Multi-Sensor Data Fusion Algorithm for Indoor Fire Early Warning Based on BP Neural Network

Round 1

Reviewer 1 Report

This paper deals with an indoor fire early warning algorithm based on back propagation neural network. The early warning algorithm fuses the data of temperature, smoke concentration and carbon monoxide, and outputs the probability of fire occurrence. After describing the phenomena of fire occurrence, the authors define the fire characteristics as the measured variables that carry the information about fire occurrence.  Then, the improved kendall -t algorithm is used to extract the trend values of preprocessed data and BP neural network to fuse the values and trends of fire parameters is considered for data classification.

The paper is well organized and easy to understand, it deals with a relevant topic and obvious usage. The experimental results show the relevance of the choice of tools. Although the tools used are known in the literature, their combination and the application support can be considered a contribution.

Cependant les points suivants peuvent être améliorés :

1- The choice of Kendall's algorithm must be better argued.As this is a filtering method, there are some assumptions to consider when setting it up.

2- The BBNN allows a short-term estimate of the outputs but does not allow a long-term estimate.Authors should specify the horizon of the estimate.That is, the time interval remaining between the generation of the alarm and the occurrence of the fire.

3- The introduction must be improved with review papers, such as,"Review of Health Indices extraction and Trend Modeling methods for Remaining Useful Life Estimation."Book Chapter Springer Nature Switzerland AG 2020.

4- Techniques for analyzing the performance of a diagnostic and prognostic algorithm exist, and which it would be good to use to evaluate the performance of the proposed approach, such as, for example, Metrics for evaluating 1142 performance of prognostic techniques, in International Conference on Prognostics and Health 1143 Management (PHM08), pp. 1–17 (2008) .

Author Response

Please see the attachment，thank you.

Author Response File: Author Response.docx

Reviewer 2 Report

Authors have illustrated some results to enhance the ability of the proposed algorithm. However, authors can compare the effects of various neurons numbers and find the best number of neurons.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

Sugestions:

1. Introduction section should be extended. Authors should analyze state-of-the-art ANN-based CO predictors.
2. Authors should add the main contribution of this paper in the end of the Introduction section. After that it would be good to see the remindner of this paper
3. Authors should provide a comparison with other ANN topologies or ML-based algorithm
4. Conclusion section should be extended using: 1) numerical results obtained in the paper; 2) limitations of the proposed approach; 3) prospects for the future research.

Author Response

Please see the attachment. thank you.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Paper can be accepted except line 235

There is no Kolrnogorov but the Kolmogorov theorem.