Reconstruction of the gray belt objects based on energy efficiency clusters

Objects redevelopment methods of the “gray belt” - industrial areas surrounding the historical district of St. Petersburg, Russia - have been considered. Information about 45 objects located in different administrative districts of the city was collected. Factors of physical wear (wear of floors, walls, roofs, types of building structural system) have been chosen as a criteria for clustering. As a result of the study, SOMs with different learning parameters were created as a result of the study. Energy efficiency calculations for two clusters were made with the selection of modern materials. Recommendations for the reconstruction according to the parameters of physical wear are given.


Introduction
The issue of the development of old industrial zones is relevant today. The reorganization of industrial buildings and their territories can create new places of attraction for visitors and citizens. This will help to stimulate the economics, disperse the tourist centre and reduce the traffic load [1][2][3][4][5]. Also the reorganization of this area will expand the network of polycentres with a developed horizontal infrastructure. This will contribute to improving the quality of life thanks to a reasonable and comfortable arrangement of the territory, creating an attractive, balanced environment for work, living and recreation of the citizens [6][7][8][9][10][11].
The industrial zone of St. Petersburg, Russia or the "gray belt", is the largest reserve for urban development near the historic centre of the city. The area of industrial buildings in St. Petersburg is about 50 million m 2 . According to various estimates, this zone occupies about 40% of the total area of the city.
Industrial territories are located in 12 administrative districts of St. Petersburg. The gray belt is outlined by the coast of the Neva Bay, it includes a port complex in the southern part of Kanonersky Island. It includes industrial and public business buildings in Kirovsky, Moskovsky, Frunzensky and Nevsky districts, within the territory outlined by the Obvodny Canal and Leninsky Prospekt, Tipanova Street and Slavy Prospekt. Also there are industrial buildings remained on the right bank of the Neva from Narodnaya Street to Energetikov * Corresponding author: talililiva@gmail.com Prospekt in the southern part of Krasnogvardeisky district. At the moment, most of the objects of the «Gray belt» are in need of reconstruction.
Achieving the level of quality and comfort of buildings, the standards of thermal protection during the reconstruction is much more difficult than with new construction. For example, it is difficult to significantly change the planning decisions, the material of the building envelope, etc.
From the point of view of architectural and construction solutions, the main indicator of energy efficiency of a building is resistance to heat transfer of building envelope. This parameter determines the temperature of the surfaces of the building envelope, affecting the feeling of comfort.
The method of calculating this indicator has changed significantly since the 1970s. In 2003, amendments appeared in the thermotechnical standard, which consider the climatic features of various regions of the country.
The purpose of this work is to create clusters according to the characteristics of physical wear and type of construction, to provide each cluster with calculation of energy efficiency during the reconstruction [12 -14].
Tasks: -to identify gray belt objects; -to determine the types of structures for each object; -to determine the wear of structures for each object; -to group the gray belt objects into clusters; -to carry out the calculation of energy efficiency during the reconstruction of each cluster.

Materials and Methods
To determine the effective strategy of redevelopment of gray belt objects, it is necessary to develop an approach for classifying objects included in these zones. In this article, it is proposed to use self-organizing maps for clustering objects.
A self-organizing map (SOM) or self-organizing feature map (SOFM) -is a type of artificial neural network (ANN), that is trained using uncontrollable learning to produce a low-dimensional (typically two-dimensional) discretized representation of the input space of the training samples, called a map, and therefore it is a method of reducing the dimension. They were developed in 1982 by Tuevo Kohonen, an honorable professor of the Academy of Finland [15]. Self-organizing map (SOM) -is an efficient tool for neural network modelling for visualization and generalization of multidimensional data. It is suitable for solving complex problems like process analysis, machine perception, control and information transfer [16].
The Self-Organizing Map algorithm can be divided into 6 steps [15]: 1. The weights of each node are determined. 2. The input vector is randomly selected from the training data set and presented to the network. 3. Each node of the network tests for matching to the input vector. Winning node becomes «Best Matching Unit» (BMU). 4. The range of the BMU is calculated. This value starts with the highest one. It is usually set as the network radius, decreasing with each iteration. 5. All nodes found in the radius of the «Best Matching Unit» neuron, calculated in the item 4, are configured so that they are more similar to the input vector. His weights change more if it's closer to the BMU. 6. Repeat item 2 for N iterations. 45 buildings of the gray belt located in different parts of the city were selected as objects of research [17]. 4-5 objects were selected from each district for further research. The distribution of the researched objects by districts is shown in Figure 1.

Fig. 1. Number of research objects by districts
The following information was collected for each object: 1. Physical wear characteristics (roof wear, floor wear, wall wear). 2. Type of construction (roof materials, floor materials and wall materials). These characteristics are shown in Table 1.

Results
Total selection of the objects were 30 buildings of Saint-Petersburg industrial areas after pre-processing of the data and evaluation of their quality in the software named Deductor. The indicators of physical wear and types of structures with the extreme values and emissions were considered. Objects with extreme values were excluded from research.
The program Deductor analyzed the factors of physical wear. The results of the analysis are shown in Figure 2. The x-axis shows the values of each test criteria. The total number of clusters is seven ( Table 2). Types of construction were also considered. During the research, the data were divided into two clusters (Table 3). Kr-5

Discussion
The program has divided the objects into 7 clusters depending on the indicators of physical deterioration of structural elements of buildings. With the increase of the cluster number, the indicators of physical wear decrease. Characteristics of physical wear clusters received in the program Deductor can be reduced thus it will reduce their number. Recommendations can be made for follow-up actions based on the results (Table 4).
It is required to conduct security and support activities. These buildings are the object of cultural heritage and need to be reconstructed V-3 Kr-5 Two clusters were identified according to the design characteristics. Then the thermotechnical calculation of the wall and selection of insulation were made for the reconstruction. Composition of the existing outer wall is shown in Table 5. Heating degree-day: According to SP 50.133330.2012 «Building heat insulation» HDD = 4749.9 °С·day Normalized heat transfer resistance of the outer wall: where α, β -design factors, according to SP 50.133330.2012 «Building heat insulation»: α = 0,00035; β = 1,4. Rn = 0.00035 ⋅ 4749.9 + 1.4 = 3.063 m 2 ⋅°С/W Design heat transfer resistance of the outer wall:  10 10 SPbWOSCE

21
The calculation value of the surface heat exchange resistance is significantly lower than normalized value. Consequently, the existing design of the outer wall does not meet modern requirements for thermal protection of buildings.
Composition of the outer wall. Selection of the insulation is shown in Table 6. The calculation value of the surface heat exchange resistance is significantly above normalized value. Consequently, the existing design of the outer meets modern requirements for thermal protection of buildings and can be used in the buildings reconstruction.
Composition of the outer wall is shown in Table 7.  The calculation value of the surface heat exchange resistance is above normalized value. It is equal to the results obtained for mineral wool boards 100 mm thick. Consequently, the existing design of the outer meets modern requirements for thermal protection of buildings and can be used in the buildings reconstruction.