CN103905560B - Oil tank area flame spread prediction and analysis method based on complex networks - Google Patents

Abstract

The invention provides an oil tank area flame spread prediction and analysis method based on complex networks. The method includes the steps of arranging sensor nodes in the positions of all oil tanks of an oil tank area and in the position between every two adjacent oil tanks, connecting the nodes to form the sensor network, establishing the weighted directed network of oil tank area flame spread in a computer system, and conducting flame spread prediction and related analysis on the basis of the established weighted directed network by means of the occurrence state of a fire disaster and environmental data monitored by the sensor nodes. According to the method, an oil tank area flame spread model is established according to a complex network thought, the fire disaster state of the oil tank area can be comprehensively monitored, the fire disaster spread tendency of the oil tank area can be predicted, related analysis is conduced, and the reference spread time is provided through the combination with a least square support vector machine model.

Description

Oil tank district fire spread prediction and analysis method based on complex network

Technical Field

The invention relates to the field of fire prediction, in particular to a method for predicting and analyzing fire spread of an oil tank region based on a complex network.

Background

Materials in petroleum production are often inflammable, explosive and the like, and the potential dangers determine that strict management and control are needed in multiple links of production, transportation, storage and the like, if careless, serious accidents can be caused, and damage and loss are caused to lives and properties of people. The oil tank area occupies a large area, the receiving and dispatching operation is frequent, the equipment maintenance times are many, and the oil tank area is a place where fire disasters frequently occur. Therefore, if modeling and research can be performed on a tank fire to predict a change trend and a future trend of combustion characteristics when the fire occurs, it is very significant to extinguish the fire and rescue the fire in the field.

At present, the related research from the theoretical and experimental points of view has been carried out in the prior art for tank fires. The method is characterized in that the method carries out numerical simulation on combustion of the diesel oil tank (heat radiation hazard characteristics [ J ] of fire of the large oil tank, 2008,8(4): 110-. The safety analysis of the oil tank fire simulation test with reduced size is carried out by adopting mathematical calculation and FDS software simulation (safety analysis [ J ] of oil tank fire model test, 122-125. 2012,31 (2)), the summer construction army and the like, and parameter data such as flame jumping frequency, inclination angle, flame height, heat radiation safety distance and the like are obtained and are compared with the actual model test. Zhao hong hai (small-size experimental research [ J ] of fire characteristics of oil tank fire science and technology, 2010(1): 26-29.) researches the combustion characteristics of oil tank fire under different oil tank diameters, oil tank heights, water cushion layer thicknesses and oil layer thicknesses and the distribution rules of ambient heat radiation flux, and analyzes the influence of the factors on the combustion process.

The above prior arts have focused on the research on the combustion characteristics of a single oil tank, and have a certain significance for the research on the spread of fire in the oil tank area, but the prior arts cannot consider the influence of the combustion of the oil tank on the surrounding environment from the perspective of the whole oil tank area, and also do not relate to the overlapping influence when a plurality of oil tanks are simultaneously fired and the mutual influence among the oil tanks in different states, and do not deeply research and analyze the complexity of the spread of fire in the oil tank area. Therefore, the prior art cannot meet the requirements of accurately and rapidly monitoring the fire state of the oil tank field and predicting the fire spread.

Disclosure of Invention

The invention provides a method for predicting and analyzing fire spread of an oil tank area based on a complex network by integrating the information of the whole oil tank area from the overall view of the oil tank area, and the method realizes the fire spread prediction, the fire spread reference time analysis and the safety distance analysis by using an oil tank area fire spread model.

In order to achieve the purpose, the technical scheme adopted by the invention is as follows:

a method for predicting and analyzing fire spread of an oil tank region based on a complex network comprises the following steps:

sensor nodes are arranged at the positions of the oil tanks in the oil tank area and the middle positions of the adjacent oil tanks, and the nodes are connected to form a sensor network;

establishing a weighted directed network of tank farm fire spread within a computer system; and

and based on the established weighted directed network, carrying out fire spread prediction and correlation analysis by using the occurrence state of the fire and the environmental data monitored by the sensor nodes.

In a further embodiment, the environmental information collected by the sensor node at least includes: temperature, humidity, carbon monoxide concentration, carbon dioxide concentration, combustible gas concentration, wind speed, and wind direction.

In a further embodiment, the establishment of the weighted directed network comprises the steps of:

step 1, abstracting each oil tank of an oil tank area as a network node to form an oil tank set V, V = (V)₁,v₂,…,v_m﹜；

And 2, giving each network node attribute for reflecting the state information of each oil tank to form an attribute set A:

A=｛p,d,o,s,q｝，

wherein: p is the node position, d is the node diameter, o is the oil storage property, s is the node state, and q is the thermal radiation intensity received by the node center;

step 3, abstracting a fire spreading path into directed edges to form an edge set E:

E=﹛e_ij︱i≤m,j≤m,i≠j﹜，

wherein the set E is a subset of V × V, and the edge E_ijIndicating a directed edge between a tank that has fired and an adjacent unlit tank as a potential fire propagation path, i.e. the fire may be directed from the start point vi of the edge towards the end point v of the edge_jSpreading;

step 4, endowing the edge weight value to form a weight value set W of the edge:

W=﹛w_ij︱i≤m,j≤m,i≠j﹜，

wherein a certain edge e_ijWeight value w of_ijIndicating a fire from a starting point v_iTo the end point v_jDegree of danger of spreading, w_ij∈[0,1]When an edge e_ijAfter the indicated fire spread does occur, w_ij=1, i.e.:

and 5, establishing a fire spreading weighted directed network G of the oil tank area, wherein G is a triple < V, E, W >.

