Feature extraction is a troublesome problem in the pipe signal degrading the classification accuracy. To address this problem, a pipe signal diagnosis method that combines the signal processing method with the intelligence algorithm is proposed. Firstly, CEEMDAN(Complete Ensemble Empirical Mode Decomposition with Adaptive Noise) algorithm is used to decompose the signal to obtain several IMFs (Intrinsic Mode Functions) and the correlation coefficient method is used to select the useful mode function components and recombine them. Then the Lempel-Ziv complexity and Margin of the reconstructed signal are calculated as feature vector. Finally, the feature vector are inputted into the ELM ( Extreme Learning Machine ) optimized by SOA ( Seagull Optimization Algorithm) for classification. And validation is performed with laboratory data. Experimental results show that comparing with conventional ELM and GA-ELM(Extreme Learning Machine Optimized by Genetic Algorithm). SOA-ELM model can identify the pipe signals effectively, and has higher recognition rate and faster diagnosis speed. 

It is difficult to select the preset scale K for VMD(Variational Mode Decomposition) and to distinguish the effective mode from the noise mode after decomposition. In order to solve the problem, a joint criterion method (VMD-HD-KT) is proposed to determine the presetscale K by HD(Hausdorff Distance) and the effective mode by Kendall correlation coefficient KT(Kendall ’s Tau). And it is used to denoise the leakage signal of natural gas pipeline. First, the HD of the last mode and the original signal when K = 2 to 8 is calculated, K is determined by evaluating HD, and then K value is input for VMD decomposition. The original signal is decomposed into K IMF(Intrinsic Mode Functions) with different characteristic time scales. IMF with KKT greater than 0. 1 is selected as the effective mode for signal reconstruction. The experimental results show that the VMD- HD-KT algorithm can accurately select the preset scale K and effective modes, and has a good denoising effect on the simulation signals and pipeline leakage signals. 

Demand side response is an important means of active distribution network optimization scheduling. Aiming at the problem of poor energy scheduling in power grid operation, firstly, based on the uncertainty characteristics of demand side response, introducing non-economic factors and characteristics of consumer psychology, the active distribution network optimization is modeled with the minimum power grid operation cost and environmental cost as the objective function; secondly, aiming at the premature problem of sparrow algorithm, latin hypercube sampling is used to improve the initial population quality, sine factor is introduced to improve the local search ability of the algorithm, and mutation operation is implemented to optimize the global search accuracy of the algorithm; finally, the improved sparrow search algorithm is applied to the solution of the active power grid optimization model. The simulation results verify the accuracy of the proposed model and the efficiency of the algorithm, and effectively solve the problem of poor energy scheduling. 

In order to improve the accuracy and robustness of SAR ( Synthetic Aperture Radar) image change detection, an unsupervised SAR image change detection method based on image enhancement and fusion is proposed. In order to obtain better effects of background noise suppression, change region enhancement and edge preservation, the log-ratio and mean-ratio differential image are constructed based on the adaptive image enhancement of the original SAR image. The differential image is fused by the fusion strategy of weighted average of low-frequency wavelet coefficients and selecting high-frequency wavelet coefficients according to the minimum local energy. The experimental results show that the fused differential image combined with fuzzy local information C-means clustering has achieved high detection accuracy and kappa coefficient on different data sets, and has strong robustness. It can be widely used in the field of SAR image change detection. 

In order to ensure economic and reliable operation of micro power grid, reduce the pollution to the environment, an optimal dispatching model for microgrid is proposed, which considered the operating cost and environmental pollution of microgrid, through the adaptive weighting method weights for different fitness function are distributed, the multi-objective problem is changed to single objective problem. The improved gull algorithm is used to find the optimal configuration scheme. Gulls experiment results show that the improved algorithm has better ability for global optimization, the solving precision and convergence speed compared with the standard algorithm. The micro grid economic operation has certain advantages in solving the problem, and reducing the micro grid operation cost and the environmental pollution, improving the reliability of the micro grid operation.

OBC(Ocean Bottom Cable) records are often limited by the limitations of data collection techniques and the environment, leading to a large amount of random noise that negatively impacts the identification of effective reflection information. Suppressing random noise requires accurate analysis of noise characteristics, therefore the multichannel method and a stability test method based on the energy distribution of the sequence are used to analyze the power spectral characteristics and stability of OBC random noise. Real noise data collected in a certain location in the South China Sea is used as the analysis dataset. The results show that OBC random noise is a colored noise sequence with noise energy mainly concentrated in the low-frequency part. The OBC noise is a weak non-stationary time sequence, and the conclusion is reasonable by combining the characteristics of the ocean exploration environment. This study has a certain practical significance and application prospects for the basic research on OBC random noise characteristics. 

