Preliminary Design and Construction Database for Laboratory Accidents

Abstract

With the growth of university chemistry experiment projects, the corresponding laboratory safety risks are increasing year by year for scientific research personnel, and specialized equipment. However, accident data are not stored systematically for lack of a safety platform to collect accident information, share the causes of accidents, and predict safety risks. To solve these problems, we designed a laboratory accident system to store and share related data, and predict risk levels. In this paper, the majority of chemistry laboratory accidents were manually collected by Python software (version 3.10.11) and were categorized based on their risk level. Moreover, the variable factors that generated risk were analyzed using Spsspro, which facilitates the construction of a meaningful forecasting model of laboratory safety via Stata. It is worth noting that the registered laboratory accident data in the proposed chemistry accident system were based on the data ownership safety architecture. The chemistry accident system can break through data barriers using confirmation and authorization key algorithms to trace non-tampered data sources in a timely manner when an emergency accident happens. Meanwhile, the proposed system can use our designed accident risk model to predict the risk level of any experimental project. It can also be recommended as an appropriate safety education module.

1. Introduction

The number of chemistry experiment accidents is increasing with the growth of laboratories, experimental projects, experimental devices, as well as new experimental science such as energetic materials, energy storage, and more. [1,2,3]. Accidents in chemistry laboratories, including laboratory fires (explosions), chemical leakages, and damage to equipment, mainly result from an improper use of reagents, experimental devices, and operational processes, which leads to significant financial losses as well as, to some extent, personal injury [4]. Generally, some accident information can be accessed from websites or literature reviews. The current accident-related systems mainly focus on managing instruments, safety education and training, and producing safety research videos. Researchers have designed a social virtual reality (VR) system which allows experimenters to perform role-learning, and social interaction for safety education [5]. To improve laboratory safety, researchers have developed a laboratory accident alarm system based on remote control technology [6]. However, the laboratory systems are adapted to manage chemicals, specialized equipment, and safety tests, and work independently, which restricts immediate access to the related data to track the cause of the accident during an emergency laboratory accident. It is necessary to establish a mechanism for securely registering procedures and sharing information in real-time.

Researchers have collected nearly two decades of accidents in chemistry laboratories from different universities [7]. They also gathered data from a literature review to analyze the vital factors which result in chemistry lab accidents [8]. However, no researcher, university, or organization has collected accident information in academic laboratories [9]. Some researchers expect to examine one or more common accidents to offer safety guidelines or create safety rules [10,11], and some emergency departments wish to review the accidents using animation. Regional incident studies, important incident case analyses, and annual accident analyses can be found in the existing accident literature [12,13,14]. Thus, there is a lack of a comprehensive and expandable accident database for collecting all reported lab accidents. Blockchain, STEM education, and VR education have gained popularity in the field of safety education [5,15], but their actual impact may not have met the expected outcomes. To recommend appropriate safety training for experiments, it is necessary to employ more quantitative methods.

Engineers have created safer technologies to improve the safety of experiments in the laboratory [16]. A remote “Data Acquisition System” is provided as a Digital Twin Lab [17]. The internet of things (IoT) is often utilized in laboratories to automatically reduce equipment-related safety issues [18]. Additionally, its feature of an engineering curriculum and its relation to Smart Lab help students enhance their artificial intelligence (AI) skills [19]. Some scholars examine the management of university laboratories and conclude that computer technology is required to operate laboratories. They created a university laboratory management system using IoT and provided a credit security evaluation methodology [20]. It presents a method for dynamically evaluating lab safety in the absence of appropriate data by linking an equation model with system dynamics [21]. A real-time smart vision-based lab safety monitoring system was designed by researchers to confirm that students were being protected by a vision-based system [22].

It is urgent to develop a system that can store relevant accident data securely and search data sources quickly and predict risk quantitatively. Therefore, the current work developed a laboratory accident system using data ownership safety architecture theory. One advantage of data ownership safety architecture (DOSA) is that data can be naturally registered, and another advantage is that data can employ key technologies to safeguard data when shared under various conditions [23]. Applications for data ownership safety architecture are advantageous for the tourism industry [24], government information resource sharing, digital copyright works protection, and trading systems [25]. Managing chemistry laboratories has led to the establishment of a supervision management data framework [26], a lab experimental data trading workflow [27], and the unconditional sharing of lab-related data with key algorithms to make sure data transfer is efficient and safe [28].

