### 1. Introduction

### 2. Unwanted Fire Experiment

### 3. 3D Numerical Analysis

### *3.1 Computational domain*

^{2}are set to produce smoke. Figure 5(c) shows the detector in the computational domain. We obtained the drawings from the smoke detector manufacturer and implemented them in the computational domain through a 3D modeling process. Based on these, we reproduced the phenomenon of smoke passing through the inside of the detector and measured the smoke concentration at the center of the detector. Figure 5(d) shows the locations of the smoke detectors. The smoke detector located directly above the insect repellent bomb was named P1, and detectors were placed at points P1 to P3 in Room 1, P4 to P6 in the corridor, and P7 to P8 in Room 2.

### *3.2 Governing equations and boundary conditions*

*ϵ*-turbulence model [7]. The governing equations [8] used in the analysis are as follows.

##### (5)

*ρ*denotes the density,

*u*and

_{i}*u*denote the velocity components of each coordinate axis, p denotes the pressure,

_{j}*μ*denotes the viscosity modulus, and R denotes the gas constant.

*σ*,

_{k}*σ*,

_{ϵ}*σ*,

_{t}*C*denote experimental constants with values of 1.0, 1.2, 1.0, and 1.9, respectively. For

_{ϵ2}*C*,

_{ϵ1}### *3.3 Mesh setup and numerical analysis method*

^{-3}or less.

### 4. Results

### 5. Conclusion

(1) At detector points P1 to P3, a smoke concentration of 15 %/m was reached in both experiments and numerical analysis, and the reaching time was also similar between the experimental and numerical analysis results.

(2) At detector points P4 to P6, the smoke concentration reached 15 %/m in the experiments but not in the numerical analysis. However, the time at which the smoke concentration began to increase was similar between the experimental and numerical analysis results.

(3) At detector points P7 and P8, a smoke concentration of 15 %/m was not reached in most of the experiments and numerical analysis, showing a similarity between the experimental and numerical analysis results.

(4) In the numerical simulation, the wall surface was assumed to be insulated. However, the temperature on the actual wall surface varies depending on the location on the wall surface. In addition, the degree of roughness on the wall surface was set to smooth in the numerical analysis, unlike the actual wall. This explains the difference in results between the experiments and numerical analysis.