A Triple-Correlation Framework for Demonstrating an Innovative Gas Warning System: A Case Study in China

This research aims to develop an innovative gas warning system for improving production safety in the underground coal mining industry. Coal mining is an important sector for China’s economic development; gas monitoring systems are widely adopted as almost 60% of coal mining accidents are caused by gas. Existing gas monitoring systems mainly focus on simply detecting real-time data obtained from gas sensors. The literature review did not nd gas monitoring systems that provide alarms or warnings from the correlation of gas with data from other sensors. This research aims to ll this gap to uncover hidden patterns and correlations between gas and temperature, wind, and dust, and incorporate data analytics into developing an innovative, integrated gas warning system. Correlational research was adopted using 328,320 data outputs obtained from ZhongXing Co. Ltd in Dec 2019. The study found strong relations between gas and temperature, wind, and dust. A Triple-Correlation Theoretical Framework and a Unied Modeling Language (UML) model were developed for an innovative gas warning system. As a result of this research, the ZhongXing Innovative Gas Warning System was developed and deployed for user acceptance testing on 28 Aug 2020. operation factors, and gas factors, including absolute gas emission, gas concentration, and gas release initial velocity. The analysis results showed that the soft and fallen coal seam's most sensitive factors, soft layer thickness variation, change of coal thickness, coal seam thickness, and geological structure but not explore any correlation analysis of gas factors and temperature, wind, and dust. Ma & Dai (2017, p.93) developed a warning index system for coal mine safety based on safety management, facility performance, behavior Monitor, and emergency rescue.


Introduction
The underground coal mining industry is an important sector for China's economic development. The proportion of coal in China's energy consumption is approximately 62% in 2020 and is expected to be 55% by 2030 (Xie et al., 2019, p.169). Most coal seams are now deep and require underground coal mining that currently accounts for about 60% of world coal production (Priyadarsini et al., 2018, p.1). Underground coal mining has hazards of roof collapse, falling rocks in high-stress mines, release, and explosions of gas Gas monitoring systems are adopted widely to reduce risks of gas exceeding concentration limits in China's coal mining industry (Li, Jiang & Zhang 2018). China's State Administration of Work Safety regulations requires a monitoring system to alert the emergency response team if the gas concentration is more than the limit (Xia, Chen & Wei 2018, p.4). In practice, existing gas monitoring systems mainly focus on simply detecting real-time data obtained from methane gas sensors -(called gas data in this research). If the gas data outputs reach the threshold limit values (TLVs), the gas monitoring system will provide alarms or warnings to the safetyresponsive team. Since gas monitoring systems were deployed in China's coal mine industry, the gas accidents have decreased from Even small gas accidents can trigger explosions resulting in injury or loss of life and damage mine tunnels, mine shafts, and expensive equipment (Cheng 2018). Underground coal exploitation is a high-risk industry linked with methane gas hazards (Tutak & Brodny 2019, p.1). Early warning systems can play a crucial role in avoiding gas accidents (Zhao et al. 2020(Zhao et al. , p.1981) and need to be continually improved and deployed in coal mining (Jo, Khan & Javaid 2019, p.183). Some systems monitor temperature sensors, wind sensors, dust sensors, etc. are communicated to gas monitoring systems; however, these are not included in the risk analysis. Current research mainly focuses on the prediction algorithms and prediction methods of gas emissions and gas concentration (Ma & Zhu 2016, p.1). There is a lack of gas monitoring systems to provide alarms or warnings triggered by any anomaly detected from the correlation analysis of gas data and data acquisition from other sensors.
This research explored whether a correlation exists between data outputs from the gas sensors and other data outputs from different sensors. Anomaly data detected from temperature, wind, and dust might be related to gas data anomaly results. This research lls this gap to uncover hidden patterns and correlations between gas and temperature, wind, and dust, and incorporate data analytics into analytics portfolios for developing the innovation gas warning system. The correlation research method was adopted in this research.
The following section focuses on the literature review. The research framework and research methodology are then presented, followed by data collection and data analysis details. A Triple-Correlation Theoretical Framework and an innovative gas warning system concept is proposed. -a large coal company in China. The ZhongXing gas monitoring system monitors gas, temperature, wind, dust, etc. (Fig.1). If gas data are normal, the data outputs are forwarded to the monitoring system. If the data outputs of gas reach the TLVs, the alarm system will immediately alert the safety-responsive team. Other data outputs of temperature, wind, and dust are also available to be communicated to the monitoring system. However, the current gas monitoring system does not include any risk analysis potentially impacted by other outputs, such as temperature sensors, wind sensors, and dust sensors.