Here, we present a manual laboratory accident system based on DOSA. Firstly, Python and manual labor were utilized to gather data on chemistry accidents. Secondly, the risk factors were extracted from those historical data with Spsspro and an accident model was created with Stata. Thirdly, a data ownership safety architecture accident system was developed that enables data owners to simply and securely register for conditional sharing. Finally, we implemented our suggested methods using this accident manual database. The developed model can be used to anticipate the experiment’s risk level and offer the most appropriate safety education course. The proposed accident manual database, when implemented effectively, has the potential to provide real-time and untampered accident data related to laboratory incidents. This timely and accurate information can be instrumental in designing effective rescue plans and responding promptly to emergencies. Furthermore, any research team can predict experimental project risk levels using our accident prediction model.

2. Materials and Methods

2.1. Manual Laboratory Data Collection Database

The keywords chemistry lab accidents, chemical explosion in chemistry labs, and accident by equipment mishandling in chemistry labs were searched in public academic databases using Python software, and a total of 854 reported chemistry laboratory accidents in the literature were obtained, as shown in Figure 1. The accident data of the reported scientific literature were manually screened and integrated into our laboratory accident database. We also collected laboratory accident data from non-academic sources.

To collect chemistry laboratory accident data from non-academic websites, the environmental engineering undergraduate safety class students of the 1st semester of 2022 were given the assignment to collect such data which was officially included in their study tasks. A total of 240 students actively participated and each participant collected 10 chemistry laboratory accidents from non-academic websites according to parameter settings including the year of the accident, season, time, level of the university, major, equipment used, chemical types, experimenters’ (instructors’, students’, etc.) degree levels, whether the responsible teacher was present or absent during the accident, lost valuable belongings, cases of personal loss or injury, and, most importantly, the data’s source as shown in

Table 1. Meta information of laboratory accidents.

Variable	Type	Descriptions
Year	Numerical	The year of the accident
Season	Character	The season of the accident
Location	Character	The location of the accident
Experimental classification	Character	Teaching (instructed by teachers); scientific research (research team); or factory and university cooperation project
Dead	Numerical	Number of people dead in accident
Injure/dead number	Numerical	Number of persons injured or dead in an accident
Accident style	Character	Accident types are divided into explosions, and chemical or mechanical injuries
Degree	Character	People in the accident who have a college, undergraduate, master, or PhD degree
Equipment	Character	Equipment is common, special, or is infrastructure equipment
Chemicals	Character	Common chemicals, hazardous chemicals, gases, or non-chemicals
Teacher in accident	Character	Are teachers in an accident (yes/no)
Accident cause public	Character	Is the cause of the accident public (yes/no)
University level	Character	University level divided by 211 and 985 project universities by Chinese criteria and university rank
Money lost	Numerical	Money lost in an accident
Data source	Character	Which website found these accidents

.

2.2. DOSA

The manually collected accident data were added to the prediction system. As many research teams did not want to share their ordinary accidents with the public, it was difficult to collect those minor or near-miss accidents on websites or in the literature [9]. It is identified that lab-related data privacy should be enforced, so, the system is constructed based on the data ownership safety architecture.

DOSA was proposed by Miao in 2008; it contained a data register center (DRC), a data authorization center (DAC), data application units (DAUs), and key technology to protect data ownership and privacy [29]. It can break through data barriers, mining more value from the data. In our previous studies, we selected a suitable key algorithm to guarantee the unconditional sharing of data between data owners and users in the DAC. Chinese domestic commercial key SM₂ was employed for lab data which needed a high degree of protection and the advanced encryption standard (AES) was employed for normal lab-related data [28].

To ensure data ownership interest, persons were contacted to authorize ownership [28]. The data owner can put his data with his public key in the DRC; meanwhile, the DRC can generate categories automatically. Data users can search targeted data from the data categories. If the data user needs some type of data, he can contact the data owner. If these data can be shared with the user, the data owner can find the data user’s public key in the DAC. Then, the data owner encrypts this data with the user’s public key stored in the DRC and authorizes the user. The data user can decrypt these targeted data using the private key to finish conditional sharing. The encryption and enquired data workflow using DOSA is shown in Figure 2.

2.3. Design Laboratory Accident System Method

We used DOSA to design a manual laboratory accident database system. Our proposed system can register accident-related data safely and allow conditional sharing with key technology. We can also use new experimental project information to predict risk levels.

