CN115099673A - Method for dynamically and intelligently evaluating dangerous chemical storehouse fire hazard risk - Google Patents
Method for dynamically and intelligently evaluating dangerous chemical storehouse fire hazard risk Download PDFInfo
- Publication number
- CN115099673A CN115099673A CN202210819137.4A CN202210819137A CN115099673A CN 115099673 A CN115099673 A CN 115099673A CN 202210819137 A CN202210819137 A CN 202210819137A CN 115099673 A CN115099673 A CN 115099673A
- Authority
- CN
- China
- Prior art keywords
- storehouse
- model
- electronic device
- electronic equipment
- kernel function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 40
- 239000000126 substance Substances 0.000 title claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 35
- 238000012706 support-vector machine Methods 0.000 claims abstract description 31
- 238000012360 testing method Methods 0.000 claims abstract description 31
- 238000013210 evaluation model Methods 0.000 claims abstract description 28
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 23
- 238000012502 risk assessment Methods 0.000 claims abstract description 12
- 230000006870 function Effects 0.000 claims description 68
- 238000011156 evaluation Methods 0.000 claims description 32
- 230000008569 process Effects 0.000 claims description 7
- 230000008859 change Effects 0.000 claims description 5
- 238000009826 distribution Methods 0.000 claims description 4
- 238000013507 mapping Methods 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 238000004880 explosion Methods 0.000 claims description 3
- 230000009471 action Effects 0.000 claims description 2
- 238000013461 design Methods 0.000 claims description 2
- 230000036039 immunity Effects 0.000 claims description 2
- 239000002245 particle Substances 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 description 18
- 238000002474 experimental method Methods 0.000 description 10
- 239000005433 ionosphere Substances 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 238000004088 simulation Methods 0.000 description 8
- 208000019622 heart disease Diseases 0.000 description 5
- 238000002790 cross-validation Methods 0.000 description 3
- 230000008602 contraction Effects 0.000 description 2
- 239000000383 hazardous chemical Substances 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 102100029469 WD repeat and HMG-box DNA-binding protein 1 Human genes 0.000 description 1
- 101710097421 WD repeat and HMG-box DNA-binding protein 1 Proteins 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012552 review Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- General Physics & Mathematics (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- General Health & Medical Sciences (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Security & Cryptography (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for dynamically and intelligently evaluating fire hazard in a dangerous chemical storehouse, which comprises the following steps of firstly, according to 7: 3, dividing the ratio into a training set and a testing set; the method comprises the steps of establishing a Support Vector Machine (SVM) model of an improved mixed kernel function, namely an NPSVM model, by linearly combining an improved Gaussian kernel function and a polynomial kernel function; optimizing the parameters of the established NPSVM model by adopting an electrostatic discharge algorithm (ESDA), and finally forming an ESDA-NPSVM prediction model; training the prediction model by utilizing a training set and a testing set which are divided in advance based on the established ESDA-NPSVM prediction model to obtain an intelligent evaluation model of the fire risk level of the storehouse; and evaluating the fire risk level of the dangerous chemical storehouse based on the established storehouse fire risk level intelligent evaluation model. The method can solve the problems that in the prior art, the fire risk assessment cost of the dangerous chemical storehouse is high, the generalization capability is weak, and overfitting is easy, and has high generalization capability and prediction accuracy.
Description
Technical Field
The invention relates to the technical field of dangerous chemical storehouse fire risk prevention and control, in particular to a method for dynamically and intelligently evaluating dangerous chemical storehouse fire risks.
Background
In the actual production and management of chemical enterprises, the stock quantity, the temperature, the humidity, the combustible gas concentration and the like in a storehouse are always in a dynamic change state, so that the overall safety state of the storehouse is changed, and the risk of fire disasters is increased. The fire risk levels (the probability of fire) corresponding to different safety states of the storeroom are different, if an enterprise can master the fire risk levels of the storeroom in real time, safety work can be pertinently developed, accidents are effectively prevented, and the safety of the enterprise is guaranteed.
The existing solution scheme in the prior art comprises a safety check list method, a road method expert review method, an event tree analysis method and the like, but the scheme needs more prior knowledge, has higher implementation cost or longer evaluation period, cannot evaluate the fire risk level according to the real-time state of the warehouse, can establish an evaluation model by means of a machine learning method for objectively and accurately mastering the dynamic fire risk evaluation result of the warehouse, and the existing evaluation model established by utilizing the machine learning has weaker generalization capability and is easy to over-fit, has poor effect and lacks an efficient solution scheme.
