Vision-based multi-level synthetical evaluation of seismic damage for RC structural components: a multi-task learning approach

Recent studies for computer vision and deep learning-based,post-earthquake inspections on RC structures mainly perform well for specific tasks,while the trained models must be fine-tuned and re-trained when facing new tasks and datasets,which is inevitably time-consuming.This study proposes a multi-task learning approach that simultaneously accomplishes the semantic segmentation of seven-type structural components,three-type seismic damage,and four-type deterioration states.The proposed method contains a CNN-based encoder-decoder backbone subnetwork with skip-connection modules and a multi-head,task-specific recognition subnetwork.The backbone subnetwork is designed to extract multi-level features of post-earthquake RC structures.The multi-head,task-specific recognition subnetwork consists of three individual self-attention pipelines,each of which utilizes extracted multi-level features from the backbone network as a mutual guidance for the individual segmentation task.A synthetical loss function is designed with real-time adaptive coefficients to balance multi-task losses and focus on the most unstably fluctuating one.Ablation experiments and comparative studies are further conducted to demonstrate their effectiveness and necessity.The results show that the proposed method can simultaneously recognize different structural components,seismic damage,and deterioration states,and that the overall performance of the three-task learning models gains general improvement when compared to all single-task and dual-task models.

Optimal Maintenance Modeling for Systems with Multiple Non-Identical Units Using Extended DSSP Method

In the optimal maintenance modeling, all possible maintenance activities and their corresponding probabilities play a key role in modeling. For a system with multiple non-identical units, its maintenance requirements are very complicated, and it is time-consuming, even omission may occur when enumerating them with various combinations of units and even with different maintenance actions for them. Deterioration state space partition (DSSP) method is an efficient approach to analyze all possible maintenance requirements at each maintenance decision point and deduce their corresponding probabilities for maintenance modeling of multi-unit systems. In this paper, an extended DSSP method is developed for systems with multiple non-identical units considering opportunistic, preventive and corrective maintenance activities for each unit. In this method, different maintenance types are distinguished in each maintenance requirement. A new representation of the possible maintenance requirements and their corresponding probabilities is derived according to the partition results based on the joint probability density function of the maintained system deterioration state. Furthermore, focusing on a two-unit system with a non-periodical inspected condition-based opportunistic preventive-maintenance strategy;a long-term average cost model is established using the proposed method to determine its optimal maintenance parameters jointly, in which “hard failure” and non-negligible maintenance time are considered. Numerical experiments indicate that the extended DSSP method is valid for opportunistic maintenance modeling of multi-unit systems.