Our research conceptual framework is shown in Figure 3. Step 1, laboratory data were collected from the literature and assignments were made to establish a manual laboratory accident database. Step 2, to increase our manual database, our group tried to design a system to store laboratory accident-related data with a key algorithm to protect the data owner’s privacy. Then, the system could automatically generate an accident category. Step 3, these related lab data contributed to real-time, untampered data when supervision departments needed check or rescue. Step 4, the manually collected lab accident data were employed to establish a risk prediction model. Through Spsspro analysis results and the Stata risk model, a new project can input related data to predict its risk level. If the risk level is high, the researcher can get a suitable safety education recommendation. To solve insufficient accident data, in Step 5, the existing predicted equation is added into the proposed system, and the system can gather more accident data because it can protect privacy.

Researchers or laboratory suppliers can encrypt their data with their public keys. Once some departments need supervision or accidents happen, data users can search from the data category to find the data owner. The data owner inputs this data with the supervision department’s public key which can be found in the DAC to authorize the sharing of data. The experimental team need not be afraid of data leakage; meanwhile, supervision departments can get fresh data and extract value from the data.

3. Results

3.1. Laboratory Accident Data Risk Analysis

After cleaning the collected accident data, lab equipment can be classified into common equipment, special equipment, gas cylinders, and non-equipment. Chemicals were divided into common chemicals, hazardous chemicals, and non-chemicals with the material safety data sheets (MSDS) classification. Experimenters were categorized based on their degrees as doctoral, master, undergraduate, and college students. Universities were classified into normal, high, and top-level using the Chinese university ranking system, such as 211 and 985. After data cleaning, the obtained 220 chemistry laboratory accident data were employed to establish databases.

Python software was applied to generate word clouds for the laboratory accident-related literature review collection, as shown in Figure 4.

The key risk variables were screened via Spsspro software (version 1.1.13) from collected factors based on historical chemistry laboratory accident databases, and some of them were identified as the dependent and independent variable. It was found that there are no noticeable differences in accident numbers across the year. It is inferred that the accidents happened with a small amount of data, perhaps indicating that at the start of the semester in September, the university exercised proper safety education (Figure 5a). According to the chemicals’ effect on accidents, hazardous chemicals have a larger effect, and the effect of common chemical accidents is close to that of gas usage (Figure 5b). According to experiment type, the proportion of teaching experiments was higher than research experiments (Figure 5c). However, the number of attendees to the scientific research is less than the number of teaching experiment students. From the collected accident types, the proportion of explosions is as high as 85%, and a leakage of chemicals was the second most common accident type. This is because in chemistry laboratories, using hazardous chemicals and special equipment in experiments can cause a larger explosion which results in injuries or economic losses. Chemical leakage can cause long-term environmental pollution which also needs to be addressed (Figure 5d). We observed that teacher or staff absence in the laboratory was two times more likely to result in an accident than if they were on site (Figure 5e). We cannot collect sufficient publicly available lab accident information to determine the cause of the accident nearly 90% of the time (Figure 5f). Figure 6 presents the trend of the number of laboratory accidents in recent years using Spsspro.

Seasonal variables were excluded, and the number of injured and deceased individuals, along with accident types, were combined to classify accident levels into five classes ranging from class 1 to class 5. A total of six factors were selected to generate a heatmap using Spsspro, as illustrated in Figure 7.

Figure 7 demonstrates a noticeable correlation between the experiment type and the level of experimental accidents. There exists a negative correlation between the number of injuries and the level of experimental accidents. Additionally, the accident level shows a positive correlation with the chemicals used. Furthermore, the presence of teachers in the lab and the type of experiment are also factors related to the accident level. Based on these analysis results, we proceeded to select variables and establish a laboratory accident prediction model.

3.2. Accident RiskModel Using Stata

Based on the aforementioned analysis, it was hypothesized that the education level of researchers, equipment selection, and chemical usage might influence the occurrence of laboratory accidents. Therefore, the experimenter’s degree of education, the equipment employed, and the chemicals utilized were considered as independent variables. Since education level, equipment, and chemicals are categorical variables, the authors transformed these variables into dummy variables. Specifically, we have converted the education level variable into four dummy variables: college students, master’s students, undergraduate students, and doctoral students. The selected equipment variable has been transformed into three dummy variables: general equipment, infrastructure, and specialized equipment. The variable for chemicals used has been converted into four dummy variables: no chemicals, regular chemicals, gas cylinders, and hazardous chemicals. The combination of the number of casualties and the type of accident was considered as the dependent variable. Regression analysis was conducted using Stata software, and the results are presented in column (1) of

Table 2. Regression results by Stata.

