Local multi-model integrated soft sensor based on just-in-time learning for mechanical properties of hot strip mill process

The mechanical properties of hot rolled strip are the key index of product quality,and the soft sensing of them is an important decision basis for the control and optimization of hot rolling process.To solve the problem that it is difficult to measure the mechanical properties of hot rolled strip in time and accurately,a soft sensor based on ensemble local modeling was proposed.Firstly,outliers of process data are removed by local outlier factor.After standardization and transformation,normal data that can be used in the model are obtained.Next,in order to avoid redundant variables participating in modeling and reducing performance of models,feature selection was applied combing the mechanism of hot rolling process and mutual information among variables.Then,features of samples were extracted by supervised local preserving projection,and a prediction model was constructed by Gaussian process regression based on just-in-time learning(JITL).Other JITL-based models,such as support vector regression and gradient boosting regression tree models,keep all variables and make up for the lost information during dimension reduction.Finally,the soft sensor was developed by integrating individual models through stacking method.Superiority and reliability of proposed soft sensors were verified by actual process data from a real hot rolling process.

Size distribution,directional source contributions and pollution status of PM from Chengdu,China during a long-term sampling campaign

Long-term and synchronous monitoring of PMIo and PM2.s was conducted in Chengdu in China from 2007 to 2013. The levels, variations, compositions and size distributions were investigated. The sources were quantified by two-way and three-way receptor models （PMF2, ME2-2way and ME2-3way）, Consistent results were found： the primary source categories contributed 63.4% （PMF2）, 64.8% （ME2-2way） and 66.8% （ME2-Bway） to PMIo, and contributed 60.9% （PMF2）, 65.5% （ME2-2way） and 61.0% （ME2-3way） to PM2.s. Secondary sources contributed 31.8% （PMF2）, 32.9% （ME2-2way） and 31.7% （ME2-3way） to PMIo, and 35.0% （PMF2）, 33.8% （ME2-2way） and 36.0% （ME2-3way） to PM2.s. The size distribution of source categories was estimated better by the ME2-3way method. The three-way model can simultaneously consider chemical species, temporal variability and PM sizes, while a two-way model independently computes datasets of different sizes. A method called source directional apportionment （SDA） was employed to quantify the contributions from various directions for each source category. Crustal dust from east-north-east （ENE） contributed the highest to both PM^o （12.7%） and PMzs （9.7%） in Chengdu, followed by the crustal dust from south-east （SE） for PMao （9.8%） and secondary nitrate ＆ secondary organic carbon from ENE for PMzs （9.6%）. Source contributions from different directions are associated with meteorological conditions, source locations and emission patterns during the sampling period. These findings and methods provide useful tools to better understand PM pollution status and tn dovolon offoctive nolhltion control gtrateMeg.