And 6, in the established weighted directed network G, based on the fire occurrence process and the node state change monitored by the sensor network, carrying out structure growth and numerical update on the weighted directed network G:

step 6-1, structure growth

When there are one or more nodes v at a time_iWhen the combustion state of the tank in the attribute changes and the degree of danger represented by the tank changes, a plurality of new adjacent nodes v are added into the weighted directed network_jAs a new potential fire spread node, the direction of the newly introduced edge is from v_iDirection v_jDetermining the edge weight value is related to the intensity of the heat radiation received by the node center;

wherein: the new node v_jFrom node v with a change in state_iA set of neighbor nodes of (1);

step 6-2, updating numerical value

a) When the node state changes, the attribute values of the nodes in the weighted directed network, the weight values of the edges and other output results are re-determined;

b) when the structure is increased, re-determining attribute values of the relevant nodes in the weighted directed network, the weights of the relevant edges and other output results;

c) and when the data collected by the sensor network is updated, re-determining the attribute values of the relative nodes in the weighted directed network, the weight values of the edges and other results.

In a further embodiment, in the step 2, the intensity of the heat radiation received by the node center is calculated as follows:

q＝τFE_act，

wherein q represents the intensity of heat radiation at x meters from the center of the tank on fire, F represents the view angle factor, E_actRepresents the actual radiation energy of the flame surface, and τ represents the atmospheric transfer coefficient, where:

a) the atmospheric transmission coefficient τ is calculated according to the following formula:

τ＝1-0.058lnx，

in the formula: x represents the distance from the center of the flame to the target;

b) coefficient of viewing angle F

The shape of the flame is equivalent to a cylinder, the bottom surface of the flame is the same as the cross section of the oil tank, and the height of the flame is H_fThe inclination angle of the flame is theta,

height H of flame_fCalculated according to the following formula:

$H_f=55D{\left(\frac{m_f^{\&prime;}}{{\mathrm{\&rho;}}_0\sqrt{\mathrm{gD}}}\right)}^{0.67}{\left(\frac{U_W}{U_C}\right)}^{-0.21},$

$U_c={\left(\frac{g{m}_f^{\&prime;}D}{{\mathrm{\&rho;}}_0}\right)}^{\frac{1}{3}},$

in the formula, ρ₀Is air density, U_WRepresenting wind speed, D representing tank diameter, m_fRepresenting the fuel combustion rate, g represents the gravitational acceleration, and is 9.81m/s²；

The viewing angle factor F is calculated according to the following formula:

$F=\sqrt{F_h^2+F_v^2},$

in the windless state, the above F_h、F_vThe following calculation was performed:

$F_h=\frac{B-1}{\mathrm{\&pi;S}\sqrt{B^2-1}}{\tan }^{-1}\sqrt{\frac{(B+1)(S-1)}{(B-1)(S+1)}}-C\frac{A-1}{\sqrt{A^2-1}}$

$F_v=\frac{1}{\mathrm{\&pi;S}}{\tan }^{-1}\left(\frac{h}{\sqrt{S^2-1}}\right)-\frac{h}{\mathrm{\&pi;S}}{\tan }^{-1}\sqrt{\frac{S-1}{S+1}}+C\frac{\mathrm{Ah}}{\sqrt{A^2-1}}$

$A=\frac{h^2+S^2+1}{2S}$

$B=\frac{1+S^2}{2S};$

$C=\frac{1}{\mathrm{\&pi;s}}{\tan }^{-1}\sqrt{\frac{(A+1)(S-1)}{(A-1)(S-1)}}$

$S=\frac{2R}{D}$

$h=\frac{{2H}_f}{D}$

under windy conditions, the above F_h、F_vThe following calculation was performed:

$F_h={\tan }^{-1}\frac{\sqrt{\frac{b+1}{b-1}}}{\mathrm{\&pi;}\sqrt{B^2-1}}-G\&CenterDot;D+\frac{\sin \mathrm{\&theta;}}{\sqrt{C}}E+F$

$F_v=I\&CenterDot;H\&CenterDot;D+\frac{\cos \mathrm{\&theta;}}{\sqrt{C}}(E+F)-I\&CenterDot;{\tan }^{-1}\sqrt{\frac{b-1}{b+1}}$

A＝a²+[b+1)²-2a(b+1)sinθ

B＝a²+(b-1)²-2a(b-1)sinθ

C＝1+[b²-1)

$D={\tan }^{-1}\sqrt{\frac{A(b-1)}{B(b+1)}}$

$E={\tan }^{-1}\frac{\mathrm{ab}-(b^2-1)\sin \mathrm{\&theta;}}{\sqrt{C(b^2-1)}}$ $F={\tan }^{-1}\frac{(b^2-1)\sin \mathrm{\&theta;}}{\sqrt{C(b^2-1)}}$

$G=\frac{a^2+{(b+1)}^2-2(b+1)+\mathrm{ab}\sin \mathrm{\&theta;}}{\sqrt{\mathrm{AB}}}$

$H=\frac{a^2+{(b+1)}^2-2b(1+a\sin \mathrm{\&theta;})}{\sqrt{\mathrm{AB}}}$

$I=\frac{a\cos \mathrm{\&theta;}}{b-a\sin \mathrm{\&theta;}}$

$a=\frac{{2H}_f}{D}$

$b=\frac{2R}{D};$

c) actual intensity of heat radiation E on flame surface_act

Actual intensity of heat radiation E on flame surface_actCalculated according to the following formula:

