Search

Deep convolutional neural networks (CNNs) have presented amazing performance in the task of semantic segmentation. However, the network model is complex, the training time is prolonged, the semantic segmentation accuracy is not high and the real-time performance is not good, so it is difficult to be directly used in the semantic segmentation of road environment images of autonomous vehicles. As one of the three models of deep learning, the auto-encoder (AE) has powerful data learning and feature extracting capabilities from the raw data itself. In this study, the network architecture of auto-encoder and convolutional auto-encoder (CAE) is improved, supervised learning auto-encoder and improved convolutional auto-encoder are proposed, and a hybrid convolutional auto-encoder model is constructed by combining them. It can extract low-dimensional abstract features of road environment images by using convolution layers and pooling layers in front of the network, and then supervised learning auto-encoder are used to enhance and express semantic segmentation features, and finally de-convolution layers and un-pooling layers are used to generate semantic segmentation results. The hybrid convolutional auto-encoder model proposed in this paper not only contains encoding and decoding parts which are used in the common semantic segmentation models, but also adds semantic feature enhancing and representing parts, so that the network which has fewer convolutional and pooling layers can still achieve better semantic segmentation effects. Compared to the semantic segmentation based on convolutional neural networks, the hybrid convolutional auto-encoder has fewer network layers, fewer network parameters, and simpler network training. We evaluated our proposed method on Camvid and Cityscapes, which are standard benchmarks for semantic segmentation, and it proved to have a better semantic segmentation effect and good real-time performance.

The bandwidth-enhanced oversampling successive approximation (BE-OSA) readout technique is proposed in this article to reduce the noise floor of the readout circuit for micro-electromechanical systems (MEMS) capacitive accelerometer while achieving high power efficiency in terms of the figure of merit (FoM). The open-loop structure has been widely used in MEMS capacitive accelerometer for the Internet of Things (IoT) applications due to its low power consumption. However, in the open-loop accelerometer, the capacitance variation in the sensing element is limited to the femto-farad level to overcome nonlinearity. As a result, the thermal noise from the parasitic capacitance becomes significant. The ability of the readout circuit to deal with thermal noise is determined by the front-end switched-capacitor capacitance-to-voltage convertor (SC CVC) rather than the back-end analog-to-digital converter (ADC). To reduce the noise floor, the traditional oversampling method increases the sampling frequency of SC CVC by increasing the transconductance of the amplifier, but this leads to low power efficiency. In this work, the BE-OSA technique provides a high power efficiency method as it increases the sampling frequency of SC CVC without increasing the transconductance of the amplifier. The SC CVC based on the BE-OSA technique is demonstrated in a readout circuit fabricated by a commercial 0.18- $\mu \text{m}$ Bipolar-CMOS-DMOS (BCD) process and tested with a femto-farad MEMS accelerometer. The measurement results show that compared with the readout circuit without using the BE-OSA technique, the readout circuit using BE-OSA reduces the noise floor from 2.5 to 0.9 aF/ $\surd $ Hz. Compared with other similar works, the proposed readout circuit achieves the best power efficiency in terms of both the absolute power efficiency ( $FoM_{2} =243$ fJ) among the switched-capacitor readout circuits and the relative power efficiency ( $FoM_{3} = 0.14$ ).

New approaches for enhanced oil recovery (EOR) with a reduced environmental footprint are required to improve recovery from mature oil fields, and when combined with carbon capture and storage (CCS) can provide useful options for resource maximisation during the net zero transition. Electrical heating is investigated as a potential EOR method in carbonate reservoirs. Samples were placed in an apparatus surrounded by a wire coil across which different DC (direct current) voltages were applied. Monitoring the imbibition of both deionized water (DW) and seawater (SW) into initially oil-wet Austin chalk showed that water imbibed into the rock faster when heated in the presence of a magnetic field. This was associated with a reduction in the water–air contact angle over time measured on the external surface of the sample. Without heating, the contact angle reduced from 127° approaching water-wet conditions, 90°, in 52 min, while in the presence of heating with 3 V, 6 V, and 9 V applied across a sample 17 mm in length, the time required to reach the same contact angle was only 47, 38 and 26 min, respectively, while a further reduction in contact angle was witnessed with SW. The ultimate recovery factor (RF) for an initially oil-wet sample imbibed by DW was 13% while by seawater (SW) the recorded RF was 26% in the presence of an electrical heating compared with 2.8% for DW and 11% for SW without heating. We propose heating as an effective way to improve oil recovery, enhancing capillary-driven natural water influx, and observe that renewable-powered heating for EOR with CCS may be one option to improve recovery from mature oil fields with low environmental footprint.

