To investigate the influence of incident and scattering angles on spherical contaminant detection on wafer surfaces， a study based on light scattering principles was conducted. The scattering field was calculated using Bobbert and Vlieger theory with Bidirectional reflectance distribution function （BRDF）. A detection system with adjustable angles was constructed to examine scattering intensity's angular dependence using 500 nm Polystyrene latex particles（PSL）. Experimental angles were limited to prevent reflection interference and detector damage. With collection angle at 20 °， the signal-to-noise ratio increased by 68.89% as incident angle decreased from 80 ° to 40 °. With incident angle at 70 °， the signal-to-noise ratio increased by 102.25% as collection angle increased from 0 ° to 50 °. The scattered field concentrated near the specular reflection angle， exhibiting higher detection sensitivity in this region. Root-mean-square errors between simulation and measurement were 0.073 1 （fixed scattering angle） and 0.094 2 （fixed incident angle）， validating the model's accuracy.
As dual-comb technology expands beyond the laboratory， its stable operation in complex temperature environments is increasingly critical. Traditional systems lock the repetition rate to a fixed reference； however， outdoor temperature fluctuations induce thermal effects in the laser cavity， causing continuous frequency drift. With a fixed reference， the phase-locked loop （PLL） must output large compensations， easily exceeding its range and causing the system to unlock. Moreover， traditional temperature controls lack the response speed needed for harsh environments. To address this， we propose a temperature-control-free digital phase-locking scheme based on variable reference frequency compensation. By locking the repetition rate to a reference that is dynamically adjusted in real-time based on loop feedback， this method suppresses temperature-induced drift over a wide range and prevents feedback saturation and unlocking. Experimental results demonstrate stable operation across a 20~40 ℃ range， achieving 8 days of continuous， phase-slip-free locking. The relative frequency stability reaches the 10-12 level at 1 second and the 10-17 level daily， offering a novel solution for the long-term stable operation of dual-comb systems in complex thermal environments.
To ensure the accuracy and consistency of the color temperature measurement system， the State Administration for Market Regulation organized a measurement comparison for national standards of color temperature in 2023. National Institute of Metrology（NIM） served as the pilot laboratory， while the National Institute of Measurement and Testing Technology（NIMTT） as a participating laboratory. Both laboratories used the national primary and secondary standards for color temperature to measure the 2 856 K color temperature point of two comparison lamps. The comparison results were evaluated using the normalized error E n value. After analysis， the absolute values of the E n values were 0.07 and 0.12， both much less than 1， indicating that the differences between the measurement results from the participating laboratory and the reference values were within reasonable expectations， and the comparison results were satisfactory. The results showed that the measurement results from the secondary standard for color temperature are consistent with those from the primary standard.
To improve the fidelity of the radio frequency/infrared （RF/IR） compound hardware-in-the-loop（HWIL） simulation system， an online calibration method for the photoelectric axis pointing of compound targets is proposed. Based on the measurement of the electromagnetic field of amplitude and phase distribution in the RF operating area and the optical axis pointing measurement method with infrared imaging， a RF/Infrared integrated compound target photoelectric axis pointing calibration device was developed. Tests and quantitative analysis of the photoelectric axis pointing measurement errors were conducted on-site in the RF/IR compound HWIL simulation system. The results show that the calibration device enables high-precision measurement and calibration of the photoelectric axes pointing. The uncertainty of the electric axis pointing measurement is 0.06°， while that of the optical axis pointing measurement is 0.56′. This calibration technology can be applied to the measurement and calibration of the photoelectric axes pointing of RF/Infrared compound hardware-in-the-loop simulation systems.
A precision optimization method for large-scale distance measurement based on BP neural network is proposed. A training set is constructed through the simulation generation model of ranging data. The strong nonlinear modeling capability of the multi-environmental parameter error compensation model is utilized to solve the coupling problem of environmental parameters. In experiments conducted on a 1.2 km standard baseline field， the error characteristics of long-distance 1.2 km and short-distance 72 m measurements are systematically analyzed. The experimental results show that for long-distance measurement， the model compensation reduces the mean error from 7.1 mm to 0.6 mm significantly， and the temperature correlation coefficient is improved from -0.99 to -0.03. For short-distance measurement， the mean error is optimized from 0.8 mm to 0.3 mm， and the temperature correlation coefficient is improved from -0.80 to 0.04. Multi-distance experimental verification shows that the average errors after compensation are all better than 0.5 mm， the error is reduced by 91.4% at 1 176 m compared with the traditional method， and the distance correlation is effectively eliminated. This method provides a reliable technical solution for high-precision field measurement.
