The accurate measurement of position and orientation for shearers is a key technology in realizing an automated, fully-mechanized, coal mining face. Since Global Positioning System (GPS) signal cannot arrive at the coal mine underground, wireless sensor network positioning system cannot operate stably in the coal mine; thus a strap-down inertial navigation system (SINS) is used to measure the position and orientation of the shearer. Aiming at the problem of the SINS accumulative error, this paper proposes a positioning error correction method based on the motion constraint-aided SINS zero velocity updated (ZUPT) model. First of all, a stationary state detection model of the shearer is built with median filter based on the acceleration and angular rate measured by the SINS. Secondly, the motion of the shearer is analyzed using coal mining technology, then the motion constraint model of the shearer is established. In addition, the alternate action between the motion constraint model and the ZUPT model is analyzed at the process of movement and cessation of the shearer, respectively; hence, the motion constraint-aided SINS ZUPT model is built. Finally, by means of the experimental platform of the SINS for the shearer, the experimental results show that the maximum position error with the positioning model proposed in this paper is 1.6 m in 180 s, and increases by 92.0% and 88.1% compared with the single motion constraint model and single ZUPT model, respectively. It can then restrain the accumulative error of the SINS effectively.

Drug-drug interaction (DDI) refers to the inhibitory or enhancing effects between different drugs. Existing DDI prediction methods primarily use graph neural networks (GNNs) to directly represent drug molecular features. However, they often ignore the 3D structures of different atoms within drug molecules and the impact of noise in GNNs on DDI prediction. Consequently, the accuracy of GNN-based DDI prediction remains unsatisfactory. To address these limitations, this study proposes a DDI prediction method based on atomic 3D position encoding and an elastic message passing graph neural network (A3DPE-EMPGNN). Firstly, we construct an atomic feature network based on an attention mechanism and a message passing neural network. This network leverages 3D position encoding based on the molecular centroid to learn the features of different atoms and their associated chemical bonds, thereby constructing a graph-based molecular representation. Secondly, we design a molecular feature network that incorporates an attention mechanism, utilizing multi-head attention to capture interaction information between different drug molecules. Thirdly, we employ an adversarial attack detection and defense strategy, integrating supervised and contrastive loss learning to optimize the model and enhance its robustness while performing DDI prediction. Lastly, we evaluate the effectiveness of A3DPE-EMPGNN on two real-world datasets. Experimental results clearly demonstrate that our method achieves over 98% accuracy across ACC, AUC, AP, and F1-score, outperforming state-of-the-art GNN-based models.

With the development of horizontal Wells and extended reach Wells in the field of oil and gas exploitation, it is of great significance to the development of MWD which can measure the drilling trajectory in real time. Therefore, in recent years, Measurement while drilling (MWD) has been widely recognized and rapidly developed in the field of precise target drilling. Accelerometers and fluxgates are related to well trajectory measurement in the MWD. The external environment and the processing and installation error of the sensor lead to some errors in the measured value. Deviation caused by environmental interference and installation error leads to the low control precision of drilling tool attitude. In view of the shortcomings of the current calibration methods and the time-consuming and complicated operation of the traditional tri-axis orthogonal inclinometer calibration, this paper proposes the attitude error calibration method of the MWD based on the support vector classifier and the K-proximity method. First, a neural network model is established based on the measured calibration results and the measurements errors. Then, the K-proximity method is used to further classify the drill tool attitude data, and a neural network model combining the K-proximity method is constructed. Secondly, the software and hardware experimental platform for on-line calibration of the slant meter while drilling is built by using the calibration frame and the geomagnetic sensor. Finally, the experimental results showed that the mean deviation of well inclination angle and tool face angle is controlled within ±0.2°, and the mean deviation of azimuth is controlled within ±0.5°, which verified that the algorithm can increase accurate of drilling tool attitude.

The accurate calculation of the attitude angle of the down-hole bit is the key technology to realize the automation and intelligence in the field of oil drilling. Therefore, in recent years, Measurement while drilling (MWD) has been widely recognized and rapidly developed in the field of precise target drilling. In order to solve the problem of azimuth error caused by the interference of the flux gate sensor by the fixed magnetic field and the induced magnetic field in the process of solving the attitude angle of the inclinometer, this paper proposes a calibration method of the inclinometer while drilling under the cooperation of multiple models. Firstly, the error analysis is carried out for the interference magnetic field of flux gate sensor. Secondly, the error compensation is carried out for the fixed magnetic field and the induced magnetic field respectively by using the ellipsoid fitting algorithm under the proposed limited conditions and the Newton iterative model under the magnetic shielding coefficient solution. Thirdly, the magnetic field environment during drilling is simulated by building an experimental platform composed of calibration frame and inclinometer carry out experimental verification. Finally, the experimental results show that compared with the accuracy of azimuth before error compensation, the maximum error of azimuth is reduced to 0.486° and the accuracy is increased to 98.58%. Thus, the interference of a fixed magnetic field and induced magnetic field on inclinometer is effectively suppressed, and the correctness and feasibility of the method proposed in this paper are verified.

Diagnosing physical function impairment (PFI) in older adults is challenging due to delayed diagnosis and cumbersome clinical assessments. We propose PFIMPA, a novel artificial intelligence method that enables seamless remote PFI assessment using physical activity (PA) data provided by wrist-worn accelerometers for the first time. To capture the complex cumulative and dynamic characteristics of PA, we designed a comprehensive feature representation method that integrates demographic information, clinical PA features, time-frequency domain features, and PA sequence conversion image features (PASCIs) based on Gramian Angular Field, Markov Transfer Field, and Recursive Plot transformations. We developed a lightweight convolutional neural network enhanced by Convolutional Block Attention Module to extract spatio-temporal features from PASCIs, further encapsulating the correlation, accumulation, and transition properties of PA. PFI detection is achieved through a three-stage feature-level fusion method combined with machine learning and hyperparameter optimization. The PASCIs module achieves an AUC of 83.72% in PFI prediction, while the multimodal approach combining clinical, time-frequency, and PASCIs features improves the AUC to 87.16%, surpassing state-of-the-art time series classification models. PFIMPA represents a significant advancement in PFI assessment for older adults, enabling seamless remote monitoring and timely intervention.