Crude oil, water and rock (CWR) surface chemistry is a key parameter in oil and gas recovery from hydrocarbon reservoirs. This paper presents the chemistry of CWR interaction in presence of two water soluble nanoparticles for carbonate rocks. Two most common nanoparticles of aluminium oxide (Al2O3) and silica (SiO2) were selected and utilised for this study. Calcite was first modified to an oil-wet system to resemble the reservoir wetting condition then treated with nano-fluids containing under study nanoparticles at different concentrations. Alteration of wettability was then quantified using contact angle measurements, and zeta potential analysis. The change in fluid chemistry of the water due to presence of nanoparticles was also monitored before and after treatment of carbonate rock. The results show that after treatment of the oil-wet samples with nano-fluids, the solution’s pH decreased for SiO2 while it increased slightly for Al2O3. The contact angle results show a decrease trend for both nanoparticles but more pronounced for Al2O3. These results are in line with zeta potential results in which very negative surface charge for an oil-wet rock was converted to less negative or even positive for certain concentrations of nano-particles. Floating phenomenon also applied to calculate the level of water percentage between floated and sank powder for the modified calcite where amount of water level increased significantly when nanoparticles added to water solutions. Comparison of all tests show that silica nano-fluid with concentration between 0.1 and 2 wt % can be efficient EOR agent. High salinity is definitely not a good option to formulate nano-fluids such as alumina and silica at high concentrations showing inverse effects.

Aiming at the problem that the vibration signals of rolling bearings working in a harsh environment are mixed with many harmonic components and noise signals, while the traditional sparse representation algorithm takes a long time to calculate and has a limited accuracy, a bearing fault feature extraction method based on the ensemble empirical mode decomposition (EEMD) algorithm and improved sparse representation is proposed. Firstly, an improved orthogonal matching pursuit (adapOMP) algorithm is used to separate the harmonic components in the signal to obtain the filtered signal. The processed signal is decomposed by EEMD, and the signal with a kurtosis greater than three is reconstructed. Then, Hankel matrix transformation is carried out to construct the learning dictionary. The K-singular value decomposition (K-SVD) algorithm using the improved termination criterion makes the algorithm have a certain adaptability, and the reconstructed signal is constructed by processing the EEMD results. Through the comparative analysis of the three methods under strong noise, although the K-SVD algorithm can produce good results after being processed by the adapOMP algorithm, the effect of the algorithm is not obvious in the low-frequency range. The method proposed in this paper can effectively extract the impact component from the signal. This will have a positive effect on the extraction of rotating machinery impact features in complex noise environments.

In the actual environment, the various deformations of the positioning holes of the mechanical parts and the different backgrounds will cause the circle to have irregular boundaries, which will cause the identification and positioning of the center of the circle to be inaccurate. Therefore, this paper proposes a centroid compensation algorithm based on the centroid of a circular hole. The algorithm first performs image preprocessing on the circular hole to enhance its contrast, and then performs binary and filtering processing to find its centroid. According to the binary image in the pixel The area occupied by the plane is used to identify, determine and compensate the center position of the circular hole. Experiments show that the algorithm can accurately locate the center of irregular holes with deformation, and has high accuracy and stability.

The structure design and material selection optimization of the secondary transmission planetary reducer gear are carried out, and the lightweight design is carried out under the condition of meeting the contact fatigue strength and bending fatigue strength; the optimized gear is simulated by using the transient dynamics module of finite element, which provides a new method to reduce the mass of planetary reducer and improve its overall power density reference resources.

The core of driving force control strategy is the research core of electric vehicle driven by wheel motor. In order to give full play to the advantages of wheel motor drive system, a reasonable control strategy must be designed. In this paper, the stability of the vehicle is determined based on the sideslip angle of the center of mass of the vehicle. Through the speed and steering wheel angle, the actual yaw rate of the vehicle at a certain speed and the theoretical yaw rate at this speed are determined as the control objectives. The PID control is used to distribute the four-wheel brake pressure, generate additional yaw moment to adjust the vehicle attitude; and the traditional PID control is used to control the vehicle attitude. In contrast, fuzzy control using human expert control experience, for the control of nonlinear and complex objects shows the advantages of good robustness and high control performance. Therefore, on the basis of PID, through neural network learning, fuzzy neural network is used to optimize the output fuzzy rules and membership function to control the vehicle. The whole process is carried out with CarSim and Simulink Simulation is carried out to verify the control optimization process.

This paper designs a close loop $\Sigma - \Delta $ readout circuit for differential MEMS accelerometer. A technique named oversampling successive approximation (OSA) is employed to build basic amplifiers and integrators. This technique can largely reduce the gain error and thus low gain amplifier such as single stage amplifier is allowed to be used. As a result, the power consumption and chip area are reduced. However, the OSA-based amplifiers and integrators are vulnerable to the interference caused by charge injection and leakage current from the specific MOSFET switches. This drawback is analyzed in detail and the interference suppressing solutions are given. The OSA-based readout circuit is fabricated in a commercial 0.18 $\mu \text{m}$ BCD process. The measurement results show that the interference is reduced by 20 dB in the circuit with interference suppressing solutions compared with the circuit without interference suppressing solutions. And the noise floor is 24 $\mu \text{g}$ /rtHz. The readout circuit achieves a 0.07% gain error with a low power consumption of 0.5 mW and 9 MHz sampling rate.