LM-CDBN超高层变形预测模型的构建与应用

doi:10.3969/j.issn.0258-2724.20180293

邱冬炜^1,2,,
王彤^1,,
段明旭¹,
罗德安^1,2,,,
王来阳¹

基金项目:国家重点研发计划资助项目（2017YFB0503700）；中国住房和城乡建设部科学技术计划资助项目（2015-K8-050）

详细信息

作者简介:
邱冬炜（1978—），男，副教授，博士，研究方向为建构筑物变形监测与结构检测、精密工程测量与数据处理，E-mail：qiudw@bucea.edu.cn

通讯作者:
罗德安（1968—），男，教授，博士，研究方向为工程测量、智能测绘技术，E-mail：luodean@bucea.edu.cn

中图分类号:P258
计量
- 文章访问数:786
- HTML全文浏览量:403
- PDF下载量:19
- 被引次数:0
出版历程
- 收稿日期:2018-05-15
- 修回日期:2018-06-06
- 网络出版日期:2018-07-08
- 刊出日期:2020-04-01

Construct and Application of LM-CDBN Deformation Prediction Model for Supertall Buildings

摘要

摘要:为提高超高层建筑变形预测精度，对附有条件的深度信念网络（conditional deep belief network，CDBN）模型中权值及阈值调整方法进行了改进，使用LM （Levenberg-Marquardt）算法作为新的模型定权机制，构建了LM-CDBN网络模型；将构建的LM-CDBN超高层变形预测模型应用于一座298 m超高层建筑中；然后用训练误差、预测值拟合度、预测结果稳定性组成的综合评价体系对模型进行了评价；最后，将LM-CDBN模型分别与深度信念模型（deep belief network，DBN）、极限学习机（extreme learning machine，ELM）、基于无迹卡尔曼滤波的支持回归向量机（unscented Kalman filter-support vector regression，UKF-SVR）进行了预测结果对比. 结果表明：在超高层建筑的变形预测中，相比DBN、ELM和UKF-SVR，LM-CDBN预测精度分别提升了32%、55%及24%，模型的信息提取稳定性及处理时变系统非线性问题的泛化能力得到了提高.
- 深度信念网络/
- 变形/
- 预测/
- 数据处理/
- 参数估计
Abstract:In order to improve the prediction accuracy of supertall building deformation, the method of adjusting the weight and threshold in the conditional deep belief network (CDBN) model was improved. The LM (Levenberg-Marquardt) algorithm was used as a weighting method to construct the LM-CDBN network model. This method was applied to the deformation prediction of a 298 m supertall building. Then, the model was fully evaluated in terms of training error, goodness of fit, and prediction stability. Finally, the prediction results of LM-CDBN model, deep belief network (DBN) model, extreme learning machine (ELM) and unscented Kalman filter-support vector regression (UKF-SVR) were compared. The result shows that the prediction performance of LM-CDBN was 32%, 55% and 24% higher than three other models respectively. LM-CDBN model improves in the information extraction stability and generalization ability of solving nonlinear problems in time-varying systems.
- deep belief networks/
- deformation/
- prediction/
- data processing/
- parameter estimation

HTML全文

图 1第1层隐含层重构误差

Figure 1.Reconstruction error of the first hidden layer

下载: 全尺寸图片幻灯片

图 2第2～4层重构误差

Figure 2.Reconstruction errors of the second to the fourth hidden layers

下载: 全尺寸图片幻灯片

图 3评价标准误差统计

Figure 3.Evaluation standard errors

下载: 全尺寸图片幻灯片

图 4R统计

Figure 4.Summary ofR

下载: 全尺寸图片幻灯片

表 1A1办公楼监测数据表

Table 1.Monitoring data of A1 office building

序号	时间	位移变形值/mm	温度/℃	风速/（m•s⁻¹）	光照强度/Lx
1	2016-08-23 00：00	28.281	24.1	5.9	2.351
2	2016-08-23 06：00	27.556	23.3	5.4	20.307
3	2016-08-23 12：00	28.428	24.2	5.5	434.243
4	2016-08-23 18：00	29.779	24.9	5.4	267.839
5	2016-08-24 00：00	30.250	24.8	4.9	4.299
6	2016-08-24 06：00	27.895	24.1	5.2	37.456
$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $	$\vdots $
139	2016-09-26 12：00	49.753	31.9	6.7	780.412
140	2016-09-26 18：00	49.341	27.6	7.5	413.485

下载: 导出CSV

表 2预测结果对比表

Table 2.Prediction comparison

观测时间	$\hat y/{\rm{mm}}$	CDBN		LM-CDBN		ELM		UKF- SVR
观测时间	$\hat y/{\rm{mm}}$	预测值/mm	X_RE/%	预测值/mm	X_RE/%	预测值/mm	X_RE/%	预测值/mm	X_RE/%
2016-08-24 12：00	47.510	45.369	4.51	46.214	2.73	41.535	12.58	52.683	10.89
2016-08-24 18：00	46.653	45.173	3.17	45.537	2.39	40.083	14.08	44.297	5.05
2016-08-25 00：00	46.025	44.345	3.65	45.027	2.17	43.629	5.21	46.913	1.93
2016-08-25 06：00	46.769	44.220	5.45	45.626	2.44	45.756	2.17	48.587	3.89
2016-08-25 12：00	46.385	44.342	4.40	45.315	2.31	49.168	6.00	45.179	2.60
2016-08-25 18：00	48.308	46.040	4.69	46.829	3.06	47.051	2.60	50.576	4.69
2016-08-26 00：00	49.739	48.576	2.34	47.891	3.72	44.774	9.98	52.940	6.44
2016-08-26 06：00	52.424	47.736	8.94	49.742	5.12	46.998	10.35	51.630	1.51
2016-08-26 12：00	49.753	47.547	4.43	47.906	3.71	48.372	2.78	48.568	2.38
2016-08-26 18：00	49.341	47.360	4.01	47.603	3.52	47.187	4.37	50.018	1.37
X_ARE/%			4.56		3.12		7.01		4.08

