CFD analysis with the RANS turbulence model is often used in the field of architectural facility design as one of the convincing design techniques. However, there is no RANS turbulence model that provides practical solutions for all flow fields. Therefore, it is necessary to select an appropriate turbulence model depending on the flow of interest to obtain reliable solutions. Hence, there have been several studies for the purpose of evaluating the performance of prediction accuracy and characteristics of the turbulence models. One of the studies reports that the RNG k-ε model is better for forced convection flow in a room. On the other hand, another study reports that the realizable k-ε model is better for an axial jet and the RNG k-ε model has the same problem for the axial jet as that in the standard k-ε model. Authors doubted that the difference of the two results had been caused by the difference in the Re number of the flow field where each study had focused. Therefore, for the isothermal axial jet in a room, we performed CFD analyses with four RANS turbulence models, namely the standard k-ε, RNG k-ε, realizable k-ε, and SST k-ω. The computational regions were two flow fields. One of those had a low Re number for the inlet, while the other had a high Re number for the inlet. The RANS turbulence models were evaluated by comparing their prediction performances with experimental data or results of theoretical calculations for the jet. Accordingly, the present study revealed a characteristic of realizable k-ε for the flow field with a low Re number. Moreover, it showed that RNG k-ε or SST k-ω was better for the axial jet in a room.

For a heat source system in a building, it is very important to detect and diagnose faults. In this research, a fault is defined as a factor that makes it impossible to perform as designed, and fault detection and diagnosis is defined as locating faults. The purpose of this research is to detect and diagnose faults in a heat source system by simulation and machine learning, and it is assumed that the heat source system in actual operation has unknown faults as a precondition. Especially in this paper, we aimed to detect and diagnose faults using convolutional neural networks (CNN). Firstly, we generated a database with appropriate labels by reasonably calculated six fault states, which are similar for a particular pump operation by detailed heat source system simulation that outputs more than 100 variables such as water flow rate, temperature and power consumption. The calculated data of 1 state 1 day was converted into a 96 × 108 matrix, which was taken as 1 data element. Next, this database was used as learning or test data to detect and diagnose the faults by CNN. We confirmed that it achieved high accuracy (more than 99 %) by using sufficient learning data. Then, as a reference, real data was input into the trained CNN. The probability of detecting the target fault was 65.5 %. Using this method, it is possible to narrow down the fault whose possibility is relatively high. However, in order to apply the proposed method in a real system, there are future tasks as below: to extend the database, to detect multiple faults and the extent of faults, and develop a method or concept that solves problems caused by the inevitable deviation between a simulation and a real system. Although we focused on fault detection and diagnosis in this paper, the proposition of strategies for rectifying faults is also included in future tasks along with the development of indices of faults for the strategies.

In buildings, the design of the heat source system and calculation of the heat source equipment capacity are carried out under the assumption of maximum load. It is important to evaluate the operation method according to the building demand after introduction, and commissioning is one of the effective measures for energy saving. As a practical example, we focused on the heat source system of a university research facility in this study. We investigated the relationship between power or gas consumption converted to primary energy consumption unit and heat demand required for the heat source system in two cases using numerical analysis. Effective information on operational improvement was obtained by examining and verifying the optimal operation method for each required heat demand.

We studied the number of occurrences of failures & troubles and these periods required for restoration, using records of the building maintenance in the hospital. This study presents the survey results of the number of failures & troubles recorded, number of those of new occurrences and number of those in which restoration was carried out the next day, and, in the sanitation of a certain large-scale hospital. The data is divided into failures & troubles found in sickrooms, staff corners, outpatient service rooms with other consultation rooms, medical checkup center, kitchens, garbage processing room, and waste water treatment equipment room, including routine works of building maintenance, and medical sanitation systems. In this report, we show the result of the first four of the above places. We show by the next report, the result of the remainder, and the analysis of the mean restoration period. The purpose of this study is to present basic materials to evaluate the availability of sanitary installation and to plan the organization of maintenance staff.