Building fire early warning management and control method and system based on Internet of Things platform

Abstract

According to the building fire early warning management and control method and system based on the Internet of Things platform disclosed by the invention, a hydrodynamic parameter field related to a building fire is converted into a knowledge ontology code of a social physical information system, and then a multi-source fusion model is constructed to generate smoke fusion state data; based on the two-dimensional field, carbon dioxide concentration and temperature parameters on the escape path are extracted, the personnel escape speed is calculated according to a preset personnel escape speed calculation model, and an escape decision model is constructed; and finally, dividing grids according to the feature size of the current building component, expanding the two-dimensional field into a square matrix, predicting each moment parameter field of the square matrix by using a trained decoder and encoder model, outputting decision information of each stage of the fire based on an escape decision model and each moment parameter field, and accessing an Internet of Things platform for real-time management and control. Therefore, scientific and reasonable escape suggestions can be provided for trapped people in real time according to the smoke diffusion state, and the current building fire early warning control effect is improved.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to a method and system for building fire warning and control based on an Internet of Things platform. Background Art

[0002] The main harm of building fires lies in toxic smoke. First, the toxic gases and particulate matter in the smoke can directly irritate the respiratory tract, causing symptoms such as coughing, wheezing, and difficulty breathing; the high-concentration smoke itself will also impede normal human breathing, increasing the risk of asphyxiation. Second, the thick smoke generated by the fire will reduce visibility, restricting escape and rescue operations; harmful substances in the smoke, such as acidic gases and fine particles, will directly damage the eyes, further increasing the risk of getting lost and stampede accidents. Then, the thick fire smoke will bring a strong visual impact and psychological pressure to people, causing emotions such as fear, anxiety, and despair, which will in turn affect people's decision-making ability and action efficiency, resulting in some people being unable to take the correct escape methods in time or delaying the escape opportunity. However, the existing prediction models have poor accuracy, are unable to accurately fuse multi-source data for prediction, and are difficult to generate reliable smoke fusion state data, resulting in inaccurate judgment of the fire development trend and unable to provide scientific and reasonable escape suggestions for trapped people based on the smoke diffusion state, thus making the current building fire warning and control effect poor, seriously threatening the safety of people's lives and property. Summary of the Invention

[0003] In view of the deficiencies in the prior art, the present invention provides a method for building fire warning and control based on an Internet of Things platform, including the following steps:

[0004] S1, converting the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of a social physical information system, where the parameter field of fluid mechanics includes a velocity field, a temperature field, and a carbon dioxide concentration field;

[0005] S2, constructing a multi-source fusion model X = F[C1, C2, C3], where F is an information fusion device, and generating smoke fusion state data by using Kalman filter to fuse the characterized data value C1, the data-driven prediction value C2, and the theoretical estimated value C3; where the data-driven prediction value C2 of the smoke state at a future moment is predicted by using a smoke prediction neural network trained with a smoke diffusion data set obtained through pre-experiments;

[0006] S3, decomposing the three-dimensional smoke parameter field into three orthogonal two-dimensional fields by projection on the horizontal and vertical planes, extracting the carbon dioxide concentration and temperature parameters on the escape path based on the two-dimensional fields, calculating the personnel escape speed according to a preset personnel escape speed calculation model, and constructing an escape decision model, where the escape decision model is used to determine a suitable current escape decision according to the personnel escape speed and the smoke diffusion speed;

[0007] S4. Divide the grid according to the characteristic dimensions of the current building component, expand the two-dimensional field into a square matrix, use the trained decoder and encoder models to predict the parameter fields at each moment of the square matrix, output the decision-making information for each stage of the fire based on the escape decision-making model and the parameter fields at each moment, and connect to the Internet of Things platform for real-time control.

[0008] Preferably, the step S3 specifically includes: calculating the personnel escape speed v according to a preset personnel escape speed calculation model, and constructing an escape decision-making model by comparing it with the smoke diffusion speed v soot Compare If the Decision value is 0, generate information suggesting staying in place for rescue; if the Decision value is 1, generate information suggesting escape.

[0009] Preferably, the step S2 specifically includes:

[0010] S21. Train a smoke prediction neural network NN through high-cost simulation pre-experiments, which can predict the smoke state at time t based on the smoke state at time t-1 Predict the smoke state C at time t 2|t :

[0011] S22. By calculating C at time t 3|t Obtain the output estimate of the fire smoke field parameters at the target time t:

[0012] Where the fusion coefficient K = P t|t-1 H T (HP t|t-1 H T +R) -1 , P t|t-1 Is the prediction covariance, R is the error covariance of the true state of the field parameters of C 3|t At time t, and H is the identity matrix.

[0013] Preferably, the step S1 specifically includes: converting the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding X of the social physical information system: = [P # E # S # A], where the parameter field of fluid mechanics includes the velocity field, temperature field, and carbon dioxide concentration field; where X is the knowledge ontology encoding of the parameter in the digital twin platform, P is the physical attribute encoding, E is the engineering attribute encoding, S is the social attribute encoding, A is the cognitive attribute encoding, the symbol: = is the definition formula for projecting the parameter onto the social physical information system, and the symbol # refers to the generalized interaction to reflect the weaving effect between the fields of P, E, S, and A.

[0014] Preferably, the step S4 specifically includes: dividing the grid according to the characteristic dimensions of the current building component, and expanding the two-dimensional smoke field into a square matrix XN*N , where the grid values at non-spatial entities of building components are set to zero; the fire process is decomposed into a multi-stage time series X N*N (ti), using the trained encoder model to extract the principal component features and the decoder model to restore the details; using the transfer matrix to replace the traditional similarity criterion, and quickly restoring the parameter field of smoke at each moment under simplified physical conditions through maximum likelihood estimation.

[0015] Preferably, the generation process of the transfer matrix includes: encoding and dimension reduction of the fire time series image through a convolutional neural network, and supplementing the detailed features of the flow field during decoding to achieve rapid restoration of the parameter field.

[0016] Preferably, it also includes tightening the social attribute S, converting the carbon dioxide concentration field into a decision-making threshold for personnel evacuation, and correcting the calculation model of personnel evacuation speed