1. Introduction
With the increasing of laboratories, experiments projects, various equipment, number of research persons, diverse method to learn safety class, university chemistry accident is also increased. Chemistry laboratory accident is involved in explosions, chemicals leakage, equipment improper operation which cause high cost and casualties. We usually find accidents from website or literature. The existing accident systems are mainly focused on instruments management, education, safety studies video. Different system is independent, such as chemicals systems, special equipment systems, safety test systems. If we have emergency accident, we need retrieve data, it costs time to trace data source. So, we need to have a system to register safely with techniques, also it can share information in real time when it needed.
Researchers have collected nearly 10 years chemistry university accident from websites [
1]. They also collected by literature reviews to analysis which parts more easily cause accidents [
2,
3]. Some researchers want to analysis one or several typical accidents which wants to get safety guide recommend or formulate safety regulators [
4,
5,
6]. Some emergency departments want to recreate animation of the accident.
Safety education is more and more attention, stem education, block chain [
7,
8]. However, it seems no expected effect. We need use more quantitative method to recommend suitable safety courses for experimenters.
Existing accident literature has regional accident research key accident case analysis or annual accident analysis [
9,
10,
11]. We lack a whole website laboratories accident collection.
To solve laboratory safety experiment, it has engineering safer systems [
12]. A remote lab for the “Data Acquisition Systems”, delivered as the digital twin lab [
13]. Internet of Things (IoT) used in many laboratories to decrease safety risks connected with equipment in an automated way [
14]. It also has engineering program and its associated Smart Lab to develop students capabilities in the areas of Artificial Intelligence (AI) [
15].Some researchers analyze the situation of university laboratory management and realizes the necessity of computer technology to manage laboratory. They used the Internet of things to build the university laboratory management system and given credit security evaluation model[
16].It has a method that connects an equation model (SEM) with system dynamics (SD) is presented to dynamically assess lab safety with the insufficient data [
17].Researchers proposed a real-time smart vision-based lab-safety monitoring system to verify the safety protection of students from video [
18].
It needs a system which can store related accident data safely and search data source quickly, meanwhile, it also needs predict risk quantitative. So, our paper proposed a chemistry accident system based on data ownership safety architecture to solve these research gap. Firstly, we collected chemistry accident using python and manual, then used these historical accident data to choose risk variables by Spsspro, and design accident model using Stata. Secondly, we design our data ownership safety architecture accident system which let data owner register safely and conditional sharing quickly with key algorithm. Finally, we put this accident manual database to our proposed systems, then used risk model in systems to predict experiment risk level, it also can recommend suitable education safety class.
Data ownership safety architecture is described one body and two wings which one body is the data combined with ownership. The one wing is that data can register innately, another wing is the data should use the key technology to protect data during conditional sharing [
19]. Data ownership safety architecture application has advantage on tourism, smart cities and laboratory management[
20].
3. Results
This section may be divided by subheadings. It should provide a concise and precise description of the experimental results, their interpretation, as well as the experimental conclusions that can be drawn.
3.1. Laboratory accident data risk analysis
By cleaning collected accident data, we classified equipment into common equipment, special equipment, gas cylinder and non-equipment. Chemicals are divided into common chemicals, hazardous chemicals, and non-chemicals by MSDS classification. Experimenter degree included doctoral, master, undergraduate, college. University level is classified into normal, high, and top-level using university rank documents and Chinese 211,985 documents. After cleaning data, we got 220 chemistry university laboratory accident databases.
We use python to collect laboratory accident-related literature review. The word cloud generated by using python is shown in
Figure 1.
We need to choose key risk variables from collected factors, and decided which factors can be dependent variable. Through historical chemistry laboratory accident databases, we used Spsspro to analysis risk factor.
From
Figure 2(a), we can observe accident happens season, spring, summer, and winter is no noticeable difference. The reason that accidents happened on autumn is repetitively minimal is September is new semester opening, university have been do concentrated safety education. According to experiment type, it shows the proportion of teaching experiment is higher than research experiment (
Figure 2b). However, the number of attends on scientific is less than the joined teaching experiment students. From percentage of experimenters, the researcher experiments are more easily happened accidents (
Figure 2c). According to chemicals to accidents effect, hazardous chemicals is most serious. Common chemicals ratio equals to gas usage; however, chemistry experiments are less use gas. Gas is more danger than common chemicals. Electric harm, injection, and mechanical injury account a smaller proportion (
Figure 2d). From collecting all accident types, the proportion of the explosion is as high as 85%, and the chemicals leakage ranked number 2. Other reasons are all around 1%. The reason is particularity of chemistry laboratory, it used hazardous chemicals and special equipment do experiments can cause lager explosion which causes injured or economic losses. Chemicals leakage can cause long-term environment pollution which also need to be attention (
Figure 2e). It shows common equipment account 75% in accident, and special equipment about 20% which includes high pressure equipment, gas, etc. And the infrastructure involves water, electric, exhaust architecture is nearly 3% in collected database (
Figure 2f). We can see teachers or staff not in laboratory is 2 times easier to happen an accident. In
Figure 3, it is shown recently is more and more accident happened.
