Yujie Sun

The compressed sensing (CS) methods have two issues when processing vibration signals of rotating machinery: (1) the data-driven method and signal priors (e.g., global/local and structure priors) are not combined well, resulting in the reconstruction not converging to high-precision results; (2) the current metrics only focus on numerical accuracy and do not consider PHM scene fitness. We propose a deep compressed sensing framework (RMVS-DCS): (1) Global-Local Feature Reconstruction (GLFR) blocks alternately reconstruct features at global and local scales. Pre-train the output layer by dictionary learning, introducing the signal structure prior to guiding the model to converge to a better one. (2) A dual evaluation system, combining the numerical indicators and prognostics and health management (PHM) task accuracy, helps the model selection in PHM scenarios. On the rotor unbalance data set, its reconstruction and prediction error are reduced by 82.60% and 89.90% (compression ratio is 0.5), demonstrating the advanced reconstruction precision and PHM compatibility of the framework.

For cylinder production, visual inspection of surface defects on cylinder joints holds great significance, yet the task is challenged by the high reflectivity on joint surfaces and the presence of numerous small defects. To address this issue, a metal surface defect detection network, named Information Augmentation and Multi-scale Feature fusion with YOLO (IAMF-YOLO) based on YOLOv8, is proposed in this paper, aiming to enhance the accuracy of joint surface defect detection. The IAMF-YOLO detection network incorporates an information augmentation strategy into the backbone network and introduces a novel backbone network called Information Augmentation Network (IAN). This IAN reduces the information loss during feature down-sampling extraction. Furthermore, to boost the detection performance of small defects on joint surfaces, improvements have been made to the Path Aggregation Network (PAN), and a Parallel Multi-scale Feature Pyramid Network (PMFPN) is introduced. PMFPN is designed to separate feature fusion into two stages: an initial feature fusion stage and an incremental feature fusion stage, of which the latter one further merges the features of the former one. Through ablation studies and comparative experiments on IAMF-YOLO, it is demonstrated that this model surpasses YOLOv8 in terms of detection accuracy and precision. Specifically, it achieves an accuracy of 88.6% on a self-constructed dataset, 91.0% on the NEU-DET dataset, and 84.2% on the GC10-DET dataset. Furthermore, this study demonstrates the robustness of the proposed model against environmental disturbances through actual defect detection under varying lighting conditions.

The orientation information is critical for the mobile robot. However, the robot’s constrained locomotion, complicated magnetic field environment, and the coupled orientation estimation present challenges for using the microelectromechanical magnetometer in real-time robot orientation determination. This article focuses on calibrating the magnetometer in constrained robot locomotion and achieving the robot’s decoupled roll, pitch, and yaw estimates. A magnitude-free and reference-free calibration approach is proposed to calibrate the magnetometer using the measurements collected from the principal plane of the robot locomotion. Moreover, a decoupled orientation estimator is created under the framework of the extended Kalman filter using the calibrated triaxial magnetic, angular rate, and gravity sensor readings. Two independent experiments validate the effectiveness of magnetometer calibration and orientation decoupling, respectively. A case study of estimating the mobile robot’s orientation shows that the proposed methods can achieve accurate heading estimation and decouple the roll and pitch from the magnetic field data.

Orientation estimation using inertial and magnetic sensors has permeated into various applications. However, robust roll and pitch estimates are still challenging for the orientation estimators when the sensors are exposed to magnetic disturbances. In this article, we proposed a decoupled orientation estimation approach (DOEA) to separate and achieve accurate roll and pitch estimates from the magnetic measurements. In the proposed DOEA, we formulate a reference vector, which is perpendicular to the gravitational vector and keeps constant size and direction, and derive its covariances to replace the magnetic field vector as the observation references. The sensor measurements of the two orthogonal homogeneous fields are further fused in the extended Kalman filter frame to correct the predicted orientation of the gyroscope integration. To validate the performance of the proposed DOEA, we compare it with three state-of-the-art approaches in four different experiments. In the magnetically disturbed environment without reliability validation of the sensors, experiment results show that the DOEA could provide robust and accurate roll and pitch measurements with 0.001- and 0.007-rad root-mean-square error (RMSE). The influence of the magnetic disturbances on the DOEA is controlled only on the yaw direction. In the wide-range static and dynamic mixed motion tracking of the robot, experiment results show that the proposed DOEA achieves 0.035-, 0.024-, and 0.018-rad RMSE of roll, pitch, and yaw estimates without singularities. Moreover, the proposed DOEA converges faster than the popular optimization filters and achieves computational efficiency 56% faster than the two-step orientation estimator, which is beneficial for realization on the embedded system.

The zero-velocity update (ZUPT) technique plays a key role in foot-mounted inertial navigation system, which can minimize the drift errors of low-cost inertial measurement unit (IMU) by resetting the velocity to zero when the foot is stationary. However, it is difficult to detect the zero-velocity interval (ZVI) in mixed locomotion patterns. This article presents a novel ZVI detector based on adaptive simulated energy consumption (SEC) curve. It can identify the ZVIs of various locomotion patterns, such as normal walking, fast walking, running, going upstairs, and downstairs. The SEC curve is generated based on the signal processing approach using the interlaced peak property of the IMU-measured angular rate and acceleration waveforms. Another benefit of the proposed method is that the IMU can be mounted on the instep without restrictions of orientation and position, which improves flexibility and convenience. Experiment results show that the proposed algorithm performs better than three other widely used methods in mixed gait patterns.

Accurate orientation estimation using low-cost inertial and magnetic sensors is important. However, since the magnetometer measurements affect pitch and roll estimates in addition to yaw determination, the resultant accuracy reduction under magnetic disturbances is adverse for special occasions where high-precision pitch and roll estimates are essential. This article presents a double-extended Kalman filter (DEKF) orientation estimator with unit quaternion as output, which decouples magnetometer effects on pitch and roll estimates by decomposing the orientation into tilt quaternion and heading quaternion. The tilt quaternion, including pitch and roll information, is estimated by the first EKF where accelerometer and gyroscope measurements are processed and the inaccurate heading information is discarded. The heading quaternion, solely describing the yaw orientation, is determined by the second EKF, which fuses the vertical component of angular rate measurements and the horizontal component of magnetic measurements. True quaternion is denoted by the multiplicative of tilt and heading quaternions. The separate property of DEKF limits the influences of magnetic disturbances on yaw estimation, making the pitch and roll angles immune to them. Experiment results show that the proposed method solves the coupled disturbance problem without losing accuracy.