Literature Review
A literature review was conducted to understand the state of research into gas monitoring systems and early warning systems in underground coal mining. The literature focuses on both international research and Chinese research in English and Chinese.
This research rst reviewed the literature published between 2016 and 2020 in Scopus. The search terms were "gas," "coal mine," "monitoring system/model/framework," "alarming system /model/framework," "warning system/model/framework," "prediction system/model/ framework." 188 papers were found. After reviewing the abstracts, most were found to focus on analyzing gas data to explore the methods and framework for providing quick responses or early warnings on any anomaly detected from gas data. A few correlational pieces of research focused on the correlation between the data obtained from different types of sensors but did not conduct any correlational analysis. Jo, Khan & Javaid (2019, p.190) conducted correlational research between temperature, humidity, gas, and CO 2 in underground coal mines and found a strong correlation between temperature and humidity. Zhao et al. (2020Zhao et al. ( , p.1982 discussed gas geology, mining effects, daily prediction, mining pressure, and gas emission dynamic analysis systems. Wang, Li & Li Overall, there appears to be a gap in research on conducting correlational analysis to provide alarms or warnings by anomaly data detected between gas and temperature, wind, and dust.

Research Framework
This research focused on Mine No.1209 in ZhongXing 19 sensors were installed, including ten gas sensors, ve temperature sensors, three wind sensors, and one dust sensor (Fig.2). Table 1 shows the codes of the sensors used in this research. The research framework focused on the correlation between gas data and data acquisition from temperature sensors, wind sensors, and dust sensors (Fig.3). Data changes from temperature, wind, and dust might be critical factors to predict the anomaly gas shift data.

Research Methodology
Reliability and validity analysis were separately conducted between gas and temperature, gas and wind, and gas and dust. The correlation analysis will then be performed between gas and temperature, gas and wind data, and gas and dust.

Reliability Analysis and Validity Analysis
In the test of data obtained between gas and temperature, the results indicated strong evidence of meeting both the reliability and validity standards of exploratory research are listed in  In the test of data obtained between gas and wind, the results indicated strong evidence of meeting both reliability and validity standards of exploratory research are listed in Table 3. Cronbach's Alpha was 0.893 and was considered to have very good reliability (above 0.8). In the test of the validity, the ratio of the number of cases to variables was 1329:1 (greater than 5:1), Kaiser-Meyer-Olkin (KMO) was 0.851 and considered to have a very good measure (great than 0.8), Bartlett's test of sphericity was 0.000 (p<0.001), the average communality was 0.788 (greater than 0.6). No two-item values are correlated above 0.75 (satis ed with discriminant validity).
In the test of gas and dust, the results met both reliability and validity standards of exploratory research are listed in Table 4. Cronbach's Alpha was 0.954 and is considered to have excellent reliability (above 0.95). In the test of the validity, the ratio of the number of cases to Page 6/12 variables was 1571:1, Kaiser-Meyer-Olkin (KMO) was 0.884, Bartlett's test of sphericity was 0.000 (p<0.001), the average communality was 0.845 (greater than 0.6). No two-item values are correlated above 0.75 (satis ed with discriminant validity).

Correlation Analysis
Correlation analysis between gas and temperature, gas and wind, and gas and dust were conducted. Correlation analysis indicated that positive or negative relations exist between gas and temperature (see Appendix 2). A negative relationship exists between gas and wind (see Appendix 3). Positive relations exist between gas and dust (see Appendix 4).