Conventional static， segmented measurement methods suffer from splicing errors and data drift on ultra-long guide rails， making one-shot dynamic continuous measurement infeasible. To address this issue， an error-compensation method based on azimuth-angle integration correction and lateral-offset trajectory reconstruction is proposed. First， azimuth-angle signals throughout forward and reverse traversals are acquired by a high-precision gyroscope. Second， attack-angle and sideslip-angle disturbances are separated by a continuous-integration calibration strategy， directly yielding a continuous， unspliced lateral-offset trajectory. Third， the lateral-offset curve is subjected to a fast Fourier transform， and， based on the root-mean-square error criterion， low-order harmonics with large amplitudes and high error-reduction contributions are adaptively selected. Finally， the trajectory is reconstructed using the selected sensitive harmonic components to obtain the compensated offset. Repeated experiments at speeds of 0.1~0.9m/s show that the maximum difference between forward and reverse lateral offsets is reduced by an average of 63.3%. Straightness measurements of a 75 m guide rail are found to be （0.446 2±0.002 8）mm （horizontal） and （0.438 2±0.002 5）mm （vertical）， outperforming traditional autocollimator and laser-interferometry methods. These results demonstrate that high precision and high repeatability are offered by the proposed approach， providing a new technical pathway for straightness measurement of ultra-long guide rails.
A theoretical analysis and simulation study on the main error sources of a grating pitch measurement system are conducted based on a direct-tracing grating interferometer. The focus is placed on the influences of pitch error， yaw error， and roll error of the grating under test on the measurement results of grating pitch. Through theoretical analysis， experimental measurement， and uncertainty evaluation， the main error sources of the system are identified. A system model is established using optical simulation software to simulate the displacement of the grating along the vector direction， extract the coherent phase at the detector， and perform phase unwrapping and grating pitch calculation. By varying the rotation angles of the grating under test around the X-， Y-， and Z-axes， the effects on measurement accuracy are systematically analyzed. Simulation results show that when the grating under test has a pitch angle of 0.01°~0.10°， the relative error of the measured grating pitch is 0.12%； when a roll angle of 0.01°~0.10° exists， the relative error reaches 0.41%； and when a yaw angle of 0.01°~0.10° is present， the range of the measured grating pitch is up to 6.9 nm.
With the development of quantum precision measurement technology， the temperature measurement method of nitrogen-vacancy （NV） centers in diamond based on solid-state quantum spin is expected to break through the bottleneck of non-destructive microscopic temperature detection technology in biomedical and chip manufacturing fields， due to its merits of ultra-high precision， micro-nano scale， stable material properties and so on. However， the method to enhance the signal-to-noise ratio by applying high-power lasers or microwave lead to significant heating effects， thereby affecting the precision of temperature measurement. To avoid increasing the external field power， a method of optimizing the polarization direction of pumping laser is proposed to improve measurement performance. Firstly， the influence of laser polarization on the electron spin polarization ratio and fluorescence contrast is theoretically analyzed， and a light path which can adjust the polarization of pumping light is designed. Moreover， `experiments are conducted to investigate the variation of fluorescence contrast along the four NV axes as a function of laser polarization direction， the optimal polarization direction is selected for optimizing microwave power and temperature measurement sensitivity. Experimental results show that the proposed method effectively reduces the required microwave power， lowers thermal heating by ~7 °C and enhances temperature sensitivity over threefold.
In industrial field environments， traditional radiation thermometers must penetrate diffuse media or isolation windows to remotely measure the surface temperature of a target， which significantly affects the temperature measurement results. By adopting the active laser radiation temperature measurement technology， experiments are designed with quartz glass as the interfering medium to measure the true surface temperature under different transmittance conditions， and the temperature measurement performance differences between traditional radiation thermometers and the active temperature measurement system are compared and analyzed. Experimental results show that the average temperature measurement deviation of the active temperature measurement system is 3.6 K under interference of different transmittances. Under the same transmittance conditions， the average deviation of the traditional LP4 radiation thermometer is 3.7 K， and its deviation increases gradually with the rise of measured temperature， which verifies that the active system possesses better penetration capability than traditional radiation thermometers. The results indicate that the active temperature measurement system can maintain high measurement accuracy even when the emissivity of the measured sample is unknown， and is theoretically capable of eliminating the transmittance influence along the optical path.