 In order to explore the driving method of LED ( Light Emitting Diode), to improve the stability of LED output brightness, a LLC resonant converter controlled by frequency conversion to control the constant current output is proposed. LLC half-bridge resonant converter has the advantages of high switching frequency, power-off retention, wide allowable input voltage range, high efficiency, light weight, small size, low EMI (Electromagnetic Interference) noise, low switching stress, etc. Therefore, LLC half-bridge resonant converter is a good topology choice for LED constant current output. And the frequency conversion control method is adopted and the simulation analysis is carried out with PSIM(Power Simulation) software. The simulation results show that the LLC resonant converter controlled by frequency conversion has good steady-state performance and transient performance, which also has far-reaching significance for engineering applications. 

 Aiming at the problem that the classical deep learning method has low accuracy in multi-step load forecasting, a short-term load forecasting model based on double attention sequence to sequence is proposed. Through the self-attention mechanism, the hidden related factors affecting the load data are effectively extracted, so that the model can better find the laws between the load data, adaptively learn the related characteristics between the load data, and the temporal-attention mechanism captures the time-related time-series characteristics. Through two actual load data experiments, the simulation results show that under the condition of (t+12) prediction, the model evaluation index MAPE(Mean Absolute Percentage Error) is 2. 09% , which is 56. 69% lower than that of LSTM(Long Short-Term Memory) model. The validity and feasibility of the model are verified. The prediction effect of the model is better than that of linear regression, LSTM model and Seq2Seq (Sequence to Sequence) model.

Based on a portable transient electromagnetic system, we use a dipole model with a differential evolutionary algorithm to investigate the effects of target size, depth and inclination angle on the accuracy of its characterization, using the electromagnetic characteristics of a calibrated unexploded bomb as a reference. The experimental results show that for a target with an inclination angle of 0毅, the characteristic response is more concentrated in the head of the target and the result is smaller than the calibration value, when for an inclination angle of 90°, the characteristic response is the sum of the characteristic responses of all parts of the target, so the characteristic response increases with the increase of the inclination angle. The larger the depth, the higher the applicability of the dipole model, and the closer the characteristic response is to the calibration value. The larger the depth, the higher the applicability of the dipole model, the closer the characteristic response to the calibration value, and the better the consistency of the inversion results under different inclination angles. The larger the size of the target, the more the target is affected by the change of inclination angle, the better the consistency of the characteristic response from all angles of the small-size target, and the better it matches the calibration results. In summary, factors such as size, depth and inclination angle lead to errors in the results of the feature response, but in general do not affect the classification of targets based on the dipole model. 

 The electromagnetic method has a good response to harmful unexploded ordnance and harmless metal targets, so the false positive detection rate is high, resulting in the subsequent cleaning work is extremely time- consuming. To solve this problem, portable and towed transient electromagnetic detection systems is used to detect multiple unexploded bombs and harmless targets, and estimate the characteristic response of underground targets. According to the estimated electromagnetic characteristics, an accurate classification of unexploded ordnance and harmless targets is achieved based on the SVM( Support Vector Machine) algorithm, and the influence of noise on the classification results is discussed. The results show that the classification model trained by the characteristic responses at different times and the fitting parameters of target response can recognize and classify the targets effectively, the accuracy of target classification has reached 100% , and 59 targets have been recognized successfully in the actual verification. In contrast, the classification method based on characteristic response has fast calculation and simple way of processing, while the classification method based on fitting parameters has strong anti-interference ability and higher accuracy. 

 In the field of electromagnetic detection, skin depth is an important parameter for electromagnetic data analysis and imaging. In practice, the induced field and the polarization field coexist. If the polarization effect of the medium is not considered, there will be obvious errors in the imaging results. In order to solve the above problems, the generalized skin depth formula of the GEMTIP(Generalized Effective-Medium Theory of Induced Polarization) model in the frequency domain is deduced based on the plane wave theory and the GEMTIP model. The accuracy of the generalized skin depth of the GEMTIP model is verified by comparison with the classical skin depth. The generalized skin depth calculation of the GEMTIP model is mainly related to the resistivity and volume fraction. The BP(Back Propagation) neural network inversion method is used to extract parameters. And by constructing a reasonable data sample set, the training error can meet the accuracy requirements, and the mapping relationship between the input and output data is obtained. Several typical three-layer geological model structures are discussed. When the polarization effect is considered, the generalized skin depth formula of the GEMTIP model is verified to improve the identification accuracy of the underground polarized medium. 

