WANG Yuqi, XUE Jing. Analysis of Palantir’s Patent Portfolio and Technological Barriers in AI Intelligence Field[J]. Technology of IoT&AI, 2026, 58(7): 1-5.
As a leading global data intelligence company, Palantir Technologies’ patent portfolio system embodies the company’s technology strategy and layout logic in the core track of “Artificial Intelligence (AI)+defense” such as ontology modeling and real-time intelligent decision-making. This study is based on a domestic patent analysis platform to quantitatively analyze Palantir’s global patent quantity, geographical distribution, and International Patent Classification (IPC) concentration. It focuses on its two main defense application systems, Gotham and Maven, and selects core patents in four key technology areas: ontology modeling, AI real-time decision-making, target recognition and tracking, and zero trust security for in-depth deconstruction. By analyzing Palantir’s technical barrier formation mechanism, it provides reference for the patent layout, underlying technology research and development, scenario innovation, and patent strategic deployment of China’s defense intelligent technology.
ZHENG Yi. Current Status and Development Exploration of Artificial Intelligence Technology Application in Power Industry[J]. Technology of IoT&AI, 2026, 58(7): 6-12.
With the development of Artificial Intelligence (AI), the power industry is undergoing a revolutionary transformation. Through literature reviews, on-site visits, and case studies, the application of AI in the power industry has been summarized, and detailed explanations have been provided on the technical applications in smart grids, power equipment operation and maintenance, load forecasting, and fault diagnosis. The research found that AI technologies such as machine learning, deep learning, and computer vision have been widely applied in the planning, operation, and maintenance of power systems, effectively improving the efficiency and reliability of power grid operation. Through comparative analysis of typical cases at home and abroad, the problems encountered in the current AI application in the power industry have been summarized, such as uneven data quality, insufficient adaptability of algorithm models, increasing cybersecurity issues, and lack of talents. In light of technological development trends and industry demands, suggestions for the development of AI technology in the power industry have been proposed, such as strengthening data governance, promoting the standardization of algorithm models, improving cybersecurity guarantees, cultivating interdisciplinary talents, and establishing a platform for industry-university-research cooperation. The research shows that AI is a powerful driving force for promoting the high-quality development of the power industry, and has broad application prospects in areas such as intelligent operation and maintenance, precise scheduling, and green energy management.
ZHANG Jiaru, YAN Xuefeng. Intelligent Logistics Distribution System in E-commerce Supply Chain Based on Internet of Things[J]. Technology of IoT&AI, 2026, 58(7): 13-16.
To address the challenges of highly fluctuating e-commerce orders and high timeliness requirements, an intelligent logistics distribution system based on the Internet of Things is designed. The system adopts a three-tier “cloud-edge-end” architecture and achieves real-time path optimization through a dynamic cost function and an adaptive routing algorithm. Simulation results show that on peak promotion days, the average order fulfillment time of the system is reduced by about 31.6%, and the delivery cost per order is decreased by about 19.3%, which significantly improves the supply chain response speed and resource utilization.
CAI Jinling. Research on Device Security Protection and Intrusion Detection Mechanisms in Diverse and Heterogeneous Internet of Things Environment in Universities[J]. Technology of IoT&AI, 2026, 58(7): 17-20.
With the advancement of smart campus construction, the integration of Internet of Things devices in universities has become increasingly complex. Issues such as unclear security boundaries, delayed anomaly detection, and inefficient risk handling arise due to dispersed devices and diverse management entities. To address these challenges, a security protection and intrusion detection mechanism tailored to the heterogeneous Internet of Things environment in universities was developed. This mechanism optimizes device access, access control, anomaly detection, and risk management, forming a closed-loop process involving device profiling, tiered access, risk assessment, and policy coordination. Experimental results demonstrate that the proposed mechanism outperforms comparative solutions in blocking unauthorized access, intercepting privilege escalation, identifying anomalies, and coordinated risk handling, offering valuable insights for Internet of Things security management in higher education institutions.
HU Xingbin, YANG Zongze. Design and Research of Security Automatic Induction Door Control System for Warehousing Based on Internet of Things Technology[J]. Technology of IoT&AI, 2026, 58(7): 21-25.