Research Findings and A Triple-Correlation Theoretical Framework
The correlation analysis results indicate 33 elations exit between data obtained from gas, temperature, wind, and dust. There are 23 relations between gas and temperature, including 20 positive relations and three negative relations. One negative relationship exists between gas and wind. There are nine positive relations between gas and dust. Thus, a Triple-Correlation Theoretical Framework was explored that alleges the triple-correlations between gas and temperature, gas and wind, and gas and dust (Fig.4).  Based on the existing the ZhongXing Gas Monitoring System (Fig.1) and the research result -a Triple-Correlation Theoretical Framework (Fig.5), a Uni ed Modeling Language (UML) model was nally developed for demonstrating ZhongXing Innovative Gas Warning system comprising of three layers: the view layer, domain layer, and data access layer (Fig.5). The view layer represents a gas warning system, including the alarming sub-system, warning sub-system, and monitoring sub-system. The data access layer represents databases containing ten gas sensors, ve temperature sensors, three wind sensors, and one dust sensor. The domain layer (is also called a business layer, or business logic layer) represents data processing of data obtained from gas databases, temperature databases, and dust database, and correlation analysis of between gas and temperature, gas and wind, and gas and dust.
The business rules of the ZhongXing gas warning system are constrained into three steps, including three decision rules in the domain layer for demonstrating the system logic ows, including data obtained, correlation analysis, and decision activities, as: First step: This step was conducted between the Data Access Layer and the Domain Layer. The data were obtained from gas databases (ten), temperature databases ( ve), wind databases (three), and dust database.
Second step: This step was run within Domain Layer. The correlation analysis was then conducted between gas and temperature, gas and wind, and gas and dust.
Third step: This step was established between the Domain Layer and the View Layer. The three decision rules were activated. If the results are normal during the correlation analysis, the data would be forwarded to the monitoring system. If the results indicate an anomaly, the warning system will warn the safety-responsive team directly. If the data outputs of gas reach TLVs, the alarm system will immediately alert the safety-responsive team.

Conclusions, Limitations, And Further Research
This research aims to develop an innovation gas warning system for increasing production safety in the underground coal mine industry. The existing gas monitoring systems focus on detecting real-time data obtained from gas sensors and do not analyze whether other sensors' outputs affect gas data. There is a lack of gas monitoring systems in the literature to provide alarms or warning by any anomaly data detected from the correlation analysis of gas data and data acquisition from temperature sensors, wind sensors, dust sensors, etc. This research lls this gap to uncover hidden patterns and correlations between gas and temperature, wind, and dust, and incorporate data analytics into analytics portfolios for developing the innovation gas warning system.
The correlation research method was adopted in this research. 328,320 data outputs were obtained from 19 sensors, including ten gas sensors, ve temperature sensors, three wind sensors, and one dust sensors. The research found strong relations between gas and temperature, wind, and dust. A Triple-Correlation Theoretical Framework was explored, including three correlations -data outputs between gas and temperature, gas and wind, and gas and dust. A UML model was developed for demonstrating the ZhongXing innovative gas warning system. As a result of this research, the ZhongXing gas monitoring system has been upgraded to the ZhongXing Innovative Gas Warning System for user acceptance test to increase coal mining safety on 28 Aug 2020.
The main limitation is that although several industry experts have been involved in this research, the result outcomes are mainly focused on the correlational analysis. Although this research does not provide evidence regarding causal mechanisms between gas and temperature, wind, and dust, the results have indicated existing signi cant relationships. There is a need to conduct further experimental research such as causal study and investigate the cause-and-effect relationships from the coal mining theory. Further research should also be conducted to test whether the Triple-Correlation Theoretical Framework and the ZhongXing Innovative Gas Warning System might be adopted successfully in other coal mining companies.
Declarations Acknowledgment: 1 Supported by the funds of the Shanxi Coking Coal Project (201809fx03). Availability of data and materials: Availability of data and materials have been provided as the appendices. The additional materials will be provided as required.
Competing interests: None. Funding: Provided on the Acknowledgement section Authors' contributions: The name list states the Authors' contributions to this research. The location of the sensors on Mine no.1209  Appendix.docx