Search

Data-driven research in Additive Manufacturing (AM) has gained significant success in recent years. This has led to a plethora of scientific literature to emerge. The knowledge in these works consists of AM and Artificial Intelligence (AI) contexts that haven't been mined and formalized in an integrated way. Moreover, no tools or guidelines exist to support data-driven knowledge transfer from one context to another. As a result, data-driven solutions using specific AI techniques are being developed and validated only for specific AM process technologies. There is a potential to exploit the inherent similarities across various AM technologies and adapt the existing solutions from one process or problem to another using AI, such as Transfer Learning. We propose a three-step knowledge transferability analysis framework in AM to support data-driven AM knowledge transfer. As a prerequisite to transferability analysis, AM knowledge is featurized into identified knowledge components. The framework consists of pre-transfer, transfer, and post-transfer steps to accomplish knowledge transfer. A case study is conducted between two flagship metal AM processes: Laser Powder Bed Fusion (LPBF) and Directed Energy Deposition (DED). The relatively mature LPBF is the source while the less developed DED is the target. We show successful transfer at different levels of the data-driven solution, including data representation, model architecture, and model parameters. The pipeline of AM knowledge transfer can be automated in the future to allow efficient cross-context or cross-process knowledge exchange.

With the proposal of the goals of carbon peaking and carbon neutrality, China’s power grid is moving towards low-carbon transformation. As a result, the power grid is increasingly affected by the dynamic response characteristics of renewable energy. This paper analyzes the Out-of-step Oscillation center characteristics of wind power connected to the power grid. And it establishes the equivalent model of three-machine system. Then we connects the equivalent model of Doubly-fed induction generator to the system. Based on the dynamic response characteristics of Doubly-fed induction generator, we study the expression of oscillation center position when the power system is in oscillation state. And we explore the influence on the migration of Out-of-step Oscillation center when the wind turbine is connected to the system. Finally, the conclusion is verified in IEEE-9 system.

As the last defense line of the power system, splitting control is of great significance to the safety and stability of power grid. However, with the increasing penetration of wind power, photovoltaic, energy storage and other power electronic power sources in the grid, the transient stability of the power system is seriously affected. Meanwhile, the operation characteristics of power electronic power source are very different from that of traditional synchronous generator. The integration of power electronic power source affects the characteristics about out-of-step oscillation, and ultimately affects the safe implementation of splitting control. In order to split the power system more adaptably, a new splitting control method is proposed in this paper. Firstly, the equivalent external characteristic model of the wind-storage system is derived, and the influence of the wind-storage system on the out-of-step center is analyzed. Secondly, according to the out-of-step oscillation characteristics when the wind-storage system is integrated, line weak-connection index(LWcI) is proposed to evaluate the stability margin of the branch. Thirdly, according to the trajectory difference about LWcI-Δθin all branches when it is unstable and stable, we propose the splitting control criterion based on LWcI-Δθand the phase angle of voltage about the branch. Finally, the effectiveness and adaptability of the splitting criterion proposed in this paper are verified by simulation analysis of the improved IEEE-39 system.

ABSTRACTDefects such as warping, which are introduced during additive manufacturing, severely compromise the quality of parts and could even damage the 3D printer. This paper proposes automated feature extraction and hyperparameter optimization in a closed-loop in-process system to monitor warping in fused filament fabrication (FFF). The feature extraction is based on G-code analysis, and map matching between the build platform and the captured image. This allows the warping detection algorithm to be applied to different camera angles, part locations, and corner geometries. Bayesian optimization is adopted to determine the best hyperparameters for the classification model. This model, based on a convolutional neural network (CNN), is executed in a Raspberry Pi pre-configured with OctoPrint, with plugins coordinating and controlling the camera, 3D printer, and microcomputer. Based on 16 tests carried out, the warping monitoring system was determined to be 99.2% accurate.

Single-cell genomics provides substantial resources for dissecting cellular heterogeneityand cancer evolution. Unfortunately, classical DNA amplification-based methods have low throughput and introduce coverage bias during sample preamplification. We developed a single-cell DNA library preparation method without preamplification in nanolitre scale (scDPN) to address these issues. The method achieved a throughput of up to 1800 cells per run for copy number variation (CNV) detection. Also, our approach demonstrated a lower level of amplification bias and noise than the multiple displacement amplification (MDA) method and showed high sensitivity and accuracy for cell line and tumor tissue evaluation. We used this approach to profile the tumor clones in paired primary and relapsed tumor samples of hepatocellular carcinoma(HCC). We identified three clonal subpopulations with a multitude of aneuploid alterations across the genome. Furthermore, we observed that a minor clone of the primary tumor containing additional alterations in chromosomes 1q, 10q, and 14q developed into the dominant clone in the recurrent tumor, indicating clonal selection during recurrence in HCC. Overall, this approach provides a comprehensive and scalable solution to understand genome heterogeneity and evolution

