocal basis function, avoiding explicitly defining the manifold coordinate function. Therefore, the significant improvement is made on the denoising ability and signal recovery accuracy. The results of the synthetic and field seismic data tests show that the CFR-LDMM can concentrate the energy of the framelet coefficients for seismic data into a certain block in the coefficient matrix, and the seismic random noise can be removed and the missing traces can be reconstructed well at low signal-to-noise ratio.

Due to the influence of instrument and equipment work, outdoor environment and other factors, there will be some random noise in the collected pipeline signal, which will make the original signal lose its characteristics, leading to the failure to accurately identify the pipeline signal. Therefore, a feature extraction method based on VMD (Variational Mode Decomposition) algorithm-entropy method is proposed. First VMD algorithm based on working condition of gathering pipeline deals with the noise signal, then from energy, impact properties, three angles, complexity of time series extracts signal characteristics under different working conditions of three kinds of signal reconstruction after the signal are calculated separately, and the energy entropy, kurtosis entropy and fuzzy entropy, and finally establishs characteristic vector input to the extreme learning machine to identify the condition. The experimental results show that the method proposed can classify and recognize pipeline working condition signals more accurately than other feature parameters, and the recognition rate is up to 98. 33% , which proves the feasibility of this method to classify and recognize pipeline leakage signals.

When the current method is used to predict the load of the power system of the distribution network, because the missing value interpolation processing of the power data is not performed before the power load prediction, the method has poor prediction accuracy, long prediction time. For the problem of poor forecasting performance, a research on the load forecasting of the distribution network power system based on the DE-ELM (Differential Evolution-Extreme Learning Machine) algorithm is proposed. This method first denoises the power data according to the wavelet transform method, completes the interpolation of the missing values of the power data according to the denoising results, and obtains a complete power data set; then divides the data set into two parts: a training set and a test set. The optimization method introduces the extreme learning machine, uses the DE-ELM algorithm to calculate the training set, builds a network model based on the results. Finally puts the test set into the constructed model for training, and realizes the load forecast of the distribution network power system based on the output results. The experimental results show that when the method is used to forecast the load of the distribution network power system, the forecasting accuracy is high, the forecasting time is short, and the forecasting performance is good.

In order to solve the problem of “information island" of science and technology resources, the deep integration and application analysis of heterogeneous data is performed. Guided by the application demand of multi perspective users for science and technology resources, the metadata characteristics of multi-source and heterogeneous science and technology resources are analyzed, and the integration design of heterogeneous data is carried out by using association aggregation method and knowledge organization tool to establish the relationship between various types of science and technology resources. Based on the application of science and technology resources metadata of information service platform of Jilin Province science and technology literature (referred to as “the platform"), the metadata storage and sharing service of multi-source heterogeneous data of the platform is realized, and the degree and effect of science and technology resources sharing service are enhanced.

Aiming at the problem that the current method does not consider the judgment of fault current, resulting in poor fault location and detection effect and low fault location accuracy, a fault locating algorithm for distribution network operation detection based on topological decoupling is proposed. According to the topological structure principle of distribution network with equivalent decoupling, based on the topological adjacency matrix of distribution network, several trunk networks with equivalent decoupling of distribution network are combined, and FTU(Feeder Terminal Unit) feeder terminal equipment is added to each trunk network to divide the single branch of distribution network with FTU. Kirchhoff current law is used to calculate the fault line and judge the fault current direction in the single branch. According to the judgment results, the fault interval judgment matrix is used to locate the operation inspection fault of distribution network. The experimental results show that this method has small error between fault location and actual location, more fault nodes and less false detection times, and its fault location and detection effect is good, which can effectively improve the fault location accuracy.

The pressure and blocking effect of the tunnel side wall on the traffic flow causes the tunnel traffic flow to deviate from the lane, reduce the operating speed or collide. The relationship between the traffic flow operating characteristics under the tunnel side wall effect and the tunnel section optimization needs to be studied. Therefor we firstly analyze the influence of tunnel sidewall effect on traffic flow. Secondly, combined with the knowledge of fluid mechanics, the theoretical modeling of the viscous characteristics of tunnel traffic flow is carried out. Finally, the relationship between lane width, traffic flow density and viscous coefficient, the relationship between traffic flow speed, sidewall spacing and viscous force are analyzed respectively. And the conventional viscous force calibration of traffic flow is carried out by taking a one-way multi-lane tunnel as an example. The relevant conclusions of the experiment will provide a theoretical guidance for tunnel speed management and section optimization.

