Xiufeng Li

Under extreme operating conditions such as high temperatures, strong corrosion, and cyclic thermal shocks, key equipment in solar tower power plants (STPPs) is prone to severe accidents and results in significant losses. To systematically quantify potential failure risks and address the methodological gaps in existing research, this study proposes a risk assessment framework combining a novel scenario propagation model covering triggering factors, precursor events, accident scenarios, and response measures with an interval type-2 fuzzy set (IT2FS) Bayesian network. This framework establishes equipment failure evolution pathways and consequence evaluation criteria. To address data scarcity, the methodology integrates actual case data and expert elicitation to obtain assessment parameters. Specifically, an IT2FS-based similarity aggregation method quantifies expert opinions for prior probability estimation. Additionally, to reduce computational complexity and enhance reliability in conditional probability acquisition, the IT2FS-integrated best–worst method evaluates the relative importance of parent nodes, combined with a leakage-weighted summation algorithm to generate conditional probability tables. The model was applied to a western Chinese STPP and the results show the probabilities of receiver blockage, pipeline blockage, tank leakage, and heat exchanger blockage are 0.061, 0.059, 0.04, and 0.08, respectively. Under normal operating conditions, the occurrence rates of level II accident consequences for all four equipment types remain below 6%, with response measures demonstrating significant suppression effects on accidents. The research results provide critical decision-making support for risk management and mitigation strategies in STPPs.

Safety barriers (SBs) are important means of reducing failure risks of process systems. As barriers vary in type and function, their effectiveness needs to be evaluated in order to find a more reasonable configuration strategy. However, in practice, there is often a lack of accurate and complete data relating to SBs, which poses a significant challenge in quantitatively assessing their effectiveness. To address this issue, in this study, we propose a semi-quantitative approach for evaluating the effectiveness of both preventive and protective barriers in process systems by integrating expert elicitation, interval-valued intuitionistic fuzzy numbers (IVIFNs), and a bow-tie model. In this approach, the bow-tie model is first applied to describe the system failure scenarios and the action phases of the barriers, and then IVIFNs with expert judgment are introduced to obtain the failure probabilities of basic events and the effects of SBs. Subsequently, the effectiveness of each barrier is measured by comparing the relative change in failure risk due to the addition of the barrier. To verify the feasibility of this approach, a natural gas storage tank with some barriers was analyzed. The results show that the regular inspection of the deformation or damage of the storage tank has the highest effectiveness, followed by the installation and regular maintenance of safety electrical equipment. Furthermore, compared to a single barrier, multiple barriers can significantly reduce the system risk.

In metallographic examination, spherular pearlite gradation, an important step in a metallographic examination, is the main indicator used to assess the reliability of heat-resistant steel. Recognition of pearlite spheroidization via the manual way mainly depends on the subjective perceptions and experience of each inspector. Deep learning-based methods can eliminate the effects of the subjective factors that affect manual recognition. However, images with incorrect labels, known as noisy images, challenge successful application of image recognition of deep learning models to spherular pearlite gradation. A deep-learning-based label noise method for metallographic image recognition is thus proposed to solve this problem. We use a filtering process to pretreat the raw datasets and append a retraining process for deep learning models. The presented method was applied to image recognition for spherular pearlite gradation on a metallographic image dataset which contains 422 images. Meanwhile, three classic deep learning models were also used for image recognition, individually and coupled with the proposed method. Results showed that accuracy of image recognition by a deep learning model solely is lower than the one coupled with our method. Particularly, accuracy of ResNet18 was improved from 72.27% to 77.01%.