PIDAS工艺评估和工艺设计模型开发  

Process assessment and process design on PIDAS data platform

作　　者：陈欢欢 刘晔[1] 闫博[1] CHEN Huanhuan;LIU Ye;YAN Bo(Research Institute,Baoshan Iron & Steel Co. ,Ltd. , Shanghai 201999, China)

机构地区：[1]宝山钢铁股份有限公司中央研究院,上海201999

出　　处：《宝钢技术》2021年第1期7-13,共7页Baosteel Technology

摘　　要：随着PIDAS系统在宝钢厚板的建立与应用,生产、工艺、设备、性能数据得到了系统的采集、整理,海量工业数据的价值得到了有效的挖掘,开始发挥重要作用。基于PIDAS,首先开发了一种工艺评估模型,与传统的性能预测模型相比,它能根据工艺参数,预测产品性能值所在的区间,以及真实值位于此区间内的概率。此模型使用了自主开发的概率区间估计算法,通过全面评估不同机器学习回归模型的效果、充分考虑历史生产数据的分布规律,来预测产品性能值所在的区间,取得了良好的效果。在此基础上,又开发了一种工艺设计模型,它能够根据化学成分、产品规格和性能目标,设计出最优的轧制工艺参数。此模型应用了自主开发的自定义权重的模拟退火算法。所提出的工艺评估和工艺设计模型均取得了良好的效果,达到产品质量预报和优化工艺设计的目的。With the deployment of PIDAS data platform,a great variety of data generated from production,process,facility and mechanical properties can be systematically collected,cleaned and analyzed.Based on PIDAS,a process assessment model is firstly developed.Different from traditional regression model in machine learning,it enables process designers to assess the interval that mechanical properties probably lie in,as well as the possibility,according to the process parameters.It is based on a probabilistic interval prediction algorithm proposed.By fully comparing the performance of different common regression algorithms and using the distribution of historical product data,this algorithm reaches a high accuracy.Based on it,a process design model was then developed.It can design the best temperature parameters according to the desired mechanical properties,sizes and ingredients.A self-designed weighted simulated annealing algorithm is proposed to build the model.Both of the two models are aimed to achieve the goal of accurate assessment of products’property and process design.

关 键 词：工艺评估 工艺设计 PIDAS 

分 类 号：TG334.9[金属学及工艺—金属压力加工]