Detection of Flying Metal Bodies Based on Photoelectric Composite Sensing.

In order to reduce the impact of the environment on the accuracy and sensitivity of detection, and to meet the requirements of concealment from detection and being lightweight, a technology for detecting flying metal objects based on photoelectric composite sensors is proposed. The method first analyzes the target's characteristics and detection environment, and then compares and analyzes the methods for detecting typical flying metal objects. On the basis of the traditional eddy current model, the photoelectric composite detection model that meets the requirements of detecting flying metal objects was studied and designed. For the problems of the short detection distance and the long response time of the traditional eddy current model, the performance of the eddy current sensor was improved to meet the requirements of detection through optimizing the detection circuit and coil parameter model. Meanwhile, to meet the goal of being lightweight, an infrared detection array model applicable to flying metal bodies was designed, and simulation experiments of composite detection based on the model were conducted. The results show that the flying metal body detection model based on photoelectric composite sensors met the requirements of distance and response time for detecting flying metal bodies and may provide an avenue for exploring the composite detection of flying metal bodies.

Experimental Study Comparing the Effectiveness of Physical Isolation and ANN Digital Compensation Methodologies at Eliminating the Stress Wave Effect Error on Piezoelectric Pressure Sensor.

Stress wave, accompanied by explosion shock wave overpressure measurement and dynamic pressure calibration on shock tube, could cause error signals in the piezoelectric pressure sensor (PPS) used for measuring and calibrating. We may call this error the stress wave effect (SWE). In this paper, the SWE and its isolation from PPS were studied by using a split Hopkinson pressure bar (SHPB). In the experimental study of SWE, when increasing the input stress, the corresponding output signal of the PPS was analyzed, and the existence of SWE was verified using the result of the spectrum analysis of the output signal. The stress wave isolation pedestal used in the stress wave isolation experiment was made of nylon and plexiglass polymer materials. The effects of the isolation pedestal's materials and length on the stress wave isolation were analyzed using the study results. Finally, an artificial neural network (ANN) was trained with the data of the SWE study and was further applied to compensate the SWE error of the PPS output signal. The compensating results were compared with the isolating results, and the advantages and disadvantages of the digital compensation and physical isolation methods were analyzed.

Modeling and Studying Acceleration-Induced Effects of Piezoelectric Pressure Sensors Using System Identification Theory.

Transient pressure testing is often accompanied by shock acceleration. Aiming at the acceleration-induced effects of pressure sensors, a dynamic compensation method combining empirical mode decomposition (EMD) with system identification theory (SIT) is proposed in this paper. This method is more effective at reducing the error of the acceleration-induced effects without affecting the sensor's sensitivity and inherent frequency. The principle and theoretical basis of acceleration-induced effects is analyzed, and the static and dynamic acceleration-induced effects on the quartz crystal of a piezoelectric pressure sensor are performed. An acceleration-induced effects dynamic calibration system is built using a Machete hammer, which generates acceleration signals with larger amplitude and narrower pulse width, and an autoregressive exogenous (ARX)mathematical model of acceleration-induced effects is obtained using empirical mode decomposition-system identification theory (EMD-SIT). A digital compensation filter for acceleration-induced effects is designed on the basis of this model. Experimental results explain that the acceleration-induced effects of the pressure sensor were less than 11% after using the digital compensation filter. A series of test data verify the accuracy, reliability, and generality of the model.

Random pruning: channel sparsity by expectation scaling factor.

Pruning is an efficient method for deep neural network model compression and acceleration. However, existing pruning strategies, both at the filter level and at the channel level, often introduce a large amount of computation and adopt complex methods for finding sub-networks. It is found that there is a linear relationship between the sum of matrix elements of the channels in convolutional neural networks (CNNs) and the expectation scaling ratio of the image pixel distribution, which is reflects the relationship between the expectation change of the pixel distribution between the feature mapping and the input data. This implies that channels with similar expectation scaling factors (δE) cause similar expectation changes to the input data, thus producing redundant feature mappings. Thus, this article proposes a new structured pruning method called EXP. In the proposed method, the channels with similar δE are randomly removed in each convolutional layer, and thus the whole network achieves random sparsity to obtain non-redundant and non-unique sub-networks. Experiments on pruning various networks show that EXP can achieve a significant reduction of FLOPs. For example, on the CIFAR-10 dataset, EXP reduces the FLOPs of the ResNet-56 model by 71.9% with a 0.23% loss in Top-1 accuracy. On ILSVRC-2012, it reduces the FLOPs of the ResNet-50 model by 60.0% with a 1.13% loss of Top-1 accuracy. Our code is available at: https://github.com/EXP-Pruning/EXP_Pruning and DOI: 10.5281/zenodo.8141065.

[Based on THz spectroscopy detection method for the concentration of 1,3-dinitrobenzene volatile gas].

For the unexpected situations occur that 1,3-dinitrobenzene volatile gas leaks in industrial production process, designed the differential characteristics absorption detection system for 1,3-dinitrobenzene volatile gas based on THz spectrum technique. The method can detect the concentration of trace 1,3-dinitrobenzene toxic gases through the THz characteristic wavelengths differential absorption method. System dealt with THz characteristics wavelengths of the two chambers, which provide difference dates, a group is standard air, and the other group tested the sample gas. Four main absorption bands of 1,3-dinitrobenzene is at 0.635, 0.912, 1.095 and 1.435 THz nearby in detection results, and according to the ratio of absorption coefficient in the corresponding absorption band it calculated the accurate amplitude from the corresponding wavelengths, at last the gas concentration was inversed. After the two traditional methods (chromatography and infrared absorption method) of experimental analysis and THz absorption detection method of the simulation study shows, precision of the detection capabilities of chromatography is high and error is small. But the chromatography separation time varies with the material, long cycle, slow; chromatography can not achieve real-time detection speed to realize real-time detection. Infrared absorption of environmental humidity is high, subject to drying. At the same time, THz absorption of environmental humidity is low; THz absorption method also has important characteristics of real time, strong anti-jamming, especially the water vapor, so it is more suitable for practical application.