 In vehicle recognition, due to different shooting angles and distances, the size of the imaged vehicle is smaller and the vehicle has different degrees of occlusion, resulting in detection error and missed detection. In order to solve this problem, based on the single stage target detection network YOLOv4(You Only Look Once version 4) algorithm, a recursive YOLOv4 target detection algorithm is proposed based on attention mechanism, namely RC-YOLOv4 algorithm. In order to improve the detection capability of the algorithm for small size vehicles after imaging, the CBAM ( Convolutional Block Attention Module ) module is added to YOLOv4 algorithm. This module combines the channel and spatial attention mechanism, which can help the network model pay more attention to the key information and small target information in the detected image. For the detection of partial occlusion of vehicles, a RFP(Recursive Feature Pyramid) structure is adopted to enhance the model’s ability to extract deep feature information. The RFP structure is similar to the human visual perception that selectively enhances or inhibits the activation of neurons. The features extracted from the backbone network are recursively fused and then fed back to the backbone network. Multiple feature fusion improves the network’s ability to extract and integrate contextual semantic information. It improves the detection accuracy of occluded vehicles. The experimental results show that the average precision of RC-YOLOv4(Recursive and CBAM You Only Look Once version 4 ) algorithm is 12. 69% higher than YOLOv4 algorithm on the self-made vehicle detection data set, and the detection speed can also meet the real-time requirements.

The improvement of consensus mechanism is a key research content in the development process of blockchain technology. In the traditional consensus mechanism, all endorsement nodes participate in endorsement, which consumes a lot of time, and has the possibility of forging and manipulating the consensus process with low security. Based on the verifiable random function, the endorsement nodes in the candidate set of endorsement nodes for trading and any endorsement node for endorsement operation are randomly selected which can effectively improve the processing efficiency and reduce the processing time of the consensus mechanism. Based on theoretical analysis and experimental verification of Hyperledger fabric model, the results show that the optimized consensus mechanism has faster transaction processing speed, lower delay time and higher security.

 Current material identification methods are mainly based on X-ray technology, ultrasound technology and radio frequency technology. However, X-ray technology relies on special equipment to transmit high frequency signals and is highly radioactive; ultrasonic technology equipment is bulky and inconvenient to carry; and RF(Radio Frequency) technology mainly relies on costly RFID(Radio Frequency Identification Devices). In order to meet the daily use in home and office scenarios, MateFi system for material identification is proposed based on WiFi(Wireless Fidelity), aiming to establish a new theoretical model to describe more specifically the attenuation state of electromagnetic waves as they penetrate different materials. The theoretical model is used to build a more robust and accurate material recognition system by combining material characteristics with machine learning techniques. The performance of the MateFi system is tested and validated in real-life scenarios. The experiments show that MateFi can recognise six different materials: wood, cardboard, nickel, thin wood, iron and titanium, with an average recognition accuracy of 96. 70% , demonstrating the system’s ability to identify materials accurately. 

Extractive text summarization is a method of extracting key text fragments from the input text to serve as the summary. In order to solve the problem of requiring sentence-level labels during training, extractive text summarization is modeled as a Q-Learning problem and DQN(Deep Q-Network) to learn the Q value function. The document representation method is crucial for the quality of the generated summarization. To effectively represent the document, we adopt a hierarchical document representation method, which uses Bidirectional Encoder Representations from Transformers to obtain sentence-level vector representation and uses Transformer to obtain document-level vector representation. The decoder considers the sentence information enrichment, saliency, position, and redundancy degree between a sentence and the current summarization. This method does not require sentence-level labels when extracting sentences, which significantly reduces workload. Experiments on CNN( Cable News Network) / DailyMail data sets show that, compared with other extraction models, this model achieves the best Rouge-L(38. 35) and comparable Rouge-1(42. 07) and Rouge-2(18. 32) performance.