$E_{\mathrm{act}}=\frac{m^{\&prime;\&prime;}{H}_c\mathrm{\&eta;}}{0.25\mathrm{\&pi;}{D}^2+\mathrm{\&pi;D}{H}_f},$

wherein m' represents a fuel combustion rate, H_cRepresenting the combustion heat of oil products, η representing the ratio of radiant heat to total combustion heat energy, wherein the value is between 0.2 and 0.5, and D represents the diameter of an oil pool;

wherein the fuel combustion rate m "is calculated according to the following formula:

m″＝m″_∞(1-e^-αD)，

in the formula, m ″)_∞Representing the burn rate at infinite tank diameter, α is a factor related to oil;

in a further embodiment, the edge weight w_ijThe value of (A) is determined according to the following formula:

$w_{\mathrm{ij}}=\left\{\begin{array}{c}1-e^{-k{Q}_{\mathrm{ij}}}\\ 1\end{array}\right.,$

in the formula, $Q_{\mathrm{ij}}={\&Integral;}_{\max (t_0,t_f-\mathrm{\&Delta;t})}^{\mathrm{tf}}{q}_{\mathrm{ij}}\mathrm{dt},$

wherein, t₀Denotes the starting time, t_fRepresenting the current time, at represents a time constant, q_ijRepresenting a node v_jReal-time receiving node v_iIntensity of released thermal radiation: if Δ t>t_f—t₀Then Q is_ijIndicating the time t from the start₀To the current time t_fAt this stageWithin, node v_jCumulative acceptance v_iA value of an amount of emitted heat radiation; if Δ t<t_f—t₀Then Q is_ijIndicating the time t from_f- Δ t to the present time t_fDuring this time, node v_jCumulative acceptance v_iA value of the amount of heat radiation is emitted.

In a further embodiment, the fire spread prediction comprises the following processes: and based on the fire occurrence state of a certain oil tank in the oil tank area and the environmental data monitored by the sensor node, carrying out fire spread prediction according to the established weighted directed network, and solving the directional edge pointed by the maximum weight in the edge weights of all the directional edges, namely the trend direction of fire spread.

In a further embodiment, the correlation analysis includes analyzing and calculating a fire spreading reference time, that is, simulation data obtained by performing a fire simulation in the tank farm using the weighted directed network, and constructing a least squares support vector machine model based on the simulation data for calculating the fire spreading reference time, and the implementation includes the following processes:

taking the environmental parameters collected by the sensor network as input data x = { x = ×)_kI k =1,2,3, … N }, input data x_kThe method comprises the steps of periodically acquiring temperature, humidity, carbon monoxide concentration and carbon dioxide concentration in a simulation process;

the actual propagation time is taken as an output value y = { y = {_kI k =1,2,3 …, N }, and output data y_kSubtracting the acquired data x from the actual fire spread moment measured by fire simulation_kThe time of (1) is obtained;

thus, a training set D of N samples is constructed: let D { (x)_k，y_k) 1,2,3, … N, where the input data is x_k∈RⁿThe output data is y_k∈ R, training data, namely the input data x_kAnd output data y_kInputting a least square support vector machine model, wherein the aim is to obtain a support vector machine model function, namely an objective discrimination function;

and finally, inputting the actually acquired fire state data into the model, and calculating to obtain the reference time for fire spreading of a certain oil tank in the actual fire occurrence process of the oil tank area.

According to the technical scheme, the oil tank area fire spread prediction and analysis method based on the complex network adopts the complex network idea to construct the oil tank area fire spread model, achieves comprehensive monitoring of the fire state of the oil tank area, predicts the fire spread trend of the oil tank area, performs correlation analysis, combines a least square support vector machine model to give spread reference time, and performs safety distance analysis. Compared with the prior art, its obvious advantage lies in:

1. and (3) accuracy: the method combines network science and oil tank fire research theories, comprehensively considers complexity and network of fire spreading in the oil tank area, and is suitable for fire spreading conditions in most oil tank areas;

2, fast speed: after the initial conditions are set, the establishment of the model can be completed by utilizing the sensor network equipment to collect data, so that human factors are reduced, and the modeling and prediction efficiency is greatly improved;

3. safety: the input quantity required by the model can be acquired by a sensor network, so that the workload of personnel in the fire disaster occurrence process is reduced, and the safety of the personnel is guaranteed.

The present invention is described in further detail below with reference to the attached drawing figures.

Drawings

Fig. 1 is a diagram showing an example of the arrangement of a tank farm sensor network.

Fig. 2 is a flow chart of modeling of a tank farm weighted directed network according to an embodiment of the present invention.

FIG. 3 is an exemplary diagram of a tank farm weighted directed net.

Fig. 4 is an equivalent tank flame diagram.

FIG. 5 is a diagram illustrating the relationship between the edge weighting and the amount of heat radiation.

FIG. 6 is a diagram of the fire spread experiment in the tank farm.

Fig. 7 is a diagram of the prediction results of the tank farm weighted directed net established in fig. 2.

FIG. 8 is a flow chart for modeling a least squares support vector machine model of a reference propagation time.

Detailed Description

In order to better understand the technical content of the present invention, specific embodiments are described below with reference to the accompanying drawings.

According to the preferred embodiment of the invention, the method for predicting and analyzing the fire spread of the oil tank region based on the complex network comprises the following steps:

sensor nodes are arranged at the positions of the oil tanks in the oil tank area and the middle positions of the adjacent oil tanks, and the nodes are connected to form a sensor network;

establishing a weighted directed network of tank farm fire spread within a computer system; and

and based on the established weighted directed network, carrying out fire spread prediction and correlation analysis by using the occurrence state of the fire and the environmental data monitored by the sensor nodes.

The following describes the implementation of the above steps of this embodiment with reference to fig. 1 to 10.