Disclosure of Invention
The invention aims to provide a method for dynamically and intelligently evaluating the fire risk of a dangerous chemical storehouse, which can solve the problems of higher cost, weak generalization capability and easiness in overfitting of the fire risk evaluation of the dangerous chemical storehouse in the prior art and has higher generalization capability and higher prediction accuracy.
The purpose of the invention is realized by the following technical scheme:
a method for dynamic intelligent risk assessment of hazardous chemical warehouse fires, the method comprising:
step 1, existing sample data in a dangerous chemical storehouse is obtained according to the following steps of 7: 3, dividing the ratio into a training set and a test set;
and 5, evaluating the fire risk level of the dangerous chemical storehouse based on the storehouse fire risk level intelligent evaluation model established in the step 4.
According to the technical scheme provided by the invention, the method can solve the problems of high fire risk evaluation cost, weak generalization capability and easiness in overfitting of dangerous chemical storehouses in the prior art, and has high generalization capability and prediction accuracy; can assess out the storehouse automatically and correspond conflagration risk level, strengthen dangerous chemicals storehouse process safety risk management and control, guarantee enterprise's safety.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for dynamic intelligent risk assessment of a dangerous chemical storehouse fire, according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of simulation of a Heart disease data experiment part with different optimization algorithms according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a part of simulation of data experiments for Ionosphere different optimization algorithms according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of comparison results of different kernel function experiments of Ionosphere data of the SVM according to the embodiment of the invention;
FIG. 5 is a diagram illustrating comparison of partial sample prediction results of various kernel function corresponding to the evaluation model according to an embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, not all embodiments, and this does not limit the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flow chart of a method for dynamically and intelligently evaluating a risk of a dangerous chemical warehouse fire, according to an embodiment of the present invention, where the method includes:
step 1, existing sample data in a dangerous chemical storehouse is obtained according to the following steps of 7: 3, dividing the ratio into a training set and a test set;
in this step, the sample data existing in the hazardous chemicals warehouse includes: inventory quantity, temperature increase, humidity, increase in combustible gas concentration to Lower Explosion Limit (LEL) ratio. Some sample data is shown in the following table:
partial sample of warehouse sample database
in this step, the kernel function is generally divided into a local kernel function and a global kernel function, and even if the kernel function is improved, a single kernel function cannot well give consideration to both learning ability and generalization ability.
The kernel functions commonly used by the current support vector machine SVM include a LINEAR kernel function (LINEAR), a gaussian kernel function (RBF), a polynomial kernel function (poly), and a Sigmoid kernel function, wherein the most commonly used gaussian kernel function becomes one of typical local kernel functions because of its strong learning ability and weak generalization ability, and the mathematical form of the gaussian kernel function (RBF) in the support vector machine SVM is as follows:
wherein, x is any point in space; y is a specified center point in space; sigma is a width parameter of the function and is used for controlling the radial action range of the function;
the RBF kernel function cannot guarantee that the expansion and contraction rate is always larger than 1, and when the RBF expansion and contraction rate is smaller than 1, the distance between two points in the sample space is larger than the distance between two points mapped to the high-dimensional space, so that the classification is not facilitated.
The gaussian kernel function is deformed, and the mathematical form of the improved gaussian kernel function is as follows:
in the formula, c is the minimum external sphere center of the training sample in the original space;
the expansion rate of the improved Gaussian kernel function NRBF is always larger than 1, the distance between local vectors can be amplified through implicit kernel mapping, but the radius of an external hypersphere cannot be enlarged, namely the whole mapping vector is contained in a feature space, and larger margin is possible, so that sample classification is facilitated;
the polynomial kernel function Poly in the support vector machine SVM is a typical global kernel function and has strong generalization capability, and the mathematical form is as follows:
K(x,y)＝(x·y)((x·y)+1) q-1 (3)
in the formula, q represents the dimension of the function;
in order to improve the performance of the SVM model, the advantages of a global kernel function and a local kernel function are integrated, an improved Gaussian kernel function and a polynomial kernel function are linearly combined to form an improved mixed kernel function NP kernel, and finally, the support vector machine SVM model NPSVM of the improved mixed kernel function is established, wherein the mathematical form of the NPSVM model NPSVM is as follows:
K(x,y)＝(1-α)K NRBF +αK Ploy (4)
wherein, K NRBF Represents a modified gaussian kernel, Kploy represents a polynomial kernel, and α represents a weight coefficient.