Then we used above analysis results, we have eliminated season variables. And we co minded injured, dead number with accident type into accident level which from class1 to class 5. We choose six factors to make a heat map using Spsspro which is shown in
Figure 4.
From
Figure 4 we can observe, Q6 experiment type is relatively strong correlation with the Q10 experimental accident level. Q8 accident injured number is negative correlation with Q10 experimental accident level. And accident level is positive correlation with use of Q18 chemicals usage. Whether the teachers are in lab or not is also relative with accident level and experiment type. So, next step we used these analysis results to choose variables and establish accident model.
3.2. Accident risk model using Stata
Through the above analysis, we choose experimenters degree, experiments used equipment and chemicals as independent variable. And dead, injured number combined with accident type set the causative variable. We used these factors to establish risk model equation using Stata. The regression results by Stata are shown in
Figure 5.
From regression model, we choose college student, non-equipment, and non-chemicals as benchmark. Because part of data is lacked, Stata certified sample size is 196.From Stata results, we can see person degree, common chemicals and infrastructure have no strong correlation for accidents. However, hazardous chemicals, gas and special equipment have relationship with chemistry laboratory accidents took them as variables, so the model equation is:
t (2.65)(2.3)( 2.1)
AL=accident level SE= Special Equipment G=Gas HC=Hazardous Chemicals
a=1.976153 b=0.7265465 c=0.8661627 d=0.499994
3.3. Design laboratory accident data system
Though manual collected database quantitative analysis, we choose strong and weak connection variables,then design whole chemistry laboratory accident structure.
We designed to have a system to store all related laboratory accident data into DRC, including consumables broken, glass consumables heating explosion, experiments explosion, etc. If an accident happened, we need to store related data, such as accident time, location, lab, experimenters, projects, chemicals, equipment, consumables, real time IOT photo and video into DRC with SM2 or AES key algorithm to authorize data ownership.
Then we can link important laboratory accident-related data according to above analysis using MySQL which is shown in
Figure 6. If an accident happened in a laboratory, we could search from Lab-id to trace person who charge of lab and project supervisor information. We also can search this project used equipment; application chemicals record to design emergency rescue plan. Though equipment-id, we can find equipment-details, including bidding contract, companies, use manual, etc. We can combine IOT video to judge whether the operator or equipment caused the accident. Using lab-id we can search laboratory infrastructure, such as water pipe material, circuit maintenance records. Meanwhile, IOT equipment can real-time monitoring which also need encrypted upload to proposed system.
These data are encrypted to register laboratory accident system based on DOSA. Only emergency happens, data owner use key algorithm to authorized emergency department which can break isolated data islands.
Supervision departments needs to search accident-related data, such as hazardous chemicals collection. Each laboratory data owner can find supervision departments public key from DAC, then encrypted chemicals data to departments’ public key. Supervision department can use their private key to decrypt data.
Table 2 presents the function and meaning. The process of encrypted accident data to DRC using SM2 is shown in Algorithm 1.
Algorithm 1. Sm2enc_to_mysql () |
Input:excel of accident data
Output:None
- 1.
db = connect_to_mysql(database, username, password)
- 2.
- 3.
excel_file = open_excel_file(“accident_data.xlsx”)
- 4.
sheet = select_sheet(excel_file, “accident”)
- 5.
accident_data = read_data(sheet)
- 6.
- 7.
encrypted_data = []
- 8.
For row in data:
- 9.
encrypted_row = []
- 10.
For cell in row:
- 11.
encrypted_cell = sm2_encrypt(cell)
- 12.
encrypted_row.append(encrypted_cell)
- 13.
encrypted_data.append(encrypted_row)
- 14.
End for
- 15.
End for
- 16.
- 17.
accidents = " accident_table"
- 18.
columns = ["column1", "column2", "column3"]
- 19.
For row in encrypted_data:
- 20.
values = []
- 21.
For cell in row:
- 22.
values.append(cell)
- 23.
sql = construct_insert_sql(accidents, columns, values)
- 24.
execute_sql(db, sql)
- 25.
End for
- 26.
End for
- 27.
- 28.
close_mysql_connection(db)
|
4. Usage of laboratory accident system
This web designed method is data ownership safety architecture. Data owner can register this system which provide a key pair. Our system can transmission private key to data register and reserve the public key to DAC. If the lab data owner wants to register accident data to this system, he can use his public key to encrypt this data to confirm ownership, meanwhile data category can automatic generated. Data user can search category to find target data which need apply to the data owner or DIY data product.
Figure 7 shows our proposed system can link related data sources and data owners, such as regulator departments, chemicals and waste liquid companies, various lab, different research teams. These data owners can encrypt their data to this system with their own public key which can confirm data ownership.
Figure 8 shows our system can collect historical lab accident or happening accident. The research team do not want to public their laboratory accident, they can encrypt their accident into system with their own public key which even administrator cannot view this information without permission. But if they have emergency or supervision, they can use supervision public key to encrypt which can authorization.
Figure 9, if any new project will do experiments in lab, they can put their data using our risk equation to predict level.
Figure 10, if the experimenter’s predicted risk is relatively high, it can choose VR experiment or recording to learn safety class.