 In order to explore the ability of imaging to diagnose benign and malignant breast lesions, and compare the value of MR(Magnetic Resonance) radiomics and traditional MRI(Magnetic Resonance Imaging) in diagnosing breast diseases, a total of 190 cases with benign or malignant breast lesions confirmed by pathological findings are collected from patients who underwent MR Plain and enhanced examination in the Department of Radiology in First Hospital of Jilin University from January 2019 to January 2022. MR radiomics is performed by building logistic regression model. The traditional MR Diagnosis is performed by a radiologist with an associate senior title. The results show that the sensitivity, specificity and AUC (Area Under Curve) of the MR radiomics test set are 0. 92, 0. 83 and 0. 92 respectively. The above values are higher than the corresponding values of traditional MR diagnosis, and the differences are statistically significant ( P = 0. 00 ). The method of MR radiomics can assist in the diagnosis of benign and malignant breast lesions, and the diagnostic ability is better than the traditional MR diagnostic mode.

A CoMaCOA(Co-evolution Multi-Objective Constrained) optimization algorithm is proposed to deal with the problem that it cannot be combined convergence and diversity effectively in multi-objective COA (Constrained Optimization Algorithms). First, a COA is transformed into the multi-objective evolutionary algorithm with dynamic constraint processing. Then, DE(Differential Evolution) is used to generate the first population. The second population is generated by the known feasible solution in the first population and coevolved with the first. The first population accelerates convergence by global search that does not deal with constraints. The second population evolves through local search to maintain and obtain more feasible solutions. Finally, the standard constrained multi-objective test function is used for experiments in order to test the performance of the proposed algorithm. The experiment result shows that the proposed algorithm achieves good results on both IGD( Inverted Generational Distance) and HV( Hypervolume), comparing with PF ( Penalty Function) method and dynamic boundary processing to constrain problem DCMaOP(Dynamic Constrained Many Objective optimization Problem). It shows that the algorithm is both effective in convergence and diversity.

 In the existing researches on closed frequent item set mining algorithms, pruning strategies are relatively single, most of which are for 1item set pruning, and there are relatively few pruning strategies for 2item set and nitem set (n逸3). However, effective pruning strategies can find and cut off a large number of hopeless item sets in advance. Therefore, improving the pruning strategy of closed frequent item set is of great help to improve the efficiency of this kind of algorithm. On the basis of ESCS(Estimated Support Cooccurrence Structure) structure, an ESCS pruning strategy for 2itemsets is proposed, and the classical closed frequent itemset mining algorithm DCI_Closed(Direct Count Intersect Closed) is improved to DCI_ESCS(Direct Count Intersect Estimated Support Cooccurrence Structure) algorithm, and the effect of ESCS pruning strategy is verified. On multiple public datasets and under different minimum support thresholds, experiments are conducted to compare the time performance of the algorithm before and after the improvement. The experimental results show that the improved DCI_ESCS algorithm performs well on long and dense data sets with long transaction and itemsets, and the time efficiency is improved to a certain extent.

In view of the large scale networking of three-dimensional lightning detectors, the location data of reception and reconciliation have been greatly increased, and there is no software specially designed for monitoring, displaying and processing three-dimensional lightning data in China, therefore a three-dimensional lightning data graphic display system is developed using B / S(Browser/ Server) architecture, ArcGIS geographic information platform, Oracle, SQLite database and other technical methods. The state monitoring subsystem is designed by using state data receiving module, analytical input module, real-time state statistics module and other functional modules. The real-time display subsystem is designed by using the function modules of location data real-time statistics, historical inquiry and three-dimensional display, and the product display subsystem is composed of a product production subsystem with data service product generation function, and the system function and operation state are verified by the national three-dimensional lightning detection network as the practical application background. The results show that the three-dimensional lightning data graphic display system is intuitive, friendly, timely display, full-featured, and rich in products. It is a useful tool in the field of lightning research.

 In order to enhance the security factor of patients’ private information and reduce the risk of data leakage, a hierarchical encryption algorithm for massive medical information based on chaotic cellular neural network and wavelet transform was proposed under the premise of considering the importance of privacy. The word repetition rate of medical information is calculated by the matching degree of word segmentation and weight matching degree, and similar data in massive information is eliminated, so as to reduce the workload of subsequent information encryption. The importance of data privacy is evaluated by grades such as medical importance, number of visits, and data size, and medical text information and image information are distinguished by data attributes. Using the chaotic features of cellular neural networks, the original medical information is converted into a parameter matrix. Logistic mapping is used to obtain the key chaotic sequence, the medical text information after primary encryption is output, the image signal is analyzed in time domain by wavelet transform to achieve secondary encryption, and the result of secondary encryption is fused to complete the hierarchical encryption of medical information. The experimental results show that the proposed algorithm has the advantages of good encryption effect, fast execution speed and high security factor, and is a suitable solution for the safe storage of medical information.

