基于PBT-DeepTCN和数字孪生的烧结终点多步预测

doi:10.12068/j.issn.1005-3026.2026.20240126

摘要/Abstract

摘要：

烧结终点位置是影响烧结矿质量和生产效率的关键参数.针对烧结终点预测中存在的指导性不足、时效性差和可视化效果弱等问题，本文构建了包括物理实体、虚拟环境、多步预测、孪生数据和虚实连接在内的数字孪生五维模型，为烧结过程提供工艺参数监控和优化指导.在预测方面，首先进行数据预处理，然后采用灰色关联度分析(GRA)筛选特征变量，最后利用基于群体的训练方法(PBT)优化的深度时间卷积网络(DeepTCN)对烧结终点进行多步预测.实验结果表明，所提数字孪生模型在不同预测步长下具有较高预测精度，为烧结领域数字化、智能化转型提供了先进思路与技术方法.

关键词: 烧结终点, 多步预测, 数字孪生, 深度时间卷积网络, 超参数优化

Abstract:

The position of the sintering terminal point is a key parameter that affects the quality and production efficiency of sinter. To improve insufficient guidance， poor timeliness， and weak visualization effect in sintering terminal point prediction， a five-dimensional digital twin model was constructed， including physical entity， virtual environment， multi-step prediction， twin data， and virtual and real connection， which provided process parameter monitoring and optimization guidance for the sintering process. In terms of prediction， the data was first preprocessed， and then the feature variables were screened by grey relation analysis （GRA）. Finally， the deep temporal convolutional network （DeepTCN）by using population based training（PBT） was constructed for multi-step prediction of the sintering terminal point. The experimental results show that the proposed digital twin model has high prediction accuracy under different prediction steps， and it provides advanced ideas and technical methods for digital and intelligent transformation in the sintering field.

Key words: sintering terminal point, multi-step prediction, digital twin, deep temporal convolutional network, hyperparameter optimization

中图分类号:

TF 046

宋小龙, 栗潇通, 杨欢, 吴朝霞. 基于PBT-DeepTCN和数字孪生的烧结终点多步预测[J]. 东北大学学报（自然科学版）, 2026, 47(1): 99-106.

Xiao-long SONG, Xiao-tong LI, Huan YANG, Zhao-xia WU. Multi-step Prediction of Sintering Terminal Point Based on PBT-DeepTCN and Digital Twin[J]. Journal of Northeastern University(Natural Science), 2026, 47(1): 99-106.

图/表 10

图1 烧结工艺流程图

Fig.1 Flow chart of sintering process

图2 数字孪生系统框架

Fig.2 Digital twin system framework

表1 特征变量灰色关联度排序 (variable)

Table 1 Grey relational degree ranking of feature

序号	变量名称	灰色关联度
1	20#风箱废气温度/ºC	0.277 1
2	焦粉下料量/( $t ⋅ h - 1$ )	0.274 2
3	7#风箱负压/kPa	0.267 4
4	7#风箱废气温度/ºC	0.243 4
5	1#风箱废气温度/ºC	0.230 6
6	w(混料氧化钙)/%	0.222 1
7	3#风箱废气温度/ºC	0.190 8
8	11#风箱负压/kPa	0.189 1
9	13#风箱废气温度/ºC	0.186 1
10	w（混料水分）/%	0.176 1
11	2#风箱废气温度/ºC	0.172 3
12	1#风箱负压/kPa	0.164 2
13	w(混料二氧化硅)/%	0.161 7
14	烧结机机速/( $m ⋅ m i n - 1$ )	0.161 3
15	2#风箱负压/kPa	0.159 7

表1 特征变量灰色关联度排序 (variable)

Table 1 Grey relational degree ranking of feature

序号	变量名称	灰色关联度
1	20#风箱废气温度/ºC	0.277 1
2	焦粉下料量/( $t ⋅ h - 1$ )	0.274 2
3	7#风箱负压/kPa	0.267 4
4	7#风箱废气温度/ºC	0.243 4
5	1#风箱废气温度/ºC	0.230 6
6	w(混料氧化钙)/%	0.222 1
7	3#风箱废气温度/ºC	0.190 8
8	11#风箱负压/kPa	0.189 1
9	13#风箱废气温度/ºC	0.186 1
10	w（混料水分）/%	0.176 1
11	2#风箱废气温度/ºC	0.172 3
12	1#风箱负压/kPa	0.164 2
13	w(混料二氧化硅)/%	0.161 7
14	烧结机机速/( $m ⋅ m i n - 1$ )	0.161 3
15	2#风箱负压/kPa	0.159 7

图3 多步预测过程（a）—自回归预测方法；（b）—非自回归预测方法.

Fig.3 Multi-step prediction process

图4 DeepTCN结构

Fig.4 Structure of DeepTCN

图5 编码器和解码器模块（a）—编码器模块；（b）—解码器模块.