in this step, the ESDA optimizing the NPSVM model parameters includes: penalty factor C, kernel function parameter g and kernel weight factor alpha in the mixed kernel;
the specific process is as follows:
(1) performing population initialization, randomly generating electronic equipment from a search space, wherein the electronic equipment is particles in an electrostatic discharge algorithm (ESDA), each electronic equipment is composed of different components similar to design variables, the fitness value of the electronic equipment is related to the position of the electronic equipment in the search space, the higher the fitness value of the electronic equipment is, the stronger the immunity of the electronic equipment is, and each electronic equipment is provided with a counter for counting the damage times of the electronic equipment, and the initialization of the electronic equipment is shown as a formula (6):
X i ＝LB+rand·(UB-LB) (6)
wherein, X i Is the ith electronic device location; UB and LB are the upper and lower bounds of the search space, respectively;
(2) after the population is initialized, iteration is carried out for a set number of times to calculate the optimal solution of the target problem, in each iteration, three electronic devices are randomly selected from the initial population, are ranked from high to low according to fitness values and are marked as an electronic device 1, an electronic device 2 and an electronic device 3;
if the generated random number r 1 >0.5, the electronic device 3 with the lowest fitness value moves to the electronic device 1 with the highest fitness value, and the new location update formula is shown as formula (7):
X 3NEW ＝X 3 +2·β 1 ·(X 1 -X 3 ) (7)
wherein, X 3NEW Is the new position of the electronic device 3, having the lowest fitness value; x 1 And X 3 Old locations of the electronic device 1 and the electronic device 3, respectively; beta is a 1 Is a random number generated from a normal distribution, with a mean μ of 0.7 and a standard deviation parameter σ of 0.2; when the electronic device 3 moves into the vicinity of the electronic device 1, it is assumed that electrostatic discharge occurs between the two objects and the electronic device 3 is damaged;
if the generated random number r 1 If the fitness value is less than or equal to 0.5, the electronic equipment 3 with the lowest fitness value moves towards the electronic equipment 1 and 2 with the higher fitness value, and the new position updating formula is shown as the formula (8):
X 3NEW ＝X 3 +2β 2 ·(X 1 -X 3 )+2β 3 ·(X 2 -X 3 ) (8)
where β 2 and β 3 are random numbers generated from a normal distribution, the mean μ is 0.7, and the standard deviation parameter σ is 0.2; when the electronic device 3 moves to the vicinity of the electronic device 1 and the electronic device 2, it is assumed that electrostatic discharge occurs and the electronic device 3 is damaged;
(3) checking the devices one by one, if the number of times of electrostatic discharge of a certain electronic device exceeds 3, considering that the electronic device is damaged, replacing the electronic device, and randomly generating a new electronic device in a search space;
if an electronic device is subjected to electrostatic discharge less than or equal to 3 times, a random number r is generated 2 If a random number r 2 <0.2, the component of the electronic equipment is considered to be damaged and must be replaced, otherwise, the component is changed; if r is 2 More than or equal to 0.2, the components of the electronic equipment are considered to be safe;
and adding the newly generated electronic equipment into the original electronic equipment population, storing, sequencing the electronic equipment in a descending order according to the fitness, and selecting the first 20 electronic equipment as the electronic equipment population of the next iteration if the specified electronic equipment population number is 20.
in the step, the training set data in the step 1 is input into an ESDA-NPSVM model for training and learning, and corresponding fire risk levels are output, so that a warehouse fire risk level intelligent evaluation model is finally formed through a large amount of training and learning.
For example, the input inventory quantity is 44.2%, the temperature increment is 1.2 ℃, the humidity is 64%, and the increment of the ratio of the combustible gas concentration to the Lower Explosion Limit (LEL) is 0.102%, the ESDA-NPSVM prediction model can output a fire risk level I, and thus, the intelligent evaluation model of the warehouse fire risk level is finally formed through a large amount of training and learning.
In the concrete implementation, the fire risk levels of the storeroom can be respectively three levels according to the actual conditions and by combining the expert opinions, and as shown in the following table, the fire risk levels corresponding to the attributes of the sample set and the attribute values of the samples are determined by referring to the actual conditions of the storeroom and the expert opinions in the chemical safety field.
Storehouse fire risk grade table
And 5, evaluating the fire risk level of the dangerous chemical storehouse based on the storehouse fire risk level intelligent evaluation model established in the step 4.