下载: 导出CSV

表 3预测结果评价表

Table 3.Evaluation for different prediction results

评价标准	CDBN	LM-CDBN	ELM	UKF-SVR
X_MSE/mm	4.397 0	2.559 0	15.559 0	5.570 0
X_MRE/mm	0.023 9	0.004 7	0.059 1	0.031 4
X_MAE/mm	0.014 0	0.003 5	0.033 3	0.017 5
X_RMSE/mm	0.017 1	0.009 5	0.041 3	0.021 3
R%	97	98	92	96

下载: 导出CSV

参考文献 (18)

邓勇,张冠宇,李宗春,等. 遗传小波神经网络在变形预报中的应用[J]. 测绘科学,2012,37(5): 183-186.

DENG Yong, ZHANG Guanyu, LI Zongchun, et al. Application of genetic wavelet neural network in deformation forecast[J]. Science of Surveying and Mapping, 2012, 37(5): 183-186.

段明旭,邱冬炜,李婉,等. 改进灰色人工神经网络模型的超高层建筑变形预测[J]. 测绘科学,2017,42(4): 141-146.

DUAN Mingxu, QIU Dongwei, LI Wan, et al. The GM-BPNN prediction research in the deformation forecasting of the super high-rise building[J]. Science of Surveying and Mapping, 2017, 42(4): 141-146.

WANG Xiaoyu, YANG Kan, SHEN Changsong. Study on MPGA-BP of gravity dam deformation prediction[J]. Mathematical Problems in Engineering, 2017, 2017(6): 1-13.

HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786): 504-507.doi:10.1126/science.1127647

CHEN H, MURRAYA F. Continuous restricted Boltzmann machine with an implementable training algorithm[J]. IEE Proceedings—Vision,Image and Signal Processing, 2003, 150(3): 153-158.doi:10.1049/ip-vis:20030362

周晓莉,张丰,杜震洪,等. 基于CRBM算法的时间序列预测模型研究[J]. 浙江大学学报(理学版),2016,43(4): 442-451.doi:10.3785/j.issn.1008-9497.2016.04.011

ZHOU Xiaoli, ZHANG Feng, DU Zhenhong, et al.A study on time series prediction model based on CRBM algorithm[J]. Journal of Zhejiang University (Science Edition), 2016, 43(4): 442-451.doi:10.3785/j.issn.1008-9497.2016.04.011

HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural computation, 2006, 18(7): 1527-1554.doi:10.1162/neco.2006.18.7.1527

CHEN Siqi, AMMAR H B, TUYLS K, et al. Conditional restricted Boltzmann machines for negotiations in highly competitive and complex domains[C]//International Joint Conference on Artificial Intelligence. Beijing: AAAI, 2013: 69-75.

SPILIOPOULOU A. Investigation of deep CRBM networks in modeling sequential data[D]. Edinburgh: University of Edinburgh, 2008.

MNIH V L H, HINTON G E. Conditional restricted Boltzmann machines for structured output prediction[C]//Twenty-Seventh Conference on Uncertainty in Artificial Intelligence. Barcelona: AUAI Press, 2011: 514-522.

周晓莉. 基于深度学习的浙江近岸船舶数据赤潮生物量趋势性预测研究[D]. 杭州: 浙江大学, 2016.

MADSEN K, NIELSEN H B, TINGLEFF O. Methods for nonlinear least squares problems[J]. Society for Industrial & Applied Mathematics, 2004, 2012(1): 1409-1415.

QUESADA-OLMO N, JIMENEZ-MATINEZ M J, FARJAS-ABADIAM. Real-time high-rise building monitoring system using global navigation satellite system technology[J]. Measurement, 2018, 123(1): 115-124.

LECUN Y, BENGIO Y, HINTON G E. Deep learning[J]. Nature, 2015, 521(7553): 436-444.doi:10.1038/nature14539

GAO Jun. Newton-Gauss curvature matrix based cdbnfor online edible fungus drying prediction model[J]. Future Generation Computer Systems, 2017, 81(1): 273-279.

HRASKO R, PACHECO A G C, KROHLING RA. Time series prediction using restricted Boltzmann machines and backpropagation[C]//Information Technology and Quantitative Management. Suzhou: Elsevier, 2015: 990-999.

BA J, HINTON G E, MNITH V, et al. Using fast weights to attend to the recent past[C]//Advances in Neural Information Processing Systems. Barcelona: Elsevier, 2016: 4331-4339.

SRIVASTAVA N, SALAKHUTDINOV R, HINTON G E. Modeling documents with a deep Boltzmann machine[C]//Proceedings of the Twenty-Ninth Conference on Uncertainty in Artificial Intelligence. Bellevue: AUAI press, 2013: 616-624.

施引文献

附加材料 (0)

访问统计