In order to solve the problem of insufficient safety control and traffic efficiency in the field of warehousing and logistics, a set of automatic sensing door control system for warehousing based on Internet of Things technology is proposed, which comprehensively realizes three core functions: personnel identification, door opening and closing, and abnormal alarm. Actual deployment verification shows that the waiting time for passage has been reduced by about 36%, the alarm response time has been reduced by about 38%, and energy consumption has been reduced by about 20%.
TAO Zhongyun. Research on Application of Internet of Things Technology in Fault Early Warning of Power Transmission and Distribution Equipment[J]. Technology of IoT&AI, 2026, 58(7): 26-30.
To address the problem of lagging fault identification and difficulty in converting warning results into operation and maintenance instructions for power transmission and distribution equipment, a fault warning system based on the Internet of Things is constructed. The system completes multidimensional data real-time collection under the edge computing architecture, uses the Generative Adversary Network (GAN) for data enhancement, solves the problem of small sample fault identification, and establishes a hierarchical early warning response mechanism that is directly connected with the operation and maintenance process. The experimental results show that the system’s warning accuracy is 91.6%, the average warning advance is
138 h, the equipment failure rate is reduced by about 56.3% compared to the traditional inspection mode, and the average fault disposal time is reduced by about 70.8%, which verifies the effectiveness of the technical closed-loop of the constructed system.
ZHANG Qingning, YANG Yunfei, SUN Jilong, ZHANG Guoxiao. Application of Internet of Things and Wireless Sensing in Railway Karst Grouting Monitoring#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 31-34.
In response to the problems such as scattered monitoring parameters, delayed information transmission, and high difficulty in process control in railway karst grouting construction, the application of the Internet of Things and wireless sensing in grouting monitoring was studied. The demand characteristics and system composition of railway karst grouting monitoring were expounded, and the monitoring index system, wireless sensing implementation method, and platform function design were introduced. Combined with a case of railway karst foundation remediation project, the system deployment plan, monitoring process, and application effectiveness were analyzed. The research results show that the designed system can not only improve the real-time, continuity, and coordination of grouting construction monitoring, but also enhance the ability of abnormal identification and construction control.
ZHOU Haoyong, YANG Ke. Design of Informatized Monitoring System for Smart Industrial Parks Based on Internet of Things Sensing#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 35-39.
In order to enhance the risk monitoring and coordinated response capabilities of the chemical industrial park, an information monitoring system based on Internet of Things sensing was designed. The results show that the designed system can collect real-time data on hazardous sources, environment, personnel and vehicles, with an alarm accuracy rate of no less than 96.00%, and the average response time does not exceed 10.5 s.
JIN Weijun, QIAN Zhemin, ZHAO Wangwu, CHEN Shijie, YE Zhou. Energy Consumption Monitoring and Energy-Saving Control Technology for Electromechanical Systems Based on Internet of Things Perception#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 40-43.
A “cloud edge end” monitoring network based on Internet of Things perception was constructed to address the energy efficiency optimization challenges caused by strong nonlinearity, time-varying coupling, and data silos in electromechanical systems. Using a physical mechanism constrained grey box model and residual gradient tracking algorithm to locate and trace energy consumption anomalies. By combining nonlinear model predictive control with distributed game strategy, equipment adaptive speed regulation and system level collaborative load distribution can be achieved. The actual test results of the industrial park show that the energy consumption baseline prediction error of the model is controlled at 2.8%, the accuracy of anomaly localization is 96.2%, and the comprehensive energy efficiency of the system has been improved by 18.9%. This technology achieves closed-loop optimization of electromechanical systems from energy consumption monitoring to collaborative control, and has good industrial application value.
WU Xuhui. Transmission Delay Control of Internet of Things Nodes Under Edge Intelligent Scheduling[J]. Technology of IoT&AI, 2026, 58(7): 44-47.
To solve the problem of high computation offloading decision overhead and wireless channel competition coupling induced transmission delay jitter in Internet of Things nodes in edge intelligent scheduling scenarios, a three-level collaborative delay control architecture for cloud edge is designed. Experimental results show that the proposed architecture suppresses the end-to-end transmission delay of hard real-time nodes to 9.7 ms under 90% edge gateway load conditions, and reduces the peak delay of burst traffic by 59.2% compared to static priority scheduling, providing a feasible technical solution for deterministic transmission in time-sensitive Internet of Things.
WU Chentong. Adaptive Sorting and Recognition of Complex Radar Signals Based on Multi-dimensional Intra-pulse and Inter-pulse Features[J]. Technology of IoT&AI, 2026, 58(7): 48-53.