 In the process of predicting the intrusion intention of abnormal information in the hospital network, there is no dimension reduction processing for the hospital network data, resulting in a long prediction time and a low prediction accuracy. Therefore, an algorithm for predicting the intrusion intention of abnormal information in the hospital network based on the improved RBF(Radical Basis Function) model is proposed. The redundancy of hospital network data is removed and sorted through correlation analysis, and the sorted data attributes are selected by RBF multilayer neural network to complete the dimensionality reduction of hospital network data. According to the data preprocessing results, the Bayesian attack graph is constructed to obtain the potential network intrusion attack path. The alarm correlation strength is calculated in the path, the intrusion alarm evidence data is extracted, the information intrusion probability is determined through the monitoring of the alarm evidence, and the prediction result of the abnormal information intrusion intention of the hospital network is obtained. The experimental results show that the proposed method has high efficiency, high accuracy and good overall effect.

 In order to improve the high definition of multi frame plane image and improve the quality of multi frame plane image, an detail enhancement algorithm of multi frame plane image texture based on visual communication is proposed. According to egdnet algorithm, the noise interference of multi frame plane image, strengthen the image edge information, supplement the image color information with color mapping equation is eliminated, completing color correction and ensuring the accuracy of color in visual communication. The image high-frequency information is extracted with Gaussian fuzzy algorithm, the super-resolution multi frame plane image is obtained through interpolation, and the super-resolution multi frame plane image is synthesized through windowing operation. And the texture detail enhancement of multi frame plane image is accomplished. The experimental results show that the image texture detail enhancement has higher definition and better image quality. 

Eliminating duplicate data is an indispensable step to ensure efficient network operation. But this process is susceptible to interference from signal strength, network device, router performance and other problems. Therefore, a minimum-hashing algorithm for single channel data elimination is proposed. First the hash function in the hash algorithm network is used for single channel data clustering, and then supervision discriminant projection algorithm is applied for clustering of data dimension reduction after processing, finally the algebraic sign estimate is used to guarantee the data between the computing cost minimum and to construct minimum hash tree generated calibration value, in the update to heavy tags. The repeated data in a single channel is completely eliminated by double-layer culling mechanism. Experimental results show that the algorithm has short execution time and low computation and storage cost.

 In order to facilitate the testing of students’ English language application ability, many schools hope to apply computers to language testing, but there are doubts about the effect of computer-assisted language testing. In view of this situation, a computer-aided language testing validity evaluation algorithm based on entropy weight method is studied. The validity evaluation index is selected by gray correlation analysis, and the evaluation index system is constructed. The entropy weight method is used to calculate the weight of each index. Through fuzzy comprehensive evaluation, the index weight and the index membership degree are fuzzy synthesized to obtain the validity evaluation score. The validity grade is obtained by referring to the principle of maximum membership degree. The results show that the validity of computer-aided language testing system in junior and senior high schools has reached a very high level, while in universities the validity has decreased, but it still reaches a high level, which shows that computers perform well in computer-aided language testing and have strong practicability.

 In order to solve the problems of high-performance cluster monitoring and management, such as system monitoring is restricted by time and place, which causes cluster administrators to be unable to find cluster abnormal situations in time and affects the running of the cluster system, the open function and message transmission mechanism of WeChat are used in combination with the cluster monitoring and management method of Linux (GNU/ Linux) operating system, a kind of simple and easy-to-use cluster monitoring and management system is developed. It is suitable for small and medium-sized clusters with the ability to expand easily. We mainly expound the system requirements, system framework and function design, technical framework and data flow, as well as the specific process of system deployment and development. At present, the system has been developed and applied in the cluster monitoring management of the institute and molecular physics of Jilin University, and has achieved good application results. The cluster administrator and users can monitor the cluster performance and job completion status through APP(Application) on the mobile phone without login system, so as to facilitate the follow-up work in time. Especially during the COVID-19 period, when the cluster access is not convenient, the implementation of this function has assisted the efficient scientific research work of the institute.