Information enables the iterative upgrading of traditional logistics industry. In order to solve the transportation problems caused by the unique logistics characteristics of automobile manufacturing industry, it considers the improvement of the speed for milk-run, reducing the cost and alleviation the traffic pressure caused by logistics vehicles in the city, based on the actual transportation demands of milk-run of automotive equipment manufacturer A in city Q. An intelligent path planning method for automobile parts transportation based on improved GA(Genetic Algorithm) algorithm is designed. The genetic algorithm is improved by using the coupling factors such as the demand of parts and components in the current month, the details of supplier orders, the capacity rate of optional transportation vehicles, the volume proportion of single vehicle appliances, and the demand of time window in the process of milk-run. In this way, the optimal path using Solomon data example is solved and compared with genetic algorithm, and the optimal distribution scheme for solving the actual transportation demands between A and the suppliers. The experimental results show that the method has some advantages in performance. The numerical simulation results illustrate the applicability of the method and the convergence in the optimization process.

With the continuous development of distributed power supply, the load in distribution network presents a trend of diversification. The reactive power regulation strategy and traditional algorithm of traditional distribution network can not meet the demand of reactive power compensation of modern distribution network. Therefore, a dynamic reactive power optimization strategy of active distribution network based on improved krill swarm algorithm is proposed to solve the above problems. First of all, the reactive power optimization process is divided into two parts, the dynamic compensation regulation and days before recently considered discrete reactive compensation devices of reactive power compensation capacity, days after fully considering scenery output to compensate the system and other continuous adjusting equipment, active power distribution network based-days before dynamic reactive power optimization model of multiple time scales. Secondly, an improved krill colony algorithm based on cosine control factor and Cauchy factor is proposed to solve the model. Finally, the feasibility and effectiveness of this strategy are verified by the modified IEEE33 node system experiment, which can ensure the smooth operation of active distribution network and realize the maximum economic benefit.

Aiming at the problems of insufficient data and unbalanced sample distribution of oil pumping unit failures, a CDCGAN(Conditional Deep Convolutional Generative Adversarial Networks) model based on self-attention mechanism is proposed. The model adds a regular term to the loss function that constrains the distribution of generated images, improves the quality and diversity of generated images and effectively prevents the occurrence of mode collapse. Using Alexnet, VGG16 and other networks to classify and test the generated pumping unit fault samples, the experimental results show that the improved network generates higher quality data, can effectively balance the pumping unit fault data, and further improves the accuracy of the pumping unit fault diagnosis rate.

In order to solve the problem that the model fitting effect is not good due to the complex coupling relationship between drilling controllable factors, a prediction model of mechanical drilling speed based on fuzzy neural network is proposed. The fuzzy control idea is used to solve the parameter coupling problem and to predict. Clustering algorithm is used to divide the data with high similarity into a fuzzy set as the initialization parameter of the second layer of fuzzy neural network. Taking an oilfield as the background, the simulation results show that the empirical knowledge extracted by fuzzy neural network conforms to the coupling relationship between controllable parameters of drilling in the oilfield, and it is suitable for most drilling operations in the region. It proves that the model has good prediction ability, and verifies the feasibility and applicability of the model, which is of great significance to improve drilling efficiency and save cost.

In order to accurately detect the fault signal of the system, and to ensure safe and stable operation, the method of building fault detection filter, the H_∞ fault detection problem of neutral systems with time-varying delays is studied. The delay-dependent Lyapunov function is selected to make the fault detection system is asymptotically stable and satisfies H_∞ performance criterion. And the LMI(Linear Matrix Inequality) technique is applied to converted the design problem of fault detection filter into a convex optimization problem of LMI. Finally, the proposed method is applied to two-stage dissolution tank system. Simulation results show that the proposed method can detect system faults quickly and accurately, and the proposed method can reduce the conservatism of system design compared with delay-independent Lyapunov function.

At present, the tunnel lighting control often adopts the method of sectional control, and the lamps are always on, which leads to excessive lighting and waste of power. In order to reduce the energy consumption of tunnel lighting, using LED ( Light Emitting Diode) lamps and stepless control, the tunnel lighting design is carried out according to the human visual characteristics and traffic flow state parameters. Based on vehicle arrival and departure data, the control strategy of “ vehicle entry light on, vehicle exit light off" is adopted to design the lamp on and off control method from the two aspects of driving safety and energy saving. The experimental results show that the tunnel lighting changes continuously and gradually, which is more suitable for the visual characteristics of human eyes, and the method can reasonably control the on-off of tunnel lamps according to the traffic flow data, so as to ensure the safety of tunnel driving and reduce energy consumption.