In the step, firstly, an enterprise safety management platform acquires the current safety state data of a storehouse in real time and transmits the current safety state data of the storehouse to an established storehouse fire risk level intelligent evaluation model;
evaluating the fire risk level of the storehouse by the storehouse fire risk level intelligent evaluation model;
then judging whether the fire risk level of the storehouse is increased, if not, judging whether the set interval time is full or not since the last evaluation, and if so, repeating the evaluation operation;
if the fire risk level of the storehouse is increased, the safety management personnel check the storehouse and judge whether the evaluation result is correct, and if the evaluation result is correct, the collected current safety state data of the storehouse is used as a sample to be added into a storehouse sample set; if the evaluation result is judged to be incorrect, the wrong sample is given a correct result and then added into the storehouse sample set;
if the change of the warehouse sample set is detected, the intelligent warehouse fire risk level evaluation model continuously conducts training learning and testing on new samples, automatic updating of the model is achieved, and therefore model evaluation performance is continuously improved.
It is noted that those skilled in the art will recognize that embodiments of the present invention are not described in detail herein.
The following description is a simulation of the effectiveness of the evaluation method according to the embodiment of the present invention with specific examples:
(1) effectiveness of the electrostatic discharge algorithm ESDA
In order to verify the effectiveness of the ESDA on model parameter optimization, the optimization performances of the commonly used GSM, GA and PSO with better performances are compared with the optimization performances of the ESDA, and the specific test scheme is as follows: firstly, dividing a sample set into a training set and a test set, wherein the training set adopts 5-fold cross validation; then, determining an optimal parameter by a parameter optimization algorithm according to the parameter fitness (5-fold cross validation average accuracy rate corresponding to the parameter), and establishing four SVM prediction models (GSM-SVM, GA-SVM, PSO-SVM and ESDA-SVM) based on the optimization algorithm according to the optimal parameter; and then, predicting the same test set by each model, and finally judging the performance of the optimization of the algorithm parameters according to the prediction accuracy of each model.
In order to ensure the effectiveness of the experiment, the population numbers and the iteration times of PSO, GA and ESDA are kept consistent, and the test data set is a public data set. For a Heart disease sample set, wherein the sample attributes are 13, the tag type is 2, the data set comprises 270 samples, and the sample attribute is as follows, namely 7: the ratio of 3 is divided into a training set and a test set, the number of training samples is 189, and the number of test samples is 81. Fig. 2 is a schematic diagram of simulation of a Heart disease data experiment part of different optimization algorithms according to an embodiment of the present invention, and the test performance comparison is shown in table 1:
table 1 comparison of experimental results of different optimization algorithm Heart disease data based on RBF kernel function
As can be seen from table 1, for the Heart disease sample set, the correct results of the models corresponding to the optimal parameters obtained by the calculation of GSM, GA, and PSO are predicted to be 69, the incorrect results to be 12, and the correct rate to be 85.19% among 81 test samples; the correct results of the optimal parameters obtained by the ESDA corresponding to model prediction are 70, the error results are 11, and the correct rate is 86.42%. As can be seen from fig. 2: the model prediction result difference corresponding to the optimized parameters of GSM, GA, PSO and ESDA is sample No. 30, in order to show the specific difference, sample No. 30-33 is selected to show the prediction result of each model, and only the model corresponding to ESDA predicts correctly for sample No. 30.
For the Ionosphere sample set, where the sample attributes are 34, tag type 2 (1 and-1, respectively), the data set includes 351 samples, as per 7: the ratio of 3 is divided into a training set and a test set, the number of training samples is 246, and the number of test samples is 105. Fig. 3 is a schematic diagram of a part of simulation of data experiments for Ionosphere different optimization algorithms according to the embodiment of the present invention, and the test performance comparison is shown in table 2:
table 2 comparison of experimental results of Ionosphere data of different optimization algorithms based on RBF kernel function
As shown in table 2, the GSM, GA, PSO and ESDA models predicted the same and correct prediction for 99 samples of the 105 test samples of Ionosphere, and the samples with differences in the model prediction results include sample nos. 39, 52, 53, 82, 95 and 100. As can be seen in fig. 3: in the above 6 samples, 4 samples of the ESDA corresponding model are predicted correctly, and the number of predicted correctly is the largest. The accuracy of ESDA was up to 98.10% in the prediction of the 105 test samples of Ionosphere.
The comparison simulation experiment can obtain that the classification accuracy of the model corresponding to the ESDA optimization parameters is higher than that of the other three optimization algorithms, and the three optimization algorithms are commonly used in practical research and have excellent performance, so that the effectiveness of the electrostatic discharge algorithm in the aspect of SVM parameter optimization can be basically proved.
(2) Comparative simulation experiment of NRBF nuclear performance
To verify the validity of the NRBF kernel function, a simple performance comparison was made of RBF and NRBF.
