Deep Learning-Driven Data Curation and Model Interpretation for Smart Manufacturing

Characterized by self-monitoring and agile adaptation to fast changing dynamics in complex production environments,smart manufacturing as envisioned under Industry 4.0 aims to improve the throughput and reliability of production beyond the state-of-the-art.While the widespread application of deep learning(DL)has opened up new opportunities to accomplish the goal,data quality and model interpretability have continued to present a roadblock for the widespread acceptance of DL for real-world applications.This has motivated research on two fronts:data curation,which aims to provide quality data as input for meaningful DL-based analysis,and model interpretation,which intends to reveal the physical reasoning underlying DL model outputs and promote trust from the users.This paper summarizes several key techniques in data curation where breakthroughs in data denoising,outlier detection,imputation,balancing,and semantic annotation have demonstrated the effectiveness in information extraction from noisy,incomplete,insufficient,and/or unannotated data.Also highlighted are model interpretation methods that address the“black-box”nature of DL towards model transparency.

Characteristics of seismic responses at liquefied and non-liquefied sites with same site conditions

To understand the characteristics of seismic response at liquefied sites, a liquefiable site and a non-liquefiable site were selected, separated by about 500 m and having the same site conditions as Class D in the National Earthquake Hazards Reduction Program （NEHRP）. A suite of earthquake records on rock sites are selected and scaled to the spectrum of the Joyner, Boore, and Fumal （JBF） attenuation model for a magnitude 7.5 earthquake at a distance of 50 km. The scaled records were then used to excite the two sites. The effect of pore-water pressure was investigated using the effective stress analysis method, and nonlinear soil behavior was modeled by a soil bounding surface model. Comparisons for spectra, peak ground acceleration （PGA）, peak ground displacement （PGD） and permanent displacement were performed. Results show that the mean ground response spectrum at the non-liquefied site is close to the estimated ground response spectrum from the JBF model, but the mean ground response spectrum at the liquefied site is much lower than the estimated ground response spectrum from the JBF model for periods of up to 1.3 s. The mean PGA at the non-liquefied site is about 1.6-1.7 times as large as that at the liquefied site, but the mean peak ground displacement （PGD） at the non-liquefied site has a slight difference with that at the liquefied site. The mean permanent displacements at the liquefied site are larger than those at the non-liquefied site, particularly at the liquefied layer.