	(1)	(2)	(3)
Md	1.826 (1.46)
PhD	1.837 (1.48)
Un	1.340 (1.09)
Inf	−0.914 (−1.54)
Se	0.702 ** (2.48)	0.726 *** (2.65)	0.673 ** (2.43)
Nc	1.064 (0.59)
Cc	0.767 (1.39)
G	1.216 ** (2.14)	0.866 ** (2.32)	0.923 ** (2.43)
Hc	1.064 ** (2.1)	0.499 ** (2.10)	0.583 ** (2.38)
Cons	−0.648 (−0.05)	1.976 *** (9.34)	1.981 *** (9.18)
N	174	196	182
R2	0.1991	0.1129	0.1138

Note: T-values in parentheses; *** p < 0.01, ** p < 0.05.” Md: master’s degree student, PhD.: Ph.D. degree student, Un: undergraduate student, Inf: infrastructure, Se: special equipment, Nc: no chemicals, Cc: common chemicals, G: gas, Hc: hazardous chemicals, Cons: Constant term, N: number of samples, and R²: coefficient of determination.

.

From the regression results in column (1) of Table 2, only three regression coefficients are significant, which are the special equipment, gas cylinders, and hazardous chemicals’ regression coefficients. Since we tried to establish a risk model to predict the risk level of laboratory accidents, we took the statistical significance of the variables into account in the model [30]. So, we retained the variables with significant coefficients in the original assumption model, and then established a new risk model, model (1).

Column (2) of Table 2 is the baseline regression result of risk model (1). It can be seen that the regression coefficients of special equipment, gas, and hazardous chemicals are all positive, and have passed the significance level test of 1%, 5%, and 10%, respectively. The above results show that using special equipment, gas, or hazardous chemicals will increase the risk of accident level (AL).

$$AL=a+b\ast Se+c\ast G+d\ast Hc$$

(1)

To verify the robustness of the risk model’s results, we eliminated abnormal historical accident samples. For example, we deleted the relatively old laboratory accidents before 2004. In addition, we deleted the accident sample that occurred in a military school in 2005 because the sample lacked sufficient information. We also eliminated two laboratory accidents happened in the Fall of 2007 and 2011, respectively, of unknown universities as specific names could not be found on the internet. The sample of an accident that occurred in a university laboratory in the United States in 2018 was deleted because the casualties were non-experimental workers. So, a total of 18 experimental accident samples were deleted. Then, we utilized the preprocessed data to perform a regression on model (1) again, obtaining the regression results in column (3) of Table 2. The regression coefficients of the three variables were still significantly positive, and the values have not changed significantly. Thus, the benchmark regression results are robust.

According to the collected historical laboratory accident data, the resulting laboratory accident risk prediction model is shown in model (2). Since our accident collection system is dynamic, coefficients and parameters will be adjusted according to subsequent accident data.

$$AL=a+b\ast Se+c\ast G+d\ast Hc$$

(2)

where AL is accident level. Subsequently, Se represents special equipment, and G indicates gas. Hc means hazardous chemicals. The coefficients are a = 1.9761 b = 0.7265 c = 0.8661 d = 0.4999.

In order to gather more laboratory accident data, a dynamical lab accident system was designed based on data ownership safety architecture. The dynamic lab accident system not only provides supervision departments with comprehensive regulatory-related data, including chemicals usage and waste liquid management, but also assists experimenters in predicting project risk levels. Furthermore, a more accurate risk model can be established using newly collected accident data on the proposed system.

3.3. Laboratory Accident Data System

The variables were selected through quantitative analysis using a manually collected accident database. The entire structure of the chemistry laboratory accident system was designed using MySQL. All related laboratory accident data stored in the DRC, including consumables broken, glass consumables in heating explosions, and experimental explosions, were designed for the proposed system. If an accident happened, lab staff needed to store the related data, such as accident time, location, experimenters, projects, chemicals, equipment, consumables, the real-time IOT equipment photo, and video into the DRC with an SM₂ or AES key algorithm to authorize data ownership [28].