Referring to fig. 1, in the present embodiment, before the tank farm fire spread model, i.e., the weighted directed network of tank farm fire spread, is established, a network of sensors needs to be arranged in the tank farm. The sensor network is the basis for realizing the prediction and analysis of the fire spread model of the oil tank field.

The sensor network comprises a plurality of sensor nodes, and the sensor nodes are mainly arranged at each oil tank position. Furthermore, a small number of sensor nodes also need to be arranged in the area between the tanks, which nodes will be used as redundant monitoring nodes in order to improve the reliability of the sensor network and the accuracy of the model. An example of the arrangement method of the sensor nodes is shown in fig. 1, wherein circles represent oil tanks, squares represent sensor nodes, and numbers are numbers of the oil tanks (or the sensor nodes), and the number of the nodes can be increased as required in a specific environment of an oil tank area.

The sensor nodes sense the surrounding environment information by various sensors, and the nodes can communicate with each other. The environmental information collected by the sensor node comprises: temperature, humidity, carbon monoxide concentration, carbon dioxide concentration, combustible gas concentration, wind speed, wind direction, and the like. The communication between the sensor nodes can be in a wired or wireless mode, and is determined according to the field environment of the oil tank field.

In other embodiments, besides the above sensor nodes, video capture devices may be provided in the sensor network for monitoring the fire condition of the oil tank, and these video capture devices may provide more comprehensive and stereoscopic environmental information of the oil tank area to make up for the deficiencies of the sensor nodes.

All information collected by the sensor network is gathered to a computer system, such as a microcomputer system (for example, a computer system formed by a single chip microcomputer or other microprocessors) serving as an upper computer, a Personal Computer (PC) system or an industrial computer system, and the upper computer processes the information by utilizing the fire spreading model of the oil tank region and stores original data and processing results.

Fig. 2 shows a modeling process of a weighted directed network in a tank farm according to an embodiment of the present invention, and as a preferred embodiment, the establishment of the weighted directed network includes the following steps:

step 1, abstracting each oil tank of an oil tank area as a network node to form an oil tank set V, V = (V)₁,v₂,…,v_m﹜；

And 2, giving each network node attribute for reflecting the state information of each oil tank to form an attribute set A:

A=｛p,d,o,s,q｝，

wherein: p is the node position, d is the node diameter, o is the oil storage property, s is the node state, and q is the thermal radiation intensity received by the node center;

step 3, abstracting a fire spreading path into directed edges to form an edge set E:

E=﹛e_ij︱i≤m,j≤m,i≠j﹜，

wherein the set E is a subset of V × V, and the edge E_ijIndicating a directed edge between a tank that has fired and an adjacent unlit tank as a potential fire propagation path, i.e. the fire may be directed from the start point vi of the edge towards the end point v of the edge_jSpreading;

step 4, endowing the edge weight value to form a weight value set W of the edge:

W=﹛w_ij︱i≤m,j≤m,i≠j﹜，

wherein a certain edge e_ijWeight value w of_ijIndicating a fire from a starting point v_iTo the end point v_jDegree of danger of spreading, w_ij∈[0,1]When an edge e_ijAfter the indicated fire spread does occur, w_ij=1, i.e.:

step 5, establishing a fire spreading weighted directed network G of the oil tank area, wherein G is a triple<V，E，W>. A tank farm weighted directed network is shown in FIG. 3, which illustrates, for example, v₁Is a fire oil tank.

And 6, in the established weighted directed network G, carrying out structure growth and numerical update on the weighted directed network G based on the fire occurrence process and the node state change monitored by the sensor network.

In the foregoing step 1, V = V₁,v₂,…,v_mThe empty sets are different, called node sets, and one tank in the tank field can be regarded as a node in G.

In the foregoing step 3, E = E_ij| i is less than or equal to m, j is less than or equal to m, i ≠ j is a subset of V × V, the subset is called a side set, the sides are directional, a side exists between an ignited oil tank and an adjacent unfired oil tank in the oil tank area, and the side can be regarded as a potential path for fire spread, namely the fire can possibly spread from the starting point V of the side_i(fired tank) edgewise end point v_j(unfired tank) spread.

In the foregoing step 4, W = W_ij| i is less than or equal to m, j is less than or equal to m, i ≠ j is a set of weight values of sides, certain side e_ijWeight value w of_ijIndicating a fire from a starting point v_iTo the end point v_jDegree of danger of spreading, w_ij∈[0,1]. When edge e_ijAfter the represented fire spread does occur, the model considers that this edge still exists, with a weight w_ij=1。w_ijThe value range is shown as formula (1):

in the step 2, the nodes v in the weighted directed network of the oil tank area form an attribute set a, and the elements in the attribute set a reflect the state information of the oil tank. The values of some elements in the attribute set may be the same or different between different nodes. Through the attribute set A, different oil tank nodes can be effectively distinguished, and information gathering and calculation can be carried out more conveniently

As previously described, the attribute set A is represented as:

A=｛p,d,o,s,q｝，

wherein: p is the node position, d is the node diameter, o is the storage oil property, s is the node state, and q is the thermal radiation intensity received by the node center.

In the present embodiment, the tank node position p represents the geographical position of the tank node in the tank farm. And information such as the distance between the oil tank nodes can be further obtained through the positions of different oil tank nodes.

The cross section of the oil tank node is equivalent to a circle, and the diameter d of the oil tank can directly represent the size of a certain oil tank node.

The stored oil material property o comprises a series of physical and chemical properties of the stored oil material, and the chemical and physical properties of different oil materials are different, so that the stored oil material is more obviously represented in the combustion process, and the influence of the property difference including density, volatility and combustion heat on the fire spreading process needs to be distinguished.

At the initial moment, the states of the oil tanks are safe; when a fire or other abnormal condition occurs, the oil tank status can be classified as: safety, danger, fire, the degree of danger rises in proper order. The state of the oil tank can be changed along with the change of the environment and the fire. For example, a tank that is safe in a tank burning state may be transformed into a dangerous state during the spread of fire, and further, may be burned to be transformed into a burning state.

The heat radiation of the flame is the most main way for transferring energy outwards after the oil tank is ignited, and the influence of the heat radiation on objects is closely related to factors such as the oil tank oil burning rate, distance, wind speed and wind direction and the like.

As an alternative embodiment, the aforementioned established weighted directed network is not static, and its structure and various attribute values will change with the fire occurrence progress and the change of the monitoring result of the sensor network, and its evolution mainly shows in both the structure growth and the value update.

Therefore, in the aforementioned step 6, in the established weighted directed network G, based on the occurrence progress of the fire and the node state change monitored by the sensor network, the weighted directed network G is subjected to structure growth and numerical value update.

Step 6-1, structure growth

When there are one or more nodes v at a time_iWhen the combustion state of the tank in the attribute changes and the degree of danger represented by the tank changes, a plurality of new adjacent nodes v are added into the weighted directed network_jAs a new potential fire spread node, the direction of the newly introduced edge is from v_iDirection v_jAnd determining the edge weight value is related to the intensity of the heat radiation received by the node center.

New node v_jFrom node v with a change in state_iOf the neighboring node.

Node v_jThe nodes to be connected in the tank farm fire-spreading network satisfy two conditions:

a) is v_jThe neighbor node of (2);

b) have been added to the fire-spreading network of the tank farm.

Step 6-2, updating numerical value

a) When the node state changes, the attribute values of the nodes in the weighted directed network, the weight values of the edges and other output results are re-determined;

b) when the structure is increased, re-determining attribute values of the relevant nodes in the weighted directed network, the weights of the relevant edges and other output results;

c) and when the data collected by the sensor network is updated, re-determining the attribute values of the relative nodes in the weighted directed network, the weight values of the edges and other results.

As an alternative embodiment, in step 2, the intensity of the heat radiation received by the node center is calculated according to the following formula (3):

q＝τFE_act(2）

wherein q represents the intensity of heat radiation at x meters from the center of the tank on fire, F represents the view angle factor, E_actRepresents the actual radiation energy of the flame surface, and τ represents the atmospheric transfer coefficient, where:

a) the atmospheric transfer coefficient τ is calculated according to the following formula (3):

τ＝1-0.058lnx （3）

in the formula: x represents the distance from the center of the flame to the target;

b) coefficient of viewing angle F

The viewing angle coefficient F is the ratio of the amount of radiation received by the radiation receiving surface from the radiation surface to the total amount of radiation, and its value is related only to the geometrical properties of the radiation surface and the receiving surface.

The calculation of the view angle coefficient is closely related to the flame shape, in this embodiment, the flame shape is equivalent to a cylinder, the bottom surface is the same as the cross section of the oil tank, the equivalent flame side surface is shown in fig. 4, and the flame height is H_fThe flame inclination angle is θ.

Preferably, the flame height H in this embodiment is based on the Thomas model proposed by Thomas_fCalculated according to the following formulas (4) and (5):

$H_f=55D{\left(\frac{m_f^{\&prime;}}{{\mathrm{\&rho;}}_0\sqrt{\mathrm{gD}}}\right)}^{0.67}{\left(\frac{U_W}{U_C}\right)}^{-0.21}---\left(4\right)$

$U_c={\left(\frac{g{m}_f^{\&prime;}D}{{\mathrm{\&rho;}}_0}\right)}^{\frac{1}{3}}---\left(5\right)$

in the formula, ρ₀Is air density, U_WRepresenting wind speed, D representing tank diameter, m_fRepresenting the fuel combustion rate, g representing the acceleration of gravity, g being 9.81m/s²。

The viewing angle factor F is calculated according to the following formula (6):

$F=\sqrt{F_h^2+F_v^2}---\left(6\right)$

in the windless state, F in the above formula (6)_h、F_vThe following calculation was performed:

$\begin{array}{c}{F}_h=\frac{B-1}{\mathrm{\&pi;S}\sqrt{B^2-1}}{\tan }^{-1}\sqrt{\frac{(B+1)(S-1)}{(B-1)(S+1)}}-C\frac{A-1}{\sqrt{A^2-1}}\\ F_v=\frac{1}{\mathrm{\&pi;S}}{\tan }^{-1}\left(\frac{h}{\sqrt{S^2-1}}\right)-\frac{h}{\mathrm{\&pi;S}}{\tan }^{-1}\sqrt{\frac{S-1}{S+1}}+C\frac{\mathrm{Ah}}{\sqrt{A^2-1}}\\ A=\frac{h^2+S^2+1}{2S}\\ B=\frac{1+S^2}{2S}\\ C=\frac{1}{\mathrm{\&pi;S}}{\tan }^{-1}\sqrt{\frac{(A+1)(S-1)}{(A-1)(S+1)}}\\ S=\frac{2R}{D}\\ h=\frac{2H_f}{D}\end{array}---\left(7\right)$

in windy conditions, F in the above formula (6)_h、F_vThe following calculation was performed:

$\begin{array}{c}{F}_h={\tan }^{-1}\frac{\sqrt{\frac{b+1}{b-1}}}{\mathrm{\&pi;}\sqrt{B^2-1}}-G\&CenterDot;D+\frac{\sin \mathrm{\&theta;}}{\sqrt{C}}E+F\\ F_v=I\&CenterDot;H\&CenterDot;D+\frac{\cos \mathrm{\&theta;}}{\sqrt{C}}(E+F)-I\&CenterDot;{\tan }^{-1}\sqrt{\frac{b-1}{b+1}}\\ A=a^2+{(b+1)}^2-2a(b+1)\sin \mathrm{\&theta;}\\ B=a^2+{(b-1)}^2-2a(b-1)\sin \mathrm{\&theta;}\\ C=1+(b^2-1)\\ D={\tan }^{-1}\sqrt{\frac{A(b-1)}{B(b+1)}}\\ E={\tan }^{-1}\frac{\mathrm{ab}-(b^2-1)\sin \mathrm{\&theta;}}{\sqrt{C(b^2-1)}}\\ F={\tan }^{-1}\frac{(b^2-1)\sin \mathrm{\&theta;}}{\sqrt{C(b^2-1)}}\\ G=\frac{a^2{(b+1)}^2-2(b+1)+\mathrm{ab}\sin \mathrm{\&theta;}}{\sqrt{\mathrm{AB}}}\\ H=\frac{a^2+{(b+1)}^2-2b(1+a\sin \mathrm{\&theta;})}{\sqrt{\mathrm{AB}}}\\ I=\frac{a\cos \mathrm{\&theta;}}{b-a\sin }\\ a=\frac{2H_f}{D}\\ b=\frac{2R}{D}\end{array}---\left(8\right)$

c) actual intensity of heat radiation E on flame surface_act

Actual intensity of heat radiation E on flame surface_actCalculated according to the following equation (9):

$E_{\mathrm{act}}=\frac{m^{\&prime;\&prime;}{H}_c\mathrm{\&eta;}}{0.25\mathrm{\&pi;}{D}^2+\mathrm{\&pi;D}{H}_f},---\left(9\right)$

wherein m' represents a fuel combustion rate, H_cRepresenting the combustion heat of oil products, η representing the ratio of radiant heat to total combustion heat energy, wherein the value is between 0.2 and 0.5, and D represents the diameter of an oil pool;

wherein the fuel combustion rate m "is calculated in accordance with the following formula (10):

m″＝m″∞(1-e-αD)， (10)

in the formula, m ″)_∞Indicating the burning rate at infinite tank diameter, α is an oil-related factor, for example, α is 2.1m when the stored fuel is gasoline^-1When the storage fuel is kerosene, α is 3.5m^-1For practical use, reference is made to the literature (Zhujiahua, Chenjunge, oil tank district pool evaluation of thermal radiation hazard [ J)]Journal of the research institute of water-transport science, 1999,1: 003.) or experimental assays.

In the weighted directed network G of the tank field, each edge is connected with each tank node, so that the tank field is transformed into a whole with close connection from the state that each tank node is respectively isolated. The weight of the edge plays a key role in the evolution of the weighted directed network.

The definition of the weighted median edge weight of the oil tank area weighted directed network is as follows: w represents the degree of risk of fire spreading from the start point vi of the edge to the end point vj of the edge. The weight of the edge and the state of the starting point and the ending point of the edge are all related: first, the starting point of the edge must be a tank node whose combustion state is on fire; secondly, the end points of the edges are non-fired tank nodes with a dangerous or safe fire status.

In the present embodiment, flame heat radiation is used as a main factor causing the spread of fire. The aforementioned edge weight value w_ijThe value form definition of (a) is determined according to the following formulas (11) and (12):

$Q_{\mathrm{ij}}={\&Integral;}_{\max (t_0,t_f-\mathrm{\&Delta;t})}^{\mathrm{tf}}{q}_{\mathrm{ij}}\mathrm{dt}---\left(11\right)$

$w_{\mathrm{ij}}=\left\{\begin{array}{c}1-e^{-k{Q}_{\mathrm{ij}}}\\ 1\end{array}\right.---\left(12\right)$

wherein, t₀Denotes the starting time, t_fRepresenting the current time, at represents a time constant, q_ijRepresenting a node v_jReal-time receiving node v_iIntensity of released thermal radiation: if Δ t>t_f—t₀Then Q is_ijIndicating the time t from the start₀To the current time t_fWithin a time period, node v_jCumulative acceptance v_iA value of an amount of emitted heat radiation; if Δ t<t_f—t₀Then Q is_ijIndicating the time t from_f- Δ t to the present time t_fWithin a time period, node v_jCumulative acceptance v_iA value of the amount of heat radiation is emitted. Thereby, Q_ijAt most the cumulative thermal radiation value over the most recent at time will be calculated. k is a constant coefficient.

In this embodiment, the constant k in the formula (12) is obtained by the following method:

thermal radiation damage judgmentThe calibration criteria reflect the damage to the equipment caused by different levels of thermal radiation intensity and exposure time, as shown in table 1 below. According to the thermal radiation damage judgment criteria and the experimental conditions of the invention shown in the table 1, the cumulative thermal radiation value received when the oil tank node receives the thermal radiation is equivalent to the thermal radiation intensity of 25.0kw/m²W =0.8 for a thermal radiation magnitude of 30s duration. This yields a constant k =0.0032, and the relationship between the edge weight w and the heat radiation amount Q is shown in fig. 5 according to equation (12).

In the actual oil tank area environment, the value of k can be properly modified according to the difference between equipment materials and the environment, so that the model has the best effect in actual use.

TABLE 1 hazards caused by different degrees of thermal radiation intensity

And combining the oil tank area weighted directed network established in the steps to obtain the real-time evolution state of the oil tank area fire spread network in the fire process. And obtaining the fire spreading trend according to the state change of the oil tank nodes in the fire spreading network of the oil tank area, the connection direction of edges among the oil tank nodes and the weight of the edges.