Fig.5 Encoder and decoder modules

表2 超参数的调整范围

Table 2 Adjustment range of hyperparameters

超参数名称	寻优范围
学习率	[0.000 1 0.01]
批处理大小	[4 32]
卷积核大小	[25]
滤波器数量	[110]
输入序列长度	[10 50]

图6 烧结终点预测结果

Fig.6 Prediction results of sintering terminal point

表3 不同预测步长下不同模型对比结果 (under different prediction steps)

Table 3 Comparison results of different models

预测步长	对比模型	$e R M S$	$e M A$	R²
t+1	CNN	0.035 1	0.028 0	0.410 2
	LSTM	0.022 8	0.018 8	0.752 6
	DeepTCN	0.013 4	0.011 3	0.913 8
	PBT- DeepTCN	0.010 4	0.008 0	0.948 3
t+4	CNN	0.038 6	0.027 7	0.288 0
	LSTM	0.029 3	0.023 8	0.590 6
	DeepTCN	0.014 0	0.010 3	0.824 5
	PBT-DeepTCN	0.018 8	0.016 3	0.882 6
t+8	CNN	0.042 9	0.031 0	0.120 2
	LSTM	0.032 9	0.027 3	0.484 4
	DeepTCN	0.023 8	0.020 3	0.725 2
	PBT-DeepTCN	0.021 0	0.017 8	0.769 1

表3 不同预测步长下不同模型对比结果 (under different prediction steps)

Table 3 Comparison results of different models

预测步长	对比模型	$e R M S$	$e M A$	R²
t+1	CNN	0.035 1	0.028 0	0.410 2
	LSTM	0.022 8	0.018 8	0.752 6
	DeepTCN	0.013 4	0.011 3	0.913 8
	PBT- DeepTCN	0.010 4	0.008 0	0.948 3
t+4	CNN	0.038 6	0.027 7	0.288 0
	LSTM	0.029 3	0.023 8	0.590 6
	DeepTCN	0.014 0	0.010 3	0.824 5
	PBT-DeepTCN	0.018 8	0.016 3	0.882 6
t+8	CNN	0.042 9	0.031 0	0.120 2
	LSTM	0.032 9	0.027 3	0.484 4
	DeepTCN	0.023 8	0.020 3	0.725 2
	PBT-DeepTCN	0.021 0	0.017 8	0.769 1

图7 数字孪生可视化界面

Fig.7 Digital twin visual interface

参考文献 12

[1]	Du S， Wu M， Chen X， et al. Intelligent integrated control for burn-through point to carbon efficiency optimization in iron ore sintering process［J］. IEEE Transactions on Control Systems Technology， 2020， 28（6）： 2497-2505.
[2]	Tamura N， Konishi M， Morita T， et al. Mathematical approach for the optimization of the sintering process operation［J］. IFAC Proceedings Volumes， 1987， 20（8）： 203-208.
[3]	Liu S， Lyu Q， Liu X J， et al. A prediction system of burn through point based on gradient boosting decision tree and decision rules［J］. ISIJ International， 2019， 59（12）： 2156-2164.
[4]	Hu J， Wu M， Cao W H， et al. Soft-sensing of burn-through point based on weighted kernel just-in-time learning and fuzzy broad-learning system in sintering process［J］. IEEE Transactions on Industrial Informatics， 2024， 20（5）： 7316-7324.
[5]	Toktassynova N， Fourati H， Suleimenov B. Modelling and control structure of a phosphorite sinter process with grey system theory［J］. The Journal of Grey System， 2020， 32（2）： 150-166.
[6]	Wang D D， Yang K， He Z J， et al. Application research based on GA-FWA in prediction of sintering burning through point［C］//Proceedings of 2018 International Conference on Computer， Communications and Mechatronics Engineering. Xiamen，2018： 378-385.
[7]	Xie Y H， He B C， Zhang X M， et al. A decomposition-based encoder-decoder framework for multi-step prediction of burn-through point in sintering process［C］//2023 IEEE 6th International Conference on Industrial Cyber-Physical Systems （ICPS）. Wuhan， 2023： 1-6.
[8]	Yan F， Yang C J， Zhang X M， et al. A 3D convolution-based burn-through point multistep prediction model for sintering process［J］. IEEE Transactions on Industrial Electronics， 2024， 71（4）： 4219-4229.
[9]	Liu M N， Fang S L， Dong H Y， et al. Review of digital twin about concepts， technologies， and industrial applications［J］. Journal of Manufacturing Systems， 2021， 58： 346-361.
[10]	Bajic B， Rikalovic A， Suzic N， et al. Industry 4.0 implementation challenges and opportunities： a managerial perspective［J］. IEEE Systems Journal， 2021， 15（1）： 546-559.
[11]	Brynjolfsson E， Mitchell T. What can machine learning do？ Workforce implications［J］. Science， 2017， 358（6370）： 1530-1534.
[12]	Lyu W J， Liu J. Artificial intelligence and emerging digital technologies in the energy sector［J］. Applied Energy， 2021， 303： 117615.