Due to the diverse modulation forms within complex radar signals and the complex variation patterns of pulse repetition intervals between pulses, traditional recognition methods based on a single feature are difficult to balance accuracy and robustness. To this end, this article proposes an adaptive sorting and recognition method based on a multidimensional intra pulse inter pulse feature space, and establishes a recognition framework of “intra pulse coarse classification inter pulse pattern recognition joint decision verification”. The experimental results show that this method has achieved performance improvement in both intra pulse and inter pulse form recognition, especially showing strong robustness under inter pulse missed detection conditions. The recognition accuracy still reaches 96.97% when the missed detection rate is 20%. The research results have verified the feasibility of the unified modeling approach between intra pulse and inter pulse systems.
OU Xianwen, LI Jie, ZHAO Lin. Research on Neural Fuzzing Test Method Based on Trusted Gradient[J]. Technology of IoT&AI, 2026, 58(7): 54-57.
Coverage-guided greybox fuzzing modifies test inputs based on runtime feedback, aiming to exercise as many program paths as possible and thereby uncover potential software defects and security vulnerabilities. Traditional random mutation is inefficient when handling magic-value comparisons, deep conditions, and structured inputs. To this end, TrustGrad Fuzz is proposed, which introduces confidence estimation into neural program smoothing, combines predictive confidence with gradient magnitude to construct a trusted-gradient score, and adaptively switches among gradient mutation, random fallback, and incremental training. Experimental results show that TrustGrad-Fuzz achieves 17 170 new edge coverages on 7 real-world programs, outperforming AFL, AFLFast, and VUzzer, which verifies the effectiveness of trusted gradients in improving path exploration.
LIU Sujun, LI Changsheng, SHAN Qingqing, DING Jiayu. Application of ARIMA, Prophet and LSTM Models in Yellow River Water and Sediment Data Prediction[J]. Technology of IoT&AI, 2026, 58(7): 58-66.
Yellow River water and sediment data present characteristics such as nonlinearity, periodicity and volatility, making it difficult for traditional prediction methods to balance prediction accuracy and model adaptability effectively. This study adopts the official water and sediment monitoring data of the Yellow River provided for problem E of the 2023 National College Students Mathematical Modeling Contest, and applies Autoregressive Integrated Moving Average (ARIMA), Prophet and Long Short-Term Memory (LSTM) models to predict water and sediment concentration. Firstly, the raw data are preprocessed to eliminate noise and fill missing values. Secondly, four indicators including Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE) and coefficient of determination are used to objectively evaluate the predicting performance of the three models. On this basis, the applicable scenarios, strengths and weaknesses of each model are analyzed. Finally, the limitations of this research are summarized, and directions for future improvement are proposed. The results show that the LSTM model achieves the best performance in the prediction of Yellow River water and sediment data. The Prophet model delivers satisfactory results in terms of accuracy and usability, while the ARIMA model has distinct advantages in linear fitting.
ZHANG Zijian. Pressure Vessel Weld Defect Recognition Strategy Based on Multi-modal Sensing#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 67-70.
In order to improve the accuracy and stability of weld defect identification of pressure vessels, the multi-modal sensing identification strategy was studied. A visual, ultrasonic, acoustic emission and infrared thermal image synchronous acquisition system is constructed. Timestamp alignment, physical feature extraction, sensitivity initialization and attention fusion methods are used, and weighted cross entropy is used to deal with sample imbalance. The results show that the accuracy of the fusion model is 94.05 %, the F1 value is 95.55 %, the false detection rate and the missed detection rate are reduced to
2.70 % and 3.24 % respectively, and the multi-source mutual verification enhances the ability of complex defect discrimination.
YAN An, YIN Hesong, HOU Wenhua, YUAN Zhigang, MA Chao. Defect Recognition Technology for Flat Steel Welds of Main Grounding Grid Based on Machine Vision[J]. Technology of IoT&AI, 2026, 58(7): 71-76.
The flat steel welds of the main grounding grid are exposed to soil corrosion for a long time, making it difficult to identify surface defects such as cracks, open surface pores and incomplete fusion at weld toes against rusted backgrounds. Traditional image detection methods perform poorly on small slender cracks and suffer from a high false detection rate. To address these issues, an improved lightweight object detection model is proposed. A P2 small-object detection enhancement layer is added at the output of the neck network, and deformable convolution is adopted to adapt to the irregular geometric shapes of cracks. An edge feature constraint module is constructed to suppress the interference of rust textures via saliency weighting on feature map edges and regular terms of the loss function. Experiments are conducted on a dataset consisting of
2 480 field-collected images of flat steel welds. The results show that the proposed method achieves an mAP@0.5 of 91.7% and a recall rate of 89.3% for crack detection, with the false alarm rate reduced to 3.8% and the average inference time of 28 ms. It can meet the requirements of real-time monitoring.
HAN Mei. Research on Early Warning Method for Abnormal Operation of Intelligent Internet of Things Equipment Based on CNN-LSTM[J]. Technology of IoT&AI, 2026, 58(7): 77-81.
Aiming at the problems of multi-source coupling of operation data, complex abnormal patterns, and insufficient warning accuracy of intelligent Internet of Things devices, a running abnormal warning method based on Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) is proposed. This method uses multi-source running data preprocessing and sample construction, extracts local abnormal features using Convolutional Neural Network (CNN), models temporal evolution laws using Long Short Term Memory (LSTM), and outputs warning results through fully connected mapping. The experimental results show that the proposed method can improve the accuracy of anomaly recognition and the stability of early warning, providing an effective method for monitoring the operation status of intelligent Internet of Things devices.
YANG Hai. Research on Intelligent Topology Control Algorithm for Unmanned Cluster Ad Hoc Networks Oriented to Energy Balance#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 82-87.
In response to the energy imbalance and topology oscillation problems in unmanned cluster ad hoc networks, this article adopts energy balance modeling and distributed intelligent topology control methods to study the link reconstruction problem under local state perception conditions. The experimental results show that in the scenarios of 30, 50, 80 nodes, the standard deviations of remaining energy are 0.091, 0.108, 0.126, respectively, and the connectivity retention rates reach 98.3%, 97.9%, 96.8%, respectively. The network lifetime and latency indicators are better than the comparative algorithms.
ZHANG Yaoqi, QIAN Chengliang, ZHANG Xuan. Dynamic Task Planning and Simulation Optimization for Intelligent Warehouse Robots Based on Multi-agent Collaboration#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 88-92.
To address the bottlenecks of dynamic task allocation, conflict avoidance, and efficiency degradation in high-density environments during multi-agent collaborative operations of Automated Guided Vehicle/Autonomous Mobile Robot (AGV/AMR) in intelligent warehouses, a state-action-task queue model is constructed to model agent states and task priorities. Dynamic task allocation is realized by integrating reinforcement learning and auction mechanisms, and local-global collaborative scheduling is introduced to optimize the execution process.Verification is carried out in the ROS/Gazebo simulation environment under scenarios of multi-task density and emergent tasks. The results show that the proposed method can effectively reduce task completion time, improve system throughput and achieve load balancing, with the conflict probability controlled within 2.0%. The system maintains stability and reliability under dynamic adjustment conditions and has engineering application value.
MOU Yan. Research on Low-Altitude Aerial Image Clarity Processing of Small Unmanned Aerial Vehicle[J]. Technology of IoT&AI, 2026, 58(7): 93-96.
To improve the clarity and interpretation stability of low-altitude aerial images captured by small unmanned aerial vehicles in complex environments, this study analyzes the issues of motion blur, haze blur, sensor noise, and illumination interference in low-altitude operations at altitudes of 50～300 m. Key technologies are elaborated, including deblurring via multi-sensor and Inertial Measurement Unit (IMU) data fusion, deep learning-based image restoration, embedded real-time sharpening, and detail recovery. Application test results show that the Peak Signal-to-Noise Ratio (PSNR) of processed images exceeds 30.00 dB, and the Structural Similarity (SSIM) is higher than 0.880. The effects of edge preservation, local contrast enhancement, and dark-region noise suppression are relatively stable, which can enhance the application value of imagery in low-altitude inspection, mapping and identification, and security monitoring.
LIU Chunhui, XIAO Man, JIN Hong. Research on AI Assisted Fault Recognition Technology for Complex Electronic Products#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 97-100.
In response to the problems existing in the identification of complex electronic products, an application research on complex electronic product fault identification technology assisted by Artificial Intelligence (AI) was carried out. Mature methods such as wavelet threshold denoising, Fast Fourier Transform (FFT), Convolutional Neural Network (CNN), and Support Vector Machine (SVM) fusion models were combined and applied to construct a complete technical application link from fault data collection to feedback of identification results. The model training and preliminary verification were completed using artificially simulated fault data. The experimental results show that the proposed scheme has better identification accuracy and efficiency in simulated fault scenarios, and can provide a reference for the intelligent practice of fault diagnosis for complex electronic products.
QIAN Qiurong. Construction of Digital Teaching Support System for Vocational School Teachers Based on AI Empowerment[J]. Technology of IoT&AI, 2026, 58(7): 101-105.
To enhance the digital support capabilities of vocational school teachers in lesson preparation, teaching, evaluation, and answering question, a teaching support system based on Artificial Intelligence (AI) technology is constructed. The system is designed around functions such as course resource organization, classroom behavior analysis, dynamic diagnosis of learning situation, intelligent answer, and permission management, and three majors including information technology, e-commerce, and mechanical processing are selected for testing. Analysis suggests that the designed system maintains high accuracy in lesson preparation resource recommendation, classroom behavior recognition, learning situation diagnosis and matching, and question answering accuracy. It can effectively improve problems such as scattered teaching data, insufficient resource adaptation, and feedback lag, providing a feasible technical path for digital teaching for vocational school teachers.
LONG Zhenwei. Research on Architecture Evolution and Cognitive Interaction Mechanism of Intelligent Information System Driven by Large Model[J]. Technology of IoT&AI, 2026, 58(7): 106-109.
To address the issues of scattered data, isolated services, and insufficient task closure in the smart Internet of Things operation and maintenance scenarios, a large-model-driven intelligent information system architecture was constructed. The system integrates semantic reconfiguration, pre-arranged permissions, intent recognition, and task re-writing mechanisms, and conducted comparative experiments in a cloud server environment of 2 units, 6 edge nodes, and 30 terminals. The experimental results show that the enhanced system performs better in terms of response latency, throughput, task completion rate, and abnormal rollback control.
LU Zaibin. Fault Early Warning and Handling in Electrical Automation Based on Multimodal Visual Fusion[J]. Technology of IoT&AI, 2026, 58(7): 110-113.
To enhance the accuracy and real-time performance of fault early warning and response in electrical automation systems, this paper adopts a multimodal vision fusion approach. Integrating visible light images, infrared thermal images, electrical parameters, vibration signals, Programmable Logic Controller (PLC) data and Supervisory Control And Data Acquisition (SCADA) data, it researches the processes of fault feature extraction, pattern recognition, root cause diagnosis, hierarchical decision-making and automatic execution. The results demonstrate that the proposed method can strengthen equipment state perception, improve fault localization accuracy and shorten the response time for fault handling, which provides technical support for the intelligent operation and maintenance of electrical automation systems.
YANG Chuhua, MEI Bing. Application of Artificial Intelligence Technology in Information Management Platform of Water Conservancy Project#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 114-117.
To address the problems of data fragmentation, delayed risk identification and insufficient dispatching linkage existing in the information management platform for water conservancy project, an application system supported by Artificial Intelligence (AI) technology is constructed. Dynamic assessment of engineering conditions is realized through multi-source fusion of hydrological and structural data, comprehensive risk scoring, health index evaluation of electromechanical equipment and closed-loop disposal of dispatching instructions. The operation results of practical cases show that after the platform deployment, the response time for personnel violations is reduced to 3.8 min, the accuracy of structural risk early warning rises to 91.5%, and the automatic recognition rate of equipment faults reaches 88.6%. The proposed method can provide technical support for refined operation and management of hydraulic engineering.
LU Qing. Research on Authentication Mechanism of Internet of Things Edge Nodes for Smart Human Resources[J]. Technology of IoT&AI, 2026, 58(7): 118-121.
In response to the security threats such as identity forgery, man-in-the-middle attacks, and illegal access faced by the Internet of Things edge nodes in the smart human resources system, a lightweight edge node identity authentication mechanism is proposed. This mechanism integrates the Physical Unclonable Function (PUF) and Elliptic Curve Cryptography (ECC) to construct a two-way authentication protocol suitable for resource-constrained edge devices in personnel scenarios. Additionally, a dynamic trust evaluation model based on reputation values is introduced to achieve the full-process identity trust verification for edge nodes such as attendance terminals and access sensors from entry authentication to continuous operation. Experimental results show that the single authentication delay in the standard test environment is 11.6 ms, and in the concurrent scenario of 800 people for attendance peak, the average delay is 14.3 ms, the average abnormal recognition time is
4.7 s, and the abnormal detection rate is 100%. It can effectively resist replay attacks, man-in-the-middle attacks, cloning attacks, and Denial of Service (DoS) attacks, providing a feasible technical path for the secure access of a large number of heterogeneous edge devices in the smart human resources system.
SONG Jiaqi, ZHENG Dan, HE Keyi. Intelligent Fault Perception Method for Meteorological Detection Equipment Under Integration of Internet of Things and Edge Computing[J]. Technology of IoT&AI, 2026, 58(7): 122-126.
Aiming at the frequent failures of meteorological detection equipment in complex environments, the problems of large communication delay, high bandwidth load and insufficient real-time response capability in the traditional cloud-based centralized processing mode. To address the existing problems, an intelligent fault perception method for meteorological detection equipment integrating the Internet of Things and edge computing is proposed. This method designs an adaptive filtering and feature extraction mechanism based on a sliding time window on the edge side. It uses the Exponential Weighted Moving-Average (EWMA) algorithm and lightweight Fast Fourier Transform (FFT) to extract time-frequency domain features, and achieves feature dimensionality reduction through the online streaming variant of Principal Component Analysis (PCA), effectively compressing the data volume. In terms of model construction, a cloud-edge collaborative architecture is adopted: a lightweight one-dimensional convolutional neural network is deployed on the edge side to achieve millisecond-level local fault discrimination. The cloud is responsible for incremental learning of difficult samples and model distillation updates, and dynamically updates the edge model through the Over-The-Air (OTA) channel. The experimental results show that the proposed method significantly reduces the core network bandwidth occupancy rate, and the end-to-end fault alarm delay is controlled at the millisecond level. It can meet the fault perception requirements of meteorological detection for high real-time performance and high reliability, and provide effective technical support for the intelligent operation and maintenance of meteorological equipment in complex environments.
YANG Bo, ZENG Jianhua, CHEN Jianxin. Safety Helmet Detection Algorithm Based on Improved YOLOv11n Model#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 127-131.
Aiming at the problems of insufficient feature extraction ability under complex backgrounds and high missed detection rate of small targets in construction site helmet wearing detection, a lightweight helmet detection algorithm based on the improved YOLOv11n model is proposed. A Coordinate Attention (CA) module is embedded at the end of the residual branch of the deep C3k2 module in the backbone network to construct the C3k2-CA module, enhancing the joint modeling ability of target spatial location and channel features. A new P2 high-resolution detection layer is added to make full use of shallow detail information and improve the detection accuracy of small-scale helmets. Experimental results show that the improved algorithm achieves an mAP@0.5 of 92.8% on the self-built dataset, which is 1.8 percentage points higher than the original algorithm, with only a slight increase in the number of parameters. The proposed algorithm effectively improves the detection accuracy in complex scenarios with only a small increase in parameters and computational overhead, achieving a better balance between accuracy and computational cost.
PAN Dinglin, TAO Yilin. Multi-dimensional Reputation Evaluation System for Brand Digital Management#br#[J]. Technology of IoT&AI, 2026, 58(7): 132-136.
Brand reputation is scattered across multi-source heterogeneous platforms and changes dynamically. Traditional single-dimensional static evaluation leads to fragmented results and delayed decision-making. A multi-dimensional reputation evaluation system is constructed, which improves the fuzzy analytic hierarchy process for dynamic weighting. The RoBERTa model and contrastive learning are used to enhance cross-platform entity fusion. The heterogeneous space is unified through an improved metapath2vec algorithm. Dynamic early warning is realized via incremental sentiment analysis, streaming clustering and time-decay isolation forest, and the GPT-2 model is employed to generate diagnostic reports. Experiments results show that the designed system achieves expected performance in terms of coverage integrity, sentiment accuracy, early warning timeliness, fusion consistency and management effectiveness, providing a multi-dimensional, real-time and operable reputation evaluation approach for brand digital management.
WANG Zhigang, YU Dongli. Computer Information Network Security Monitoring Technology Based on Artificial Intelligence Algorithms#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 137-140.