Referring to fig. 2 and fig. 3, in this embodiment, the foregoing fire spreading prediction specifically refers to using the established model, and then inputting the actual fire occurrence situation of one or more oil tanks and the environmental data monitored by the sensor network lock into the established oil tank area fire spreading model, that is, the weighted directed network, to obtain the directional edge pointed by the maximum weight value among the edge weight values of the directional edges, that is, the directional edge represents the trend direction of fire spreading.

The prediction of the fire spread tendency in the tank farm will now be described in detail with reference to fig. 6-7.

(1) Simulation fire spreading actual condition analysis

The simulation situation of the fire spreading in the oil tank area is shown in fig. 6, the wind direction of the experimental environment is from top to bottom in the graph, and the fire spreading state of the oil tank area at three time nodes in the experimental process is selected in fig. 6.

Initial moment, ignition tank node is v₂(ii) a At 100 seconds, the fired tank node is v₂、v₅Showing that the fire disaster starts from v within the time period of 0-100 seconds₂Spread to v₅(ii) a At 200 seconds, the fired tank node is v₂、v₅、v₈This means that the fire is from v within a period of 100 to 200 seconds₅Spread to v₈. The fire spreading direction of the oil tank area is consistent with the environmental wind direction.

(2) Model prediction result analysis based on the embodiment

And inputting the data acquired and stored in real time in the simulation process into the tank farm fire spread model to obtain a model result, wherein as shown in fig. 7, the output results of the tank farm fire spread model at three different time nodes are intercepted in fig. 7.

At 20 seconds, the fired tank node v₂Connected to the other 5 nodes in the tank farm, where v₂And v₅The edge weight between is maximal, which indicates that v is a period of time thereafter₅The possibility of fire occurrence is the greatest, which coincides with the spreading result at 100 seconds in fig. 6. At 120 seconds, the tank fire has spread to v₅Removing the edge which has already spread the fire, wherein the starting point of the edge with the maximum weight value in the weighted directed network of the oil tank area is v₅End point is v₈This means that the direction of fire spread most likely to occur thereafter is from v₅To v₈The state of the tank farm at 200 seconds in fig. 6 verifies this prediction. By comparing the weight changes of the rest edges at three moments of 20 seconds, 120 seconds and 220 seconds, the edge (such as e) with the tank node on the two sides as the starting point and the tank nodes on the two sides as the end points can be seen₂₁、e₂₄Etc.) tend to stabilize over time, indicating that the likelihood of fire spread does not remain constant over time, as in the previous experiment shown in fig. 6The results also verify this.

Therefore, in the oil tank area weighted directed network, the direction of the side with the larger weight is more consistent with the wind direction and almost the same as the fire spreading direction.

The weight of part of edges tends to be stable after a period of time, and the possibility of fire spreading is relatively low under the condition that no more heat radiation nodes are added.

Referring to fig. 2 and 3, in the embodiment, the correlation analysis includes analyzing and calculating a fire spreading reference time, that is, simulation data obtained by performing a fire simulation in the tank farm using the weighted directed network, and constructing a least squares support vector machine model based on the simulation data, so as to calculate the fire spreading reference time.

In the embodiment, a least square support vector machine model is constructed by using experimental data obtained by a fire simulation experiment of the oil tank field, and the model is used for calculating fire spreading reference time and providing help for fire fighting work of the oil tank field.

Specifically, as a preferred embodiment, the determination of the fire spread reference time is performed according to the following steps:

taking the environmental parameters collected by the sensor network as input data x = { x = ×)_kI k =1,2,3, … N }, input data x_kThe method comprises the steps of periodically acquiring temperature, humidity, carbon monoxide concentration and carbon dioxide concentration in a simulation process;

the actual propagation time is taken as an output value y = { y = {_kI k =1,2,3 …, N }, and output data y_kSubtracting the acquired data x from the actual fire spread moment measured by fire simulation_kThe time of (1) is obtained;

thus, a training set D of N samples is constructed: let D { (x)_k，y_k) 1,2,3, … N, where the input data is x_k∈R_nThe output data is y_k∈ R, training data, namely the input data x_kAnd output data y_kInputting least squaresA support vector machine model, wherein the aim is to obtain a support vector machine model function, namely an objective discriminant function;

and finally, inputting the actually acquired fire state data into the model, and calculating to obtain the reference time for fire spreading of a certain oil tank in the actual fire occurrence process of the oil tank area.

As an exemplary approach, the aforementioned least squares support vector machine model is described as follows:

consider a training set D of N samples: let D { (x)_k，y_k) 1,2,3, … N, where the input data is x_k∈RⁿThe output data is y_k∈ R. the goal of the support vector machine model is to construct a discriminant function in the following format.

Finding the optimal parameters w, b so that the function value y corresponding to the sample x can be approximated by f (x). Here, the non-linear mappingThe input data is mapped to a high-dimensional feature space. Thus, the least squares support vector machine algorithm can be transformed to solve the following optimization problem:

$\underset{w,s}{\min }J(w,e)=\frac{1}{2}{w}^{T}w+\frac{1}{2}\mathrm{\&gamma;}{\mathrm{\&Sigma;}}_{k=1}^N{e}_k^2---\left(14\right)$

e_kdenotes the error term, k =1,2, …, N.

Satisfying the equality constraint:

whereinIs from the input space RⁿTo a feature space R^mThe mapping of (a) to (b) is,w∈R^m。

the first term on the right side of the formula (14) reflects the confidence range, the more complex the model structure is, the larger the term is, the second term reflects the training error, and the front term and the rear term reflect the principle of minimizing the structural risk. Gamma is a penalty factor, the function of the penalty factor is to control the degree of penalty to the error division sample, the degree of importance of the model to the error division is reflected, and the larger the penalty is, the heavier the penalty is.