In the experiment, the value range of the parameter C is 2 -10 -2 10 Step size of 2 0.5 G has a value range of 2 -10 -2 10 Step size of 2 0.5 . Wherein, the optimal parameter result in the table reserves four digits after the decimal point, and the test accuracy reserves one digit after the decimal point. The training set is cross-validated by 5-fold, as shown in fig. 4, a schematic diagram of comparison results of different kernel function experiments of Ionosphere data of the SVM according to the embodiment of the present invention is shown, and the comparison results are shown in table 3:
table 3 comparison of experimental results of different kernel function Ionosphere data based on grid algorithm
As can be seen from Table 3, the prediction accuracy of the NRBF core is slightly higher than that of the RBF core. The prediction results of 100 samples of 105 test samples of Ionosphere of the RBF and NRBF corresponding models are the same and correct, and the samples with different model prediction results are No. 52, 53, 82, 95 and 100 samples. As can be seen in fig. 4: in the above 5 samples, 3 samples of the NRBF corresponding model are correctly predicted, and 2 samples of the RBF corresponding model are correctly predicted.
(3) Example static simulation experiment
In order to verify the feasibility of the method, the method is applied to the stabilized chemical industry park, a stabilized chemical industry park class A library sample database is established, the database comprises 150 samples, and the method comprises the following steps of: the ratio of 3 is divided into a training set and a test set, the number of training samples is 105, and the number of test samples is 45.
In the application, 45 class a library test samples are predicted by combining 5 SVM models with different kernel functions with ESDA respectively, 5-fold cross validation is adopted in a training set, as shown in FIG. 5, a comparison diagram of prediction results of partial samples of various kernel functions corresponding to an evaluation model is shown, and samples with the same model prediction results are not shown in the diagram. The prediction results of the data of the models are shown in table 4, and the data prediction accuracy of the different models is shown in table 5:
TABLE 4 prediction results of different models for the class A library test set data of the stabilized chemical industry park
TABLE 5 ESDA-based comparison of results of fire dynamic risk assessment for different Kernel function class A libraries
From table 4, it can be seen that the evaluation models corresponding to different kernel functions have specific prediction results for 45 test samples, where the evaluation results of the models are the same and the prediction is correct except for sample nos. 5, 11, 28, 30, 38, and 41.
Fig. 5 shows the specific prediction results of each model for the above 6 samples, and it can be seen from fig. 5 that the prediction results of the NP core for the model to be evaluated are the most consistent with the actual results (actual label).
It can be seen from table 5 that the accuracy of the NP nuclear model assessment is 97.78% at the highest. Experiments show that the ESDA-NPSVM model has good intelligent evaluation capability on fire risk level of a dangerous chemical class A library.
(4) Dynamic simulation experiment of ESDA-NPSVM model
In actual production and management of chemical enterprises, due to the conditions of warehouse entry and exit and the like, safety indexes such as the quantity of stocks in dangerous chemical storehouses and the temperature of the storehouses are always in a dynamic change state, and in order to objectively and accurately master the storehouse fire risk, the application provides an ESDA-NPSVM storehouse fire dynamic risk assessment model for dynamic fire risk assessment.
The time interval for evaluating the fire risk of the storehouse is set to be 10 minutes, the time interval is variable, and enterprises can set the time interval according to actual conditions in application. Although the SVM is suitable for a small sample set, 150 samples are used for model training, the number of training samples is relatively small, and in order to enable the ESDA-NPSVM evaluation model to be continuously perfect in practical application, a reasonable dynamic risk evaluation sample set updating method is designed according to the actual situation of a dangerous chemical storehouse, so that the sample set is continuously updated and expanded in the actual evaluation process of the model. When the fire risk level of the storehouse is increased, the safety management personnel check the storehouse and judge whether the evaluation result is correct, if the evaluation result is wrong, the group of data is endowed with a correct level result and added into a class A storehouse sample set, and if the evaluation result is correct, the group of data is directly added into the sample set. Once the change of the sample set is detected, the model is automatically updated, and the evaluation performance of the model is continuously perfected.
In order to test the effectiveness of the ESDA-NPSVM storehouse fire dynamic risk intelligent evaluation model in practical application, in the practically collected real-time data of the storehouse of the dangerous chemicals in the class A storehouse, 10 groups of the data are randomly selected to verify the performance of the ESDA-NPSVM evaluation model, and the results are shown in Table 6:
TABLE 6 evaluation results of ESDA-NPSVM model for real-time data of storehouse
In order to verify the accuracy of the evaluation result of the ESDA-NPSVM evaluation model on 10 groups of data, 5 experts in the chemical industry field are invited to evaluate the 10 groups, and the evaluation result is shown in Table 7:
TABLE 7 expert evaluation results of class A library real-time data
As can be seen from tables 6 and 7: the evaluation result of the ESDA-NPSVM storeroom fire dynamic risk intelligent evaluation model on the actually acquired 10 groups of data is consistent with the expert evaluation result, and the model is effective in the actual storeroom fire risk intelligent evaluation and has an actual application value.