To address issues such as fragmented multi-source logs, insufficient identification of unknown threats, and delayed alarm responses, an artificial intelligence-based security monitoring system has been developed. The system integrates mirrored traffic data, firewall logs, host proxy records, and identity authentication data. After edge preprocessing that includes time calibration, field normalization, and session reconstruction, feature extraction generates fusion vectors. Threat detection is achieved through a combination of isolated forest algorithms, Bidirectional Long Short-Term Memory (BiLSTM) networks, and access chain scoring. Experimental results demonstrate that the fusion approach achieves a detection accuracy of 95.1%, an F1 score of 94.2%, an false positive rate of 2.7%, an average detection latency of 2.1 s, and an alarm response time reduced to
3.8 min, significantly enhancing real-time monitoring capabilities and response precision in complex attack scenarios.
LI Shuangyi, LIU Ping. Heterogeneous Network Integrated Communication Technology for Smart Parks[J]. Technology of IoT&AI, 2026, 58(7): 141-144.
To improve communication access and collaborative transmission capabilities in multi-service scenarios of smart parks, this study adopts layered architecture design, edge gateway adaptation, multi-link scheduling, and security control methods to investigate a heterogeneous network integrated communication technology system. The results show that the proposed system can realize unified access of multiple types of terminals, differentiated data transmission, and stable support for key services, providing technical support for the construction of smart park communication systems.
ZHOU Zhuoyang. Research on Access Security and Identity Authentication Mechanism for Internet of Things Devices in Smart Campuses#br#
#br#
[J]. Technology of IoT&AI, 2026, 58(7): 145-149.
Smart campus Internet of Things devices have strong heterogeneity and a large number, and their access security and identity authentication issues are prominent. Traditional authentication schemes are not suitable for resource constrained terminals, while lightweight protocols have security shortcomings. To this end, a hierarchical lightweight identity authentication mechanism HiLA is proposed, which constructs device fingerprints through hardware Physical Unclonable Function (PUF) and radio frequency features, combines Elliptic Curve Diffie Hellman (ECDH) and hash chains to achieve bidirectional lightweight authentication, and dynamically evaluates trust values based on device behavior. The actual test results show that this mechanism has low authentication latency, low overhead, can effectively resist various network attacks, controllable false rejection rate, and is suitable for smart campus Internet of Things scenarios.
XU Zhuangzhi, XU Sisi, PENG Yida. Research on Distributed Network Communication Traffic Anomaly Detection Based on Graph Neural Networks[J]. Technology of IoT&AI, 2026, 58(7): 150-153.
A method based on graph neural network is proposed to address the issues of weak adaptability and low accuracy in anomaly detection of communication traffic in distributed networks. Construct a network topology diagram, design a graph neural network that integrates attention mechanism to extract spatiotemporal features, and introduce reconstruction error as a discrimination mechanism to achieve efficient and accurate detection. The experimental results show that this method has significant advantages in accuracy and robustness, verifying its excellent performance and application potential in complex distributed network environments.
YANG Zhaomeng, LI Xianyuan, WANG Qin. Research on Generative AI-Driven Derivative Design of Anshun Miao Batik Costume Patterns[J]. Technology of IoT&AI, 2026, 58(7): 154-161.
As the material carrier of the ethnic culture in central Guizhou,the patterns on Anshun Miao batik clothing, with their representative style characteristics from different periods, can serve as a symbolic system to represent the social culture of the region. In response to the existential crisis of traditional crafts in the digital intelligence era, this study transcends the archival limitations of conventional intangible cultural heritage preservation and proposes a tripartite methodological framework of “decoding, translating, and reconstructing” for intelligent heritage transmission. Firstly, it analyzes the cultural coupling between Artificial Intelligence Generated Content (AIGC) and intangible cultural heritage products, and deconstructs the modular morphological features of Anshun Miao batik costume patterns. Secondly, it employs the Low-Rank Adaptation (LoRA) algorithm to achieve the decoupling of artistic style features. Finally, it develops an intelligent design framework based on style transfer. AIGC provides a reusable methodology for intangible cultural heritage preservation, while the LoRA model enables precise extraction and transfer of traditional pattern artistic styles. The proposed intelligent design framework carries dual value in cultural inheritance and innovation, offering a new cognitive schema for the living transmission of intangible cultural heritage.
ZHAI Jiajia, GAO Meng. Research on Energy-Saving Regulation Methods for Building Automation Based on Multi-source Sensing Data[J]. Technology of IoT&AI, 2026, 58(7): 162-166.