Important laboratory accident-related data can be linked according to the above analysis using MySQL, as shown in Figure 8. If an accident happens in a laboratory, the emergency department can search the Lab-id to trace the person in charge of the lab and find project supervisor information. They can also search for this project’s used equipment, and the application chemicals record to design emergency rescue plans. Though equipment-id and equipment-details, including bidding contracts, supplier companies’ information, etc., can be found, the emergency department can also combine IOT videos to judge whether the operator or equipment caused the accident. Using the lab-id, we can search laboratory infrastructure, such as water pipe material, and circuit maintenance records. Meanwhile, IOT equipment provides real-time monitoring which also needs to be encrypted when uploading to the DRC.

These data are encrypted to register the laboratory accident system based on the DOSA. Only when an emergency happens does the data owner use a key algorithm to authorize the emergency department, so that isolated data islands can be accessed.

Supervision departments need to collect accident-related data such as hazardous chemicals usage. Every laboratory data owner can find the supervision department’s public key from the DAC, then access encrypted chemicals data with the departments’ public key. The supervision department can use its private key to decrypt data. The underlying logic of the laboratory manual accident database system is encapsulated in the proposed Algorithm 1.

Table 3. Function and meaning.

Function	Meaning
connect_to_mysql	Connect to MySQL database
open_excel_file	Open Excel file
select_sheet	Select a worksheet in an Excel file
read_data data. analyze	Read data in an Excel file Parse keywords form a row of data
algorithm_encrypt	Encrypt cells using an encryption algorithm
construct_insert_sql	Build a SQL insert statement
execute_sql	Execute a SQL statement
close_mysql_connection	Close the MySQL database connection

provides the functional implementation details of each function call in Algorithm 1. Algorithm 1 demonstrates the process of encrypting accident data using key algorithms such as SM₂, AES, or RSA, and automatically generates a directory. This process ensures the secure storage and access of accident data.

Algorithm 1. Encrypt the manual database and generate a catalog automatically

Input: The Excel “accident_data.xlsx” composed of manually collected accident data Output: catalog db = connect_to_mysql(database, username, password) excel_file = open_excel_file(“accident_data.xlsx”) sheet = select_sheet(excel_file, “accident”) accident_data = read_data(sheet) encrypted_data = [] catalog = [] 7. For row in data: 8. encrypted_row = [] 9. key,words = data.analyze(row) 10. content = (key+“,”+words) 11. catalog.append(content) 12. For cell in row: 13. encrypted_cell = algorithm_encrypt(cell) 14. encrypted_row.append(encrypted_cell) 15. encrypted_data.append(encrypted_row) 16. End for 17. End for 18. accidents = “ accident_table” 19. columns = [“column1”, “column2”, “column3”] 20. For row in encrypted_data: 21. values = [] 22. For cell in row: 23. values.append(cell) 24. sql = construct_insert_sql(accidents, columns, values) 25. execute_sql(db, sql) 26. End for 27. End for 28. close_mysql_connection(db) 29. return catalog

Algorithm 1 encompasses the following implemented steps: (1) Establishing a lab accident database that is interconnected with relevant lab data. (2) Reading and analyzing the manual database for each incident to extract keywords, facilitating the creation of an accident directory. (3) Encrypting the accident data and securely storing it in the dynamic database, ensuring the privacy and interests of the data owner. (4) Automatically generating a catalog within the laboratory accident system that allows users to query and access the data. These steps collectively constitute the functionality of Algorithm 1.

4. Usage of the Laboratory Accident System

This section presents the actual results obtained from implementing the components of the laboratory manual accident system. Our intention is to gather additional laboratory accident data in order to develop a more accurate risk model. Many research teams do not publicize their small incidents, near-misses, or instrument misuse; so, laboratory accident data are hard to collate [9]. Furthermore, we aim to mitigate accidents resulting from various sources by implementing measures such as conducting effective safety studies. The system also incorporates targeted safety education based on the results derived from the risk model. Here, we present a laboratory manual database accident system that can predict laboratory experiment project risk levels based on gathering historical accident data, as shown in Figure 9. Any student or researcher who wants to enter the laboratory can input experiment-related data into the system, which will calculate the experiment’s risk level according to the existing risk model.

Figure 10 shows that the system can link accident-related data sources based on the DOSA, which can break data barriers. These data owners can encrypt their lab data with their public key in this system which confirms data ownership.

Laboratories, supervision departments, and suppliers can register and obtain a unique key pair within the manual accident database system. Public keys can be securely stored in the DAC, and private keys need to be saved separately. Lab data owners can store chemical data, experimental data, experiment projects, or accident data with the data owner’s public key to authorize ownership and protect privacy. Indeed, a lot of information can be made publicly available and does not require key protection, such as equipment, chemicals, and waste liquid bidding information. Traditional websites often provide easy access to these public data. However, we offer a convenient platform specifically designed for storing all laboratory accident-related data.

Figure 11 illustrates how our system can collect both historical laboratory accident data and real-time accident data. If the research teams do not want to publicize their laboratory accident, they can encrypt their accident data into the system with their public key which even administrators cannot view without permission. However, if the lab needs emergency security or supervision, data can be authorized using a key algorithm [28].

Our system can collect the accident’s time of occurrence, location, and the experimental project’s properties, which includes whether the project is for scientific research, teaching, or in collaboration with companies. The accident type is divided into explosion, damaged consumables, hazardous chemical leakage, broken water pipes, etc. In the proposed system, it is essential to record the equipment and chemicals utilized in the experiments. Additionally, important details such as the accident time of occurrence, the educational level of team personnel, and whether supervisors were present in the laboratory should be stored. In the unfortunate event of an accident resulting in casualties, it is crucial for the laboratory staff to accurately input data into the system.

Figure 12 demonstrates that if the experimenter’s predicted risk level is relatively high, he can choose a VR experiment or relevant recording module to learn the relevant safety material. One benefit of our proposed system is that researchers can study targeted high-risk-level safety education modules.

5. Discussion

Prior research on laboratory accidents has concentrated on specific regions [31,32], such as the analysis of accident causes among Chinese students in laboratories. In a similar vein, lab management systems have developed specialized modules or majors, including bio-risk [33], equipment, and safety education. Given the insufficiency of traditional laboratory information management systems (LIMS) in terms of safety [34], we adopted a method that binds a person and data to authorize data ownership using a key algorithm.

To the best of our knowledge, there are scarcely any systems that can link manual accident databases with related lab data. Compared with earlier lab accident research, the authors established a manual laboratory accident database that can be improved dynamically by users.

The system can link all laboratory accident-related data, which can break through data barriers. It provides a quantitative prediction model, which will decrease accident probability and improve safety learning efficiency. The constructed system was based on data ownership safety architecture using a key algorithm. So, if any emergency accident occurs, it can quickly trace related data using this proposed system. It can also provide real-time and non-tampered data to the rescue team. Finally, the suggested system function can be extended and applied to other events, such as smart cities, tourism management, etc.

6. Conclusions

The authors propose a laboratory manual database accident system based on the data ownership safety architecture by using a prediction model which can guarantee the data owner’s interests, as well as overcoming accident-related data barriers. Not only can lab experimenters enter their data into the DRC, but related suppliers and supervision departments can also view and share data in our proposed system. The advantage is that if some accidents happen, information on who is in the labs and which companies’ chemicals and equipment were used can be quickly obtained. So, emergency departments can get immediate data to design secure projects in real-time. We designed a confirmation and authorization method by binding persons and data using SM₂ or RSA. We tried to decrease lab accidents by predicting risk levels using a risk model based on the manual accident database. Through the use of predicted results, students can enter experiment information and obtain a risk level. This makes safety education targeted, which can improve efficiency. Our database can be extended because we can get more accident data using a key algorithm to keep data private.

The research limitations of the study are: (1) It is hard to get detailed information about the accident that happened which is not public or published. (2) Some websites’ information is different from others or unclear. (3) Many literature reviews, books, and sample materials often concentrate on specific types of laboratory accidents or directly present the analysis results. So, the authors tried to conduct assignment methods to collect more information from non-academic channels.

The practical implications of our study are a dynamic manual database that could be extended to establish equipment or hazardous chemical usage predictions by using the proposed method. By analyzing the statistics of experimental accidents, teachers can enhance their focus on safety during the teaching of safety courses. Furthermore, our method can be applied to more scenarios, such as emergency correlation databases in the medical industry, which can quickly find a real data source to break data barriers. In future research, we aim to explore the utilization of blockchain technology to ensure the integrity and traceability of students’ safe learning experiences. By implementing blockchain, we intend to create a tamper-proof system that reduces the likelihood of laboratory risks from their very source.