Aiming at the problem that the development of fissures in the rock mass of grotto temple is slow and the influencing factors are diverse, it is difficult to predict the development of fissures. A new prediction network for the development of fissures in rock mass based on deep learning is proposed. It is a hybrid network with parallel self-attention mechanism. It models temporal correlations through local convolution modules and global recurrent modules to capture temporal patterns at different time scales accurately. Self-attention mechanism is introduced to model the complex dependencies between different sequences in multivariate time series data. To further improve the robustness of the model, traditional autoregressive processing is followed. We constructed the first dataset in this field based on the monitoring data of fissure development-related factors in Cave No. 32 of North Grotto Temple in Q City. Comparative experiments on this dataset show that the proposed model has a better performance in fracture development prediction of grotto rock mass.

In order to enable users to understand the integrity of cloud documents, a new retrievability proof scheme M-POR ( Proof of Retrievability-Based Message Authentication Codes) is proposed based on MACs (Message Authentication Codes). The MACs are generated by data blocks as authenticators, which can verify data integrity and locate the error data block. An original document is uploaded to the cloud server after encoding. Users can verify data integrity by using challenge-response-verification algorithms. The error-correcting code is introduced in encoding. If the error data is less than the threshold, the data can be recovered. Performance analysis shows that M-POR scheme can provide data integrity proof, and has low storage cost and calculation cost.

To solve the problems that the difference in crowd image background and the change in crowd scale caused by perspective effect have a serious impact on the accuracy of crowd counting, a multi-level fusion and attention mechanism based crowd counting algorithm is proposed, which includes two sub networks: scale attention extraction and multi-level fusion. The scale attention extraction network adopts coder-decoder structure, which is responsible for scale extraction to combat the problems of crowd scale change and crowd occlusion in complex crowd scenes; the multi-level fusion network adds a feature fusion operation before each convolution block to fuse the attention map with the input of each convolution block to remove the redundant image information, and then generate a high-quality crowd density map. Compared to other excellent crowd counting algorithms, the MAE(Mean Absolute Error) and MSE(Mean Squared Error) of the proposed algorithm on the ShangHaitech dataset Part _ B are increased by 17% and 25% , respectively, and the MAE on Part _ A is increased by 1. 7% . The MAE is increased by 7% on the UCF_CC_50 dataset. The experimental results show that the proposed algorithm has high accuracy and robustness in dealing with complex crowd scenes.

For the online learning problem of dynamic data flow, the traditional online BLS ( Broad Learning System) algorithm can not accurately capture the latest change trend of the data. Therefore, two online BLS algorithms with forgetting mechanism, one is based on forgetting factor ( FF-OBLS: Online Broad Learning System based on Forgetting Factor) and other is based on sliding window ( SW-OBLS: Online Broad Learning System based on Sliding Window), are proposed. FF-OBLS reflects the different contributions of old and new samples to the learning model by adding forgetting factors to old samples in the online learning process, SW-OBLS eliminates the impact of old samples on the learning model by deleting old samples in the online learning process, so as to enable the learning model to accurately analyze and predict the subsequent trend of dynamic data flow. In order to verify the effectiveness of the proposed two algorithms, dynamic regression data sets are used in the experiment. The experimental results show that the online BLS models with forgetting mechanism are better than the traditional online BLS model in the perspective of prediction accuracy and time cost, therefore they are more suitable to deal with dynamic data flow problems.

Aiming at the problems of poor security and slow encryption efficiency of color image, a spatial color image composite encryption algorithm based on chaotic mapping is designed. The depth residual network is constructed, the jump connection is added in the residual unit block, the nonlinear mapping of noisy image is created, and the activation function is moved to the convolution layer to accelerate the convergence rate of the network achieving the goal of image denoising. The wavelet transform operation is adopted for the image to analyze the spatial characteristics of the image, the Doppler wave is used to scramble the high and low frequency coefficients of the image, the diffusion operation is used to calculate the adjacent area of the symmetrical points of the image, and the pixels to be diffused are arranged and combined according to the zero mean normalized cross-correlation score, so as to make the chaotic sequence have sufficient coupling correlation with the image and strengthen the image security. The mapping parameters are introduced into Logistic chaotic mapping, the random values of mapping variables are adjusted, the probability distribution function of chaotic system sequence is calculated, and the composite encryption of color image is completed. Simulation results show that the proposed method can effectively eliminate the pixel feature information of color image, and the encryption effect is good and feasible.