In summary, the classification performance of the SVM algorithm adopted by the method in the embodiment of the present invention depends on kernel functions and parameter selection, and in the aspect of parameter selection, because an electrostatic discharge algorithm (ESDA) has the advantages of fast convergence rate and high optimization precision, the method of the present invention adopts the ESDA with better performance to perform parameter optimization, and improves the SVM performance in the aspect of parameter selection; meanwhile, the method improves the traditional Radial Basis Function (RBF), and provides an NP (nuclear processor) mixed kernel which has strong learning capacity and generalization capacity and can effectively improve the evaluation accuracy of an evaluation model.
According to the method and the system, the corresponding fire risk level of the storehouse can be automatically evaluated, the safety of an enterprise is ensured, the dynamic fire risk level of the storehouse is mastered in real time, and accidents are effectively prevented; meanwhile, the method and the system have high prediction precision, can effectively identify the process risk, have high safety dynamic intelligent risk assessment capability of the dangerous chemical storage process, and provide guidance for enterprise safety management.
The above description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are also within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims. The information disclosed in this background section is only for enhancement of understanding of the general background of the invention and should not be taken as an acknowledgement or any form of suggestion that this information forms the prior art that is already known to a person skilled in the art.
Claims (6)
1. A method for dynamic intelligent risk assessment of dangerous chemical storehouse fire, which is characterized by comprising the following steps:
step 1, existing sample data in a dangerous chemical storehouse is obtained according to the following steps of 7: 3, dividing the ratio into a training set and a testing set;
step 2, establishing a Support Vector Machine (SVM) model of the improved mixed kernel function, namely an NPSVM model, by linearly combining the improved Gaussian kernel function and the polynomial kernel function;
step 3, optimizing the NPSVM model parameters established in the step 2 by adopting an electrostatic discharge algorithm (ESDA), and finally forming an ESDA-NPSVM prediction model; the method for optimizing the NPSVM model parameters by the electrostatic discharge algorithm ESDA comprises the following steps: penalty factor C, kernel function parameter g and kernel weight factor alpha in the mixed kernel;
step 4, training the prediction model by utilizing the training set and the testing set divided in the step 1 based on the ESDA-NPSVM prediction model established in the step 3 to obtain an intelligent evaluation model of the fire risk level of the storehouse;
and 5, evaluating the fire risk level of the dangerous chemical storehouse based on the storehouse fire risk level intelligent evaluation model established in the step 4.
2. The method for fire dynamic intelligent risk assessment of dangerous chemical storehouse according to claim 1, wherein in step 1, the sample data existing in the dangerous chemical storehouse comprises: inventory quantity, temperature increase, humidity, increase in combustible gas concentration to lower explosion limit ratio.
3. The method for dynamic intelligent risk assessment of dangerous chemical storehouse fire according to claim 1, wherein the process of step 2 specifically comprises:
the mathematical form of the gaussian kernel function in the support vector machine SVM is as follows:
wherein x is any point in space; y is a specified center point in space; sigma is a width parameter of the function and is used for controlling the radial action range of the function;
the Gaussian kernel function cannot guarantee that the expansion rate is always greater than 1, and when the RBF expansion rate is less than 1, the distance between two points in a sample space is greater than the distance between two points mapped to a high-dimensional space, so that the classification is not facilitated;
the gaussian kernel function of equation (1) is transformed, and the mathematical form of the modified gaussian kernel function is as follows:
in the formula, c is the minimum external sphere center of the training sample in the original space;
the expansion rate of the improved Gaussian kernel function NRBF is always larger than 1, the distance between local vectors can be amplified through implicit kernel mapping, but the radius of an external hyper-sphere cannot be enlarged, namely the whole mapping vector is contained in a feature space, and sample classification is facilitated;
the polynomial kernel Poly in the support vector machine SVM is a typical global kernel, and its mathematical form is as follows:
K(x,y)＝(x·y)((x·y)+1) q-1 (3)
in the formula, q represents the dimension of the function;
in order to integrate the advantages of the global kernel function and the local kernel function, the improved Gaussian kernel function and the polynomial kernel function are linearly combined to form an improved mixed kernel function NP kernel, and finally a support vector machine SVM model NPSVM of the improved mixed kernel function is established, wherein the mathematical form of the model NPSVM is as follows:
K(x,y)＝(1-α)K NRBF +αK Ploy (4)
wherein, K NRBF Represents a modified gaussian kernel function; kcloy represents a polynomial kernel; and alpha represents a weight coefficient.