The above problem can be converted into a problem of solving a system of linear equations. The present invention solves this problem using the lagrangian method, as shown in equation (16).

Wherein, a_kAnd k is 1,2, … N is a lagrange multiplier.

According to the optimization conditions:

$\frac{\&PartialD;L}{w}=0,\frac{\&PartialD;L}{b}=0,\frac{\&PartialD;L}{e}-0,\frac{\&PartialD;L}{a}=0,---\left(17\right)$

it is possible to obtain,

defining a kernel functionThe kernel function K (xi, xj) is a symmetric function meeting the Mercer condition, and the invention selects the RBF kernel function: $K(x_i,x_j)=\exp (-{\left|\right|x_i-x_j\left|\right|}^2/{\mathrm{\&sigma;}}^2)---\left(20\right)$

wherein, $\left|\right|x_i-x_j\left|\right|=\sqrt{{\mathrm{\&Sigma;}}_{k=1}^N{(x_i^k-x_j^k)}^2},$ σ is the nucleus width.

According to the equations (18) and (19), the optimization problem is converted into the solution of a linear equation:

and finally obtaining a nonlinear model:

$f\left(x\right)=\underset{k=1}{\overset{N}{\mathrm{\&Sigma;}}}{a}_{k}K(x,x_k)+b---\left(22\right)$

and f (x) is the output quantity, namely the fire spread reference time.

As shown in the flow of establishing the least squares support vector machine model in fig. 8, the calculation of the fire spread reference time in this embodiment includes the following processes:

inputting training data (namely the input data x and the output data y) into a least square support vector machine model, obtaining model parameters meeting requirements by using formulas (16-21), substituting the model parameters into the least square support vector machine model (namely formula (22)), and finally calculating the reference time for fire spreading of a certain oil tank in the actual fire occurrence process of the oil tank area based on the actually acquired fire state data and the input model.

Although the present invention has been described with reference to the preferred embodiments, it is not intended to be limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, the protection scope of the present invention should be determined by the appended claims.

Claims (2)

1. A method for predicting and analyzing fire spread of an oil tank region based on a complex network is characterized by comprising the following steps:

sensor nodes are arranged at the positions of the oil tanks in the oil tank area and the middle positions of the adjacent oil tanks, and the nodes are connected to form a sensor network;

establishing a weighted directed network of tank farm fire spread within a computer system; and

based on the established weighted directed network, fire spreading prediction and correlation analysis are carried out by utilizing the occurrence state of the fire and the environmental data monitored by the sensor nodes;

the environmental information collected by the sensor node at least comprises: temperature, humidity, carbon monoxide concentration, carbon dioxide concentration, combustible gas concentration, wind speed, and wind direction;

the establishment of the weighted directed network comprises the following steps:

step 1, abstracting each oil tank of the oil tank area as a network node to form an oil tank set V (V ═ V-₁,v₂,…,v_m﹜；

And 2, giving each network node attribute for reflecting the state information of each oil tank to form an attribute set A:

A＝{p,d,o,s,q}，

wherein: p is the node position, d is the node diameter, o is the oil storage property, s is the node state, and q is the thermal radiation intensity received by the node center;

step 3, abstracting a fire spreading path into directed edges to form an edge set E:

E＝﹛e_ij︱i≤m,j≤m,i≠j﹜，

wherein the set E is a subset of V × V, and the edge E_ijIndicating a directed edge between a tank that has fired and an adjacent unlit tank as a potential fire propagation path, i.e. a starting point v from which a fire may start_iTowards the end point v of the edge_jSpreading;

step 4, endowing the edge weight value to form a weight value set W of the edge:

W＝﹛w_ij︱i≤m,j≤m,i≠j﹜，

wherein a certain edge e_ijWeight value w of_ijIndicating a fire from a starting point v_iTo the end point v_jDegree of danger of spreading, w_ij∈[0,1]When an edge e_ijAfter the indicated fire spread does occur, w_ij1, namely:

step 5, establishing a fire spreading weighted directed network G of the oil tank area, wherein G is a triple < V, E, W >;

and 6, in the established weighted directed network G, based on the fire occurrence process and the node state change monitored by the sensor network, carrying out structure growth and numerical update on the weighted directed network G:

step 6-1, structure growth

When there are one or more nodes v at a time_iWhen the combustion state of the tank in the attribute changes and the degree of danger represented by the tank changes, a plurality of new adjacent nodes v are added into the weighted directed network_jAs a new potential fire spread node, the direction of the newly introduced edge is from v_iDirection v_jDetermining the edge weight value is related to the intensity of the heat radiation received by the node center;

wherein: the new node v_jFrom node v with a change in state_iA set of neighbor nodes of (1);

step 6-2, updating numerical value

a) When the node state changes, the attribute values of the nodes in the weighted directed network, the weight values of the edges and other output results are re-determined;

b) when the structure is increased, re-determining attribute values of the relevant nodes in the weighted directed network, the weights of the relevant edges and other output results;

c) and when the data collected by the sensor network is updated, re-determining the attribute values of the relative nodes in the weighted directed network, the weight values of the edges and other results.

2. The complex network based tank farm fire spread prediction and analysis method of claim 1, wherein the fire spread prediction comprises the following processes: and based on the fire occurrence state of a certain oil tank in the oil tank area and the environmental data monitored by the sensor node, carrying out fire spread prediction according to the established weighted directed network, and solving the directional edge pointed by the maximum weight in the edge weights of all the directional edges, namely the trend direction of fire spread.