The design dataset is the backbone of data-driven design. Ideally, the dataset should be fairly distributed in both shape and property spaces to efficiently explore the underlying relationship. However, the classical experimental design focuses on shape diversity and thus yields biased exploration in the property space. Recently developed methods either conduct subset selection from a large dataset or employ assumptions with severe limitations. In this paper, fairness- and uncertainty-aware data generation (FairGen) is proposed to actively detect and generate missing properties starting from a small dataset. At each iteration, its coverage module computes the data coverage to guide the selection of the target properties. The uncertainty module ensures that the generative model can make certain and thus accurate shape predictions. Integrating the two modules, Bayesian optimization determines the target properties, which are thereafter fed into the generative model to predict the associated shapes. The new designs, whose properties are analyzed by simulation, are added to the design dataset. This constructs an active learning mechanism that iteratively samples new data to improve data representativeness and machine learning model performance. An S-slot design dataset case study was implemented to demonstrate the efficiency of FairGen in auxetic structural design. Compared with grid and randomized sampling, FairGen increased the coverage score at twice the speed and significantly expanded the sampled region in the property space. As a result, the generative models trained with FairGen-generated datasets showed consistent and significant reductions in mean absolute errors.

Fused Filament Fabrication (FFF) is an additive manufacturing technology that can produce complicated structures in a simple-to-use and cost-effective manner. Although promising, the technology is prone to defects, e.g. warping, compromising the quality of the manufactured component. To avoid the adverse effects caused by warping, this paper utilizes deep-learning algorithms to develop a warping detection system using Convolutional Neural Networks (CNN). To create such a system, a real-time data acquisition and analysis pipeline laid out. The system is responsible for capturing the print layer-by-layer and simultaneously extracting the corners of the component. The extracted region-of-interest s then passed through a CNN outputting the probability of a corner being warped. If a warp s identified, a signal is sent to pause the print, creating a closed-loop in-process detection system. The underlying model tested in an experimental set-up yielding a mean accuracy of 99.3 %.

Single-cell RNA-seq technologies require library preparation prior to sequencing. Here, we present the first report to compare the cheaper BGISEQ-500 platform to the Illumina HiSeq platform for scRNA-seq. We generate a resource of 468 single cells and 1297 matched single cDNA samples, performing SMARTer and Smart-seq2 protocols on two cell lines with RNA spike-ins. We sequence these libraries on both platforms using single- and paired-end reads. The platforms have comparable sensitivity and accuracy in terms of quantification of gene expression, and low technical variability. Our study provides a standardized scRNA-seq resource to benchmark new scRNA-seq library preparation protocols and sequencing platforms.

Machine learning (ML) has recently enabled many modeling tasks in design, manufacturing, and condition monitoring due to its unparalleled learning ability using existing data. Data have become the limiting factor when implementing ML in industry. However, there is no systematic investigation on how data quality can be assessed and improved for ML-based design and manufacturing. The aim of this survey is to uncover the data challenges in this domain and review the techniques used to resolve them. To establish the background for the subsequent analysis, crucial data terminologies in ML-based modeling are reviewed and categorized into data acquisition, management, analysis, and utilization. Thereafter, the concepts and frameworks established to evaluate data quality and imbalance, including data quality assessment, data readiness, information quality, data biases, fairness, and diversity, are further investigated. The root causes and types of data challenges, including human factors, complex systems, complicated relationships, lack of data quality, data heterogeneity, data imbalance, and data scarcity, are identified and summarized. Methods to improve data quality and mitigate data imbalance and their applications in this domain are reviewed. This literature review focuses on two promising methods: data augmentation and active learning. The strengths, limitations, and applicability of the surveyed techniques are illustrated. The trends of data augmentation and active learning are discussed with respect to their applications, data types, and approaches. Based on this discussion, future directions for data quality improvement and data imbalance mitigation in this domain are identified.

Single-cell genomics provides substantial resources for dissecting cellular heterogeneityand cancer evolution. Unfortunately, classical DNA amplification-based methods have low throughput and introduce coverage bias during sample preamplification. We developed a single-cell DNA library preparation method without preamplification in nanolitre scale (scDPN) to address these issues. The method achieved a throughput of up to 1800 cells per run for copy number variation (CNV) detection. Also, our approach demonstrated a lower level of amplification bias and noise than the multiple displacement amplification (MDA) method and showed high sensitivity and accuracy for cell line and tumor tissue evaluation. We used this approach to profile the tumor clones in paired primary and relapsed tumor samples of hepatocellular carcinoma(HCC). We identified three clonal subpopulations with a multitude of aneuploid alterations across the genome. Furthermore, we observed that a minor clone of the primary tumor containing additional alterations in chromosomes 1q, 10q, and 14q developed into the dominant clone in the recurrent tumor, indicating clonal selection during recurrence in HCC. Overall, this approach provides a comprehensive and scalable solution to understand genome heterogeneity and evolution