4. The method for dynamic intelligent risk assessment of dangerous chemical storehouse fire according to claim 1, wherein the process of step 3 specifically comprises:
(1) performing population initialization, randomly generating electronic equipment from a search space, wherein the electronic equipment is particles in an electrostatic discharge algorithm (ESDA), each electronic equipment is composed of different components similar to design variables, the fitness value of the electronic equipment is related to the position of the electronic equipment in the search space, the higher the fitness value of the electronic equipment is, the stronger the immunity of the electronic equipment is, and each electronic equipment is provided with a counter for counting the damage times of the electronic equipment, and the initialization of the electronic equipment is shown as a formula (6):
X i ＝LB+rand·(UB-LB) (6)
wherein, X i Is the ith electronic device location; UB and LB are the upper and lower bounds of the search space, respectively;
(2) after the population is initialized, iteration is carried out for a set number of times to calculate the optimal solution of the target problem, in each iteration, three electronic devices are randomly selected from the initial population, are ranked from high to low according to fitness values and are marked as an electronic device 1, an electronic device 2 and an electronic device 3;
if the generated random number r 1 >0.5, the electronic device 3 with the lowest fitness value moves to the electronic device 1 with the highest fitness value, and the new location update formula is shown as formula (7):
X 3NEW ＝X 3 +2·β 1 ·(X 1 -X 3 ) (7)
wherein, X 3NEW Is the new position of the electronic device 3, having the lowest fitness value; x 1 And X 3 Old locations of the electronic device 1 and the electronic device 3, respectively; beta is a 1 Is a random number generated from a normal distribution, with a mean μ of 0.7 and a standard deviation parameter σ of 0.2; when the electronic device 3 moves into the vicinity of the electronic device 1, it is assumed that electrostatic discharge occurs between the two objects and the electronic device 3 is damaged;
if the generated random number r 1 If the fitness value is less than or equal to 0.5, the electronic equipment 3 with the lowest fitness value moves towards the electronic equipment 1 and 2 with the higher fitness value, and the new position updating formula is shown as the formula (8):
X 3NEW ＝X 3 +2β 2 ·(X 1 -X 3 )+2β 3 ·(X 2 -X 3 ) (8)
where β 2 and β 3 are random numbers generated from a normal distribution, the mean μ is 0.7, and the standard deviation parameter σ is 0.2; when the electronic device 3 moves to the vicinity of the electronic device 1 and the electronic device 2, it is assumed that electrostatic discharge occurs and the electronic device 3 is damaged;
(3) checking the devices one by one, if the number of times of electrostatic discharge of a certain electronic device exceeds 3, considering that the electronic device is damaged, replacing the electronic device, and randomly generating a new electronic device in a search space;
if an electronic device is subjected to electrostatic discharge less than or equal to 3 times, a random number r is generated 2 If a random number r 2 <0.2, the component of the electronic equipment is considered to be damaged and must be replaced, otherwise the component is replacedChanging; if r is 2 More than or equal to 0.2, the components of the electronic equipment are considered to be safe;
and adding the newly generated electronic equipment into the original electronic equipment population, storing, sequencing the electronic equipment in a descending order according to the fitness, and selecting the first 20 electronic equipment as the electronic equipment population of the next iteration if the specified electronic equipment population number is 20.
5. The method for dynamically and intelligently evaluating the fire risk of the dangerous chemical storehouse according to the claim 1, wherein in the step 4, based on the ESDA-NPSVM prediction model established in the step 3, the training set data in the step 1 is input into the ESDA-NPSVM model for training and learning, and the corresponding fire risk level is output, so that the storehouse fire risk level intelligent evaluation model is finally formed through a large amount of training and learning.