Because the existing algorithms fail to calculate the semantic similarity, the detection results are not ideal, and an automatic detection algorithm for the deviation of the composition subject for college English teaching is proposed. In the college English teaching environment, combining distributed semantic space and structured semantic space, a semantic representation model is constructed to obtain the semantic similarity between English words and phrases. Through the LDA(Latent Dirichlet Allocation) model, all documents are trained, and the probabilistic weighted summation of each subject and feature words in the document is carried out, and the composition of the subject deviation is detected according to the set reasonable threshold. The results of simulation experiments show that the proposed algorithm can obtain high-precision automatic detection results of composition subject deviation.

At present, State Grid Jilin Electric Power Co. , Ltd. relies on confidential personnel to manually mark the confidentiality level of documents, and its accuracy depends on the professional quality of relevant personnel, which is easy to cause the problem of inaccurate labeling. Therefore, we establish an enterprise document security classification system based on the transformer model, which can automatically extract the feature expression of text security information and intelligently assist the decision-making of enterprise secret documents. The proposed model is trained and tested on the data set constructed by the internal core commercial secret files, ordinary commercial secret files and non secret files of State Grid Jilin Electric Power Company Limited. The accuracy rate is 97. 37% and the recall rate is 98. 67% . The results show that the model achieves high recognition effect and can effectively prevent the disclosure of secret files.

In order to explore the status quo and development trend of electronic payment research, an analysis of electronic payment research hotspots is performed. We use bibliometric method to further grasp the development status of electronic payment field from aspects of research overview, research strength and discipline distribution in the field of electronic payment, and uses the method of scientific knowledge mapping to visualize the research hotspots. The results show that China and the United States produce the most papers in the electronic field. The main output institutions are Beijing University of Posts and Telecommunications and Central South University. The research focuses on the theory and practical application of the security of electronic payment system.

The aim is to complete the master data management requirements of colleges and universities, comprehensively analyze the current situation and management requirements of business data, and build a master data management platform based on the micro service distributed architecture. A platform is developed and deployed based on spring boot microservice architecture and devops mode to realize master data related microservices such as basic microservices, business microservices and interface microservices, so as to meet the requirements of master data life cycle management. The actual operation shows that the platform can operate stably and efficiently in the daily management of colleges and universities, and meet the effectiveness and integrity management objectives of all kinds of data in university.

In order to help undergraduates understand the technology of satellite remote sensing and master machine learning algorithms, combined with college student innovation training program in Jilin University, a project named “Research on Change Detection of Buildings around Campus Based on Remote Sensing Images" is designed. GF-2 satellite images and JL1-3B night time glimmer images are used as experimental data, and the area for experiment is around a primary school in China. Various machine learning algorithms are used to extract the message of buildings in the area for experiment in different periods, and the precision of the results is analyzed. The results of building extraction are compared with ground truth data. Finally, the changes of buildings in different periods are gained. The JL1-3B night time glimmer images are used to analyze the buildings around the school and the activities of residents. The experimental results show that buildings in the remote sensing images can be effectively discerned by random forest algorithm and VGG(Visual Geometry Group) neural network algorithm. The number of buildings in different periods and the results of change detection of lamp light show the influence of campus on the development of surrounding area and provide reference information for city planning.

The traditional mobile robot based on DWA ( Dynamic window Approach ) algorithm exists the following deficiencies: longer obstacle avoidance time and the inability to optimize the local path planning in obstacle-intensive dynamic zone. Aimed at the problems mentioned above, a hybrid path algorithm combined A * algorithm with VO(Velocity Obstacle) is proposed to optimize the velocity of obstacle avoidance for mobile robots. Via the comparative experiment combined the DWA algorithm with the VO hybrid path algorithm in the case of three obstacles, the ROS (Robot Operating System) adopted the modular software design is put into practice to test the obstacle avoidance effect of the hybrid path planning algorithm. The results of simulation experiment in multiple environments clearly indicate that the obstacle avoidance effect will be significantly improved via the VO hybrid path algorithm in the scenarios scattered with multiple dynamic obstacles, and it has high speed of the obstacle movement and low frequency of radar scanning.