A small-sample calibration strategy based on endpoint control of natural gas flow computers is proposed. A metrological database covering 533 mainstream flow computers from field measurements across China's major gas transmission networks in 2025 was constructed. The linearity of indication errors across the full scale was analyzed using generalized linear models. Based on this， a linear prediction model relying solely on the minimum flow point Q min and maximum flow point Q max was established. The interpolation error and uncertainty components introduced by the simplified model were quantitatively evaluated through Monte Carlo simulation and in accordance with JJF 1059.1—2012. Results show that 67.54% of the sampled devices exhibit extremely high linear correlation between indication error and standard flow， with a mean coefficient of determination R² of 0.99. Compared to the traditional five-point method， the interpolation error introduced at intermediate flow points remains below 0.02% at a 96% confidence level—significantly lower than the maximum permissible error. Both domestic and international mainstream equipment demonstrate consistent excellent linear characteristics. This small-sample calibration strategy significantly reduces on-site calibration time and natural gas loss while maintaining measurement accuracy and traceability， thereby enhancing the economic efficiency of pipeline metrological management.
To address the challenge of rapidly and accurately identifying interface transitions during oil product switching in pipeline transportation， a real-time detection method for pipeline oil switching interfaces based on sound velocity compensation is proposed. First， a standard-state sound velocity compensation model is constructed based on fundamental principles of sound propagation to suppress the influence of medium temperature， pressure， and other conditions on sound velocity measurements. Building upon this， a Kalman filter algorithm dynamically updates the baseline sound velocity， enabling adaptive matching to variations in oil batches and pipeline operating conditions. Finally， the cumulative sum （CUSUM） method is integrated to detect abrupt changes in the compensated sound velocity， establishing a high-precision oil interface detection model. To validate the method's effectiveness， field tests were conducted on typical diesel-gasoline interface transitions.Results showed that the compensated baseline velocity difference between diesel and gasoline ranged from 142.0~177.1 m/s. This method accurately identified the interface transitions in all four experimental sets， with results consistent with actual operating conditions.
To address the vibration and noise issues in electro-hydraulic modules caused by the drive of cycloidal gear pumps and fluid impact， a noise， vibration， and harshness （NVH） detection system was designed. Utilizing pneumatic control components and a programmable logic controller （PLC） based automation platform， an automated tooling mechanism was developed for product installation and automated sensor loading. Based on the working principle of the electro-hydraulic module， a hydraulic pipeline was designed to simulate oil flow. The oil flow and rate within the module were controlled via the hydraulic pipeline， a transmission control unit （TCU）， an ABB servo motor， and an Inovance electronic pump to simulate real-world vehicle operating conditions. Vibration sensor data and motor speed were acquired using an NI USB-4431 high-speed data acquisition card. The test environment and operating conditions were monitored in real-time， with test data being returned for analysis. Time-domain analysis was performed on the vibration signals. The pulse signals from the encoder cable were converted into rotational speed， which was then combined with the vibration signals for order analysis. Required analysis orders were extracted through order slicing. For two different operating conditions， the dataset consisted of 489 faulty components and 490 normal components for each condition. The data for each condition was split into training and testing sets in a 7∶3 ratio. Fault diagnosis was performed using the support vector machine （SVM） algorithm. The diagnosis accuracy for both operating conditions exceeded 98.98%， while the detection time was very short. Both the accuracy and real-time performance meet the requirements for enterprise online detection.
To optimize the pretension of the beryllium copper belt transmission system in a free-fall absolute gravimeter， a finite element model of the system was established using Abaqus. The model's validity was verified through a multi-faceted validation approach， including pretension verification and comparison of the displacement and velocity of the falling platform. The relative errors between the finite element results and experimental data for displacement and velocity were both less than 5.0%， confirming the model's accuracy. Based on this， a parametric simulation was conducted to analyze the variations in cumulative displacement relative error and system slip rate under different pretension levels. The results show that when the pretension is below 105 N， the slip rate exceeds 2.5%， indicating significant belt slippage； when the pretension exceeds 107.5 N， the slip rate stabilizes around 0.58%. However， as the pretension increases to 120 N， the cumulative displacement error rises， and the system becomes more sensitive to environmental disturbances due to increased stiffness. Considering both transmission accuracy and system rigidity， the optimal pretension was determined to be 108 N， achieving a balance between high precision and appropriate stiffness.
To solve the problems of insufficient feature extraction and low fault diagnosis accuracy in traditional rolling bearing fault diagnosis models， a bearing fault diagnosis method based on dual-domain feature fusion network is designed. Firstly， the one-dimensional vibration signal is transformed into two two-dimensional images with obvious features through the Gram angle field to improve the visualization ability of the signal； then it is input into the dual-path lightweight feature extraction module， which uses patch embedded image processing and introduces deep separable convolution to improve the computational efficiency； a feature fusion module is designed， and the information interaction between dual-domain features is dynamically adjusted by using the adaptive weight learning mechanism to improve the feature expression ability of the model. The experimental results show that the average fault recognition accuracy of the proposed model can reach 99.64%， and compared with traditional models such as CNN and ResNet-18， it can identify rolling bearing faults more efficiently.
