PSI - Issue 50

I. Shardakov et al. / Procedia Structural Integrity 50 (2023) 257–265 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

258

2

lead to local subsidence of the earth's surface and cause structural damage. Deformation processes that occur on the earth's surface and in underlying soils are usually controlled by geodetic and geotechnical monitoring methods, Zhang and Zhang (2016), Mazzanti (2017). The deformation state of individual buildings or structures located in such areas is most often assessed by monitoring uneven settlements (vertical displacement) of foundations. For this, various geodetic methods can be used, Pardo et al (2016), including the method of hydrostatic leveling, Shardakov et al. (2015), (2016). Usually, the implementation of instrumental observations or the installation of systems for continuous monitoring of the structures is associated with the appearance of visible damage to the structure, the occurrence of cracks, etc. At the same time, users are primarily interested in the question of how long the structure can be used in the conditions of developing geophysical processes. Thus, researchers are faced with the task, based on data on deformation processes on the earth's surface observed over a relatively short period, to predict their development in the future and make a conclusion about the possibility of safe operation of the structure for a certain period. A reliable forecast of deformation processes allows timely implementation of the necessary repair measures (repair, strengthening of the structure or timely dismantling of the structure) and avoid undesirable events. The processes of soil subsidence and settlement of building foundations can develop according to different scenarios. They can proceed slowly over a long period (over several years or decades), or, on the contrary, begin to develop rapidly. One of the typical evolutionary scenarios is described by the “S - shaped curve”, Nie et al. (2015) . The initial stage of this scenario is characterized by an increase in the sediment rate from zero to a certain constant value, then the sediment rate remains constant for a certain time, and finally it decreases to almost zero. Seasonal changes in the state of the soil can be added to these processes; moreover, the recorded data have a random scatter caused by the measurement features. Observational data on changes in the settlement of building foundations are time series, i.e., ordered sequences of data that change over time. In general, time series describe an extremely wide range of phenomena, for example, changes in economic indicators, parameters of various physical processes, etc. Statistical analysis of time series allows you to identify the main patterns of change in the studied value and, on their basis, make a forecast of its development. Currently, statistical methods and predictive modeling are being actively developed for the analysis of such data arrays, Hyndman and Athanasopoulos (2018). In the practice of analyzing the state of building structures, methods based on models of exponential smoothing, autoregressive moving average neural networks, are most widely used, see, for example, Vorobev and Kashevarova (2013), Cui and Ren (2014), Tang et al. (2019), Yan et al. (2020). They are used to predict vertical displacements and deformations, Yanget al. (2019), Yang et al. (2019). This article discusses the experience of using some statistical forecasting models to analyze the trends in the development of uneven building settlements. The data for such a study were obtained during long-term monitoring of a group of 40 residential buildings located on the territory above the potash ore mining zone, see Shardakov et al. (2016), (2019), Gusev et al. (2022). To monitor the deformation state of these buildings, an automated monitoring system was installed on each of them, based on the measurement of foundation settlements using hydraulic level sensors. During the operation of this system, the values of relative vertical displacements at the characteristic points of the foundation were recorded. This data were then used in the analytical block of the monitoring system to simulate the stress-strain state of the entire building using the finite element method, Glot (2013). Foundation settlements were monitored continuously for 10 years with a frequency of measurements several times a day. It was noticed that the nature of deformation processes in different parts of the undermined territory is different. Among the observed group of buildings there are both objects with a uniform settlement rate, and with accelerating or decelerating settlements. Figure 1a shows the layout of hydraulic level sensors at one of the monitoring objects, which is a brick 3-storey building 30 m long and 12 m wide. Figure 1b represents the change in relative sediment recorded by each sensor during 2 years of observation. The numbers of the curves correspond to the numbers of the sensors in Figure 1a. It is interesting to analyze the obtained extensive monitoring data on the basis of statistical methods and to predict the development of the observed processes. The purpose of such a study is to select an adequate statistical forecasting model and assess the possibility of including the results of forecasting in the monitoring system. Long term monitoring data can be used to test various statistical prediction models. In this case, one part of the monitoring data is used for statistical analysis of the settlement process and forecasting, and the other part is used to compare the predicted and actually observed data.