6. The method for dynamic intelligent risk assessment of dangerous chemical storehouse fire according to claim 1, wherein the process of step 5 specifically comprises:
firstly, an enterprise safety management platform acquires the current safety state data of a storehouse in real time and transmits the current safety state data of the storehouse to an established storehouse fire risk level intelligent evaluation model;
evaluating the fire risk level of the storehouse by the storehouse fire risk level intelligent evaluation model;
then judging whether the fire risk level of the storehouse is increased, if not, judging whether the set interval time is full or not before the last evaluation, and if so, repeating the evaluation operation;
if the fire risk level of the storehouse is increased, the safety management personnel check the storehouse and judge whether the evaluation result is correct, and if the evaluation result is correct, the collected current safety state data of the storehouse is used as a sample to be added into a storehouse sample set; if the evaluation result is judged to be incorrect, the wrong sample is given a correct result and then added into the storehouse sample set;
if the change of the warehouse sample set is detected, the intelligent warehouse fire risk level evaluation model continuously conducts training learning and testing on new samples, automatic updating of the model is achieved, and therefore model evaluation performance is continuously improved.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210819137.4A CN115099673A (en) | 2022-07-13 | 2022-07-13 | Method for dynamically and intelligently evaluating dangerous chemical storehouse fire hazard risk |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210819137.4A CN115099673A (en) | 2022-07-13 | 2022-07-13 | Method for dynamically and intelligently evaluating dangerous chemical storehouse fire hazard risk |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115099673A true CN115099673A (en) | 2022-09-23 |
Family
ID=83297286
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210819137.4A Pending CN115099673A (en) | 2022-07-13 | 2022-07-13 | Method for dynamically and intelligently evaluating dangerous chemical storehouse fire hazard risk |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115099673A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115375204A (en) * | 2022-10-25 | 2022-11-22 | 中国人民解放军陆军装甲兵学院 | Performance evaluation method for vehicle-mounted intelligent micro-grid |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109242223A (en) * | 2018-11-26 | 2019-01-18 | 武汉理工光科股份有限公司 | The quantum support vector machines of city Public Buildings Fire Risk is assessed and prediction technique |
-
2022
- 2022-07-13 CN CN202210819137.4A patent/CN115099673A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109242223A (en) * | 2018-11-26 | 2019-01-18 | 武汉理工光科股份有限公司 | The quantum support vector machines of city Public Buildings Fire Risk is assessed and prediction technique |
Non-Patent Citations (1)
| Title |
|---|
| 王浩: "基于化工园区火灾风险等级安全巡检机器人路径优化", 《中国优秀硕士学位论文全文数据库》, 15 February 2022 (2022-02-15) * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115375204A (en) * | 2022-10-25 | 2022-11-22 | 中国人民解放军陆军装甲兵学院 | Performance evaluation method for vehicle-mounted intelligent micro-grid |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | How to design rating schemes of risk matrices: a sequential updating approach | |
| Esmaeili et al. | Attribute-based safety risk assessment. II: Predicting safety outcomes using generalized linear models | |
| Lu et al. | An MDL approach to the climate segmentation problem | |
| US20120084251A1 (en) | Probabilistic data mining model comparison | |
| Carmona et al. | Overview on evolutionary subgroup discovery: analysis of the suitability and potential of the search performed by evolutionary algorithms | |
| Zhu et al. | [Retracted] Analysis of the Impact of Climate Change on National Vulnerability Based on Fuzzy Comprehensive Evaluation | |
| CN110472678A (en) | Optimization method based on the information system for power enterprise abnormality detection scheme for improving SVM | |
| KR102550014B1 (en) | System and Method for providing early warning to prevent false alarm | |
| CN115099673A (en) | Method for dynamically and intelligently evaluating dangerous chemical storehouse fire hazard risk | |
| Briand et al. | Modeling and managing risk early in software development | |
| Say et al. | Planning in factored state and action spaces with learned binarized neural network transition models | |
| Akyar | Fuzzy risk analysis for a production system based on the Nagel point of a triangle | |
| CN115203167A (en) | Data detection method and device, computer equipment and storage medium | |
| Gondia et al. | Machine learning–based decision support framework for construction injury severity prediction and risk mitigation | |
| Nadjafi et al. | Dynamic fault tree analysis using fuzzy LU bounds failure distributions | |
| Camelia et al. | A Computational Grey Based Model for Companies Risk Forecasting. | |
| Jakšić et al. | Fuzzy approach in ranking of banks according to financial performances | |
| US10867135B1 (en) | System and method for combining expert knowledge and deep learning to identify and classify regulatory maintenance compliances | |
| CN117333011A (en) | Fire risk assessment method, device, equipment and medium | |
| CN111767212A (en) | Software code quality evaluation method, device, equipment and storage medium | |
| Yahyaei et al. | Multi-criteria logistic hub location by network segmentation under criteria weights uncertainty | |
| Zhigirev et al. | Introducing a Novel Method for Smart Expansive Systems' Operation Risk Synthesis. Mathematics 2022, 10, 427 | |
| Niloofar et al. | Data-driven modelling of repairable fault trees from time series data with missing information | |
| Hasanin et al. | Experimental Studies on the Impact of Data Sampling with Severely Imbalanced Big Data | |
| Wu et al. | A fuzzy identification procedure for nonlinear time series: With example on ARCH and bilinear models |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |