8 December 2023 in IEEE Transactions on Instrumentation and Measurement
Noise suppression is of great significance in high-precision micro thrust calibration and measurement. Existing methods cannot simultaneously recover both the nonlinear transitions (generated by thrust) and smooth trends of the signals. Additionally, they exhibit poor robustness to noise model changes. To address the issues, in this research we propose a novel self-supervised contrastive denoising autoencoder (SCDAE) model. The SCDAE features end-to-end signal denoising based on a multi-task joint learning framework combining contrastive representation learning and denoising autoencoder (DAE) without the need for a clean signal reference. Specifically, SCDAE consists of a representation learner (RPL) and a denoising learner (DEL). In addition, considering the multi-staircase property of the measurement signal, a novel parallel hybrid data augmentation strategy is designed in the RPL to create positive sample pairs and greatly force the network to extract robust content-invariant and disentangled features. Besides, in order to solve the blindness of training target creation for the DEL, a new training target creation strategy is proposed based on the variational mode decomposition (VMD). An attention module is introduced to adaptively reinforce the content-invariant disentangled features from the RPL, guiding the DEL to generate the essential dynamic nature from the raw input signal, thus suppressing noise information. Extensive simulations and experimental case study were carried out to demonstrate the effectiveness and advancement of the proposed method in recovering signal details (nonlinear transitions and smooth trends) with comparison to existing classical methods.
21 August 2023 in IEEE Transactions on Instrumentation and Measurement
Micro-Newton thrusters are widely utilized in the field of astronautics. Typically, micro-newton thrust signal processing aims at the restoration of piecewise constant signals. Improving the amplitude accuracy of the denoised signal remains a major difficulty. In this research, we proposed a new nonlinear total variation denoising (TVD) algorithm with an adjustable majorization-minimization (MM) upper bound function for optimizing the iterative solution process of TVD. The influences of the upper bound function of the TVD on denoising are analyzed in detail. It is demonstrated that the proposed aTVD method shows extraordinary robustness for denoising signals with different noise levels. In addition, this algorithm can also adjust the sensitivity of the jumps in signals for optimized amplitude accuracy effectively. Numerical examples show that the proposed method has 40.4% and 18.2% accuracy improvement over conventional TVD using traditional quadratic upper bound and enhanced TVD using exponential upper bound function. Moreover, the advantages of the proposed method are validated in an experiment of measured thrust signal processing. The proposed method can be further adopted to improve the performance of signal denoising algorithms developed from the TVD method.
30 November 2023 in Vacuum
28 January 2025 in IEEE Transactions on Instrumentation and Measurement
Thrust is one of the key parameters that determines the performance of thrusters in the propulsion systems of microsatellites. Thrusters are widely used in space exploration, quantum communication and other fields. At present, the study on thrust measurement under low load has made great progress, meeting the requirements of high precision. However, it is still challenging to achieve high precision under heavy load. In this paper, a flexible inverted pendulum based on multiple elliptic flexure hinges is proposed to achieve high accuracy thrust measurement under heavy load. The theoretical model between sensitivity, stability and pendulum parameters under heavy load is established. The principle that the pendulum can balance heavy load and high accuracy is explained as well. The pendulum is calibrated using precision machined electro-static combs. The performance of the pendulum was evaluated by a cold gas thruster and a Hall thruster. Experimental results show that the resolution of the pendulum is better than 0.3 μN and the measurement range covers 0.3-1300 μN for loads up to 4 kg.
2 June 2025 in IEEE Transactions on Instrumentation and Measurement
Sub-micro-Newton thrusters on space satellites are widely used in spaceborne gravitational wave detection, making accurate measurement of micro-thrust-generated thrust essential. Moreover, micro-propulsion systems often exhibit substantial mass, posing challenges for thrust measurement. This research prompted the development of a sub-micro-thrust measurement device based on the inverted pendulum principle, which with-stands 8 kg loads and boasts a large thrust-to-weight ratio exceeding 109. Experimental results demonstrate that adjusting the stiffness of the inverted pendulum can effectively enhance the resolution of the device. Piecewise constant signal (PCS) thrust force was introduced, and the nonlinear total variation denoising (TVD) algorithm method was adopted for signal denoising. It is validated that thrust measurement in the range of 0–261 μN is achieved with a minimum resolution of 24 nN. The device design and stiffness adjustment method proposed in this paper, as well as the data denoising method utilized, greatly enhance the thrust resolution measurement. This approach enables the development of an accurate micro-thrust model for spaceborne gravitational wave detection missions while simultaneously offering a novel solution for sub-micro-newton-level thrust measurement under heavy load conditions.
12 November 2024 in IEEE Sensors Journal
Effective denoising of micro-thrust measurement signals (MTMSs) is crucial for analyzing the dynamic characteristics of micro-thrusters. Existing methods struggle to suppress noise while accurately reconstructing smooth trend features and step edge features. These issues can lead to potential misjudgments and omissions of micro-thrust step effects. In this paper, we propose a novel perspective on self-supervised blind denoising algorithm with advantage of feature-adaptive supervision for MTMSs denoising. Specifically, the cascaded learning framework comprising a blind aware network (BAN), a local perception network (LPN), and a feature-adaptive blind denoising network. BAN employs the "noise2noise" denoising concept, utilizing a patch-Transformer model to capture short-term and long-term smooth trend features, providing denoising supervision for smooth regions. Besides, LPN uses the local receptive fields of the convolutional layer to capture step edge features and provide denoising supervision for the step regions. Through the proposed feature-adaptive regulatory factor, the LPN is selectively supervised by the BAN’s output. Finally, the blind denoising network adaptively integrates supervisions from both smooth and step regions through the regulatory factor, achieving effective denoising of MTMSs. Numerical example shows that the proposed method has 50.6% and 23.4% root mean square error (RMSE) reductions over the low-pass Butterworth filter and the total variation denoising (TVD) algorithm. Moreover, experimental results demonstrate that the proposed method outperforms existing baseline methods in denoising performance and effectively reconstructs smooth trend and step edge features.