A multi-source sensor data-driven energy-saving control method for building automation is proposed to address issues such as frequent fixed time control, insufficient environmental perception, and ineffective equipment operation in building operation. This method collects data on temperature and humidity, light intensity, CO2 concentration, personnel activity status, and equipment energy consumption. After cleaning, completion, time synchronization, and normalization, a spatial operation status evaluation value is constructed, and air conditioning, lighting, and ventilation equipment are coordinated and adjusted. Based on the operation data of a typical area of an office building, comparative analysis showed that after regulation, the daily average electricity consumption decreased by about 15.61%, the air conditioning operation time decreased by about 17.14%, the ineffective lighting operation time has been shortened to 1.1 h, and the abnormal response time has been shortened to 8 min. The results indicate that this method can reduce ineffective operation of equipment and improve the level of energy-saving control in building automation.
ZHANG Xiaosheng. Research on Construction Termite Early Warning and Comprehensive Management Platform Based on Intelligent Monitoring[J]. Technology of IoT&AI, 2026, 58(7): 167-170.
To address the issues such as the concealment of termite activities in construction projects, the lag of manual inspections, and the difficulty in tracing the treatment process, an intelligent monitoring, early warning and comprehensive governance platform has been constructed. The platform collects temperature and humidity, soil moisture content, bait disturbance and micro-vibration data through sensing nodes, combines historical disposal records to establish a multi-source risk integration and level determination method, and links with the governance work order for closed-loop response. The engineering verification results show that the constructed platform reduces the response time for abnormal events from 12.6 min to 3.8 min, and the accuracy of component risk warning has increased to 91.5%. This platform can support the pre-identification of termite risks and the refined governance.
LI Bingxin, ZHANG Hao. Research on Corrosion Rate Prediction Technology for Oil and Gas Pipelines#br#[J]. Technology of IoT&AI, 2026, 58(7): 171-174.
Aiming at the problem of insufficient data discretization and prediction ability in the multi mechanism coupled corrosion environment of oil and gas pipelines, a big data corrosion monitoring method that integrates multi-source heterogeneous data is proposed. Through the spatiotemporal alignment and quality control of resistance probes, polarization data, and Supervisory Control And Data Acquisition (SCADA) operating parameters, key features are extracted, and Corrosion Rate (CR) prediction and risk warning are achieved based on random forest and Long Short-Term Memory (LSTM) models. The results show that the proposed method not only improves the prediction accuracy and output stability under complex working conditions, but also reduces the risk of misjudgment and missed judgment in corrosion warning, realizing the transformation from experience driven to data-driven, and providing reliable support for pipeline integrity management.
MIAO Huili, WANG Xianfeng, LIU Wenqing. Design of Wear State Recognition and Early Warning System of Elevator Traction Wheel Based on Edge Computing[J]. Technology of IoT&AI, 2026, 58(7): 175-178.
The traction wheel is an important component of the elevator transmission system. Its wear condition directly affects the normal operation and service life of the entire system. The traditional centralized detection method has a large delay and heavy network resource occupation, cannot timely determine the state of the traction wheel. Therefore, a method for identifying the wear state of the traction wheel and implementing graded alarms based on edge computing is proposed, and the effectiveness of this method is verified through experiments. The experimental results show that the proposed method has high recognition accuracy and fast recognition speed for different types of wear, which is conducive to improving the maintenance efficiency of elevators.
ZHENG Yongyi, LOU Hang, HE Li, PAN Longjiang, LUO Xiying. Low-Power Design of Data Acquisition and Remote Transmission Module for Transformer Oil Chromatography Online Monitoring Device[J]. Technology of IoT&AI, 2026, 58(7): 179-184.
In response to the problems such as high energy consumption during long-term operation of the online transformer oil chromatography monitoring device, including high data transmission pressure and insufficient node endurance capacity, a low-power data acquisition and remote transmission module design method is proposed. This method adopts pulse intelligent power supply control, adaptive wake-up scheduling, wavelet packet decomposition and feature entropy compression algorithms, and combines Long Range Radio (LoRa) adaptive remote transmission strategy and lightweight hash chain verification mechanism to optimize the entire process of acquisition, processing and transmission. Experimental results show that this method can effectively reduce the average power consumption of the system, decrease the number of invalid acquisitions, improve the data compression efficiency, and ensure the reliability of data transmission in complex environments. The designed data acquisition and remote transmission module can enhance the low-power operation capability of the transformer oil chromatography online monitoring device, providing technical support for the long-term stable monitoring of intelligent power equipment.