To address the problems of relying on labeled data， high energy consumption of wireless sensor networks （WSNs） nodes， and large data transmission in bearing fault detection in industrial fields， an unsupervised bearing fault detection method based on convolutional spiking neural networks （CSNN） and sensor computing is proposed. The method utilizes CSNN， which effectively integrates the advantages of convolutional neural networks （CNN） and spiking neural networks （SNN） in local feature extraction and dynamic feature extraction， respectively. By using unsupervised learning， it eliminates the dependence on labeled data. Furthermore， the method uses WSNs sensing computing to reduce the data transmission and energy consumption of WSNs sensor nodes. In this paper， a CSNN-based bearing fault detection model is implemented using C programming and embedded in WSNs sensor nodes for experimental testing. The results show that the proposed method achieves a fault detection accuracy of 99.2%， reduces data transmission by 99.8% compared to direct wireless transmission of bearing vibration raw data， and lowers node energy consumption by approximately 29.3%.
Chemical mechanical polishing （CMP） is a key technology in semiconductor manufacturing and has been successfully applied to the planarization of insulating layers in superconducting quantum devices. First，the polishing rate was characterized as a function of the polishing cycle through multi-sample insulating layer polishing experiments， and the polishing rate was stabilized at 1 nm/s by adjusting the pre-polishing parameters. The measured non-uniformity after polishing was less than 5%， and the roughness after cleaning could reach within 1 nm， meeting the requirements for device fabrication. To evaluate the planarization performance of different morphologies， we designed test structures with graded pattern densities and used scanning electron microscopy （SEM） and atomic force microscopy （AFM） to characterize the local morphological features. Ultimately， we found that a pattern density of 50% could achieve better polishing results. Based on these polishing insights， the optimized planarization process was integrated into the manufacturing workflow of programmable Josephson junction arrays. Finally， through analysis and comparison， it was found that the planarization process could increase the operating current of the device at low temperatures to 50 mA.
Focusing on the critical on-wafer S-parameter calibration techniques for microwave and millimeter-wave chip testing， an analysis and comparison of commonly used calibration algorithms： SOLT， SOLR， LRM， LRRM， TRL， and Multiline-TRL is provide. It elaborates on the error model， frequency applicability， advantages and disadvantages of each algorithm in terms of calibration test accuracy and efficiency. The composition， parameter definitions， and purposes of the corresponding on-wafer calibration kits for each algorithm are described. Within the 1 GHz to 67 GHz frequency band， these calibration techniques are experimentally validated using internationally accepted comparison methods on the same on-wafer S-parameter measurement system. Using the highly accurate Multiline-TRL algorithm as the benchmark， the maximum deviations in the S-parameters of passive device under test （DUT） obtained from the other algorithms relative to the benchmark is calculated. This work provides a reference for selecting appropriate calibration techniques in different on-wafer test scenarios.
To address the specific challenges of class-level feature confusion （negative transfer） and low recognition accuracy of hard boundary samples caused by large-span valve opening variations in governor valve actuators， a cross-valve-opening fault diagnosis method based on adaptive feature alignment and domain adversarial transfer learning is proposed. This method utilizes a feature extractor with Resformer as its backbone to deeply mine highly discriminative fault features across various valve-opening conditions. Considering the limitations of traditional loss functions in handling hard-to-classify samples， an adaptive threshold focal loss （ATFL） function is introduced to drive the model to actively focus on ambiguous samples at the classification boundaries. Furthermore， to overcome the defect that global alignment easily destroys the sub-domain structure， a refined class-level feature adaptation module combining local maximum mean discrepancy （Local-MMD） and a cosine annealing strategy is designed. By synergizing with an auxiliary domain adversarial network， this module achieves highly efficient domain adaptation. Experimental results demonstrate that the proposed method achieves an average accuracy of 86.14% in the cross-valve-opening fault diagnosis task， validating its superiority and robustness in solving cross-condition diagnostic problems in complex industrial scenarios.
The travelling thief problem （TTP）， combining the travelling salesman problem and the knapsack problem， is an NP-hard problem whose high complexity poses solving challenges. A two-stage framework is proposed， the first stage generates high-quality routes via the Lin-Kernighan （LK） heuristic； the second stage models item selection as a multi-objective optimization problem， solved by the elite non-dominated sorting differential evolution （E-NSDE）， simultaneously optimizing total item value and travel time. By introducing an elite selection strategy， E-NSDE expands the range of individual selection while enhancing population diversity. Experiments on standard TTP benchmarks demonstrate that E-NSDE outperforms NSDE on the vast majority of instances， with growing advantages as problem scale increases； hypervolume （HV） comparisons further confirm E-NSDE's superiority over both NSDE and NSGA-Ⅱ in convergence and diversity. The proposed algorithm offers an effective approach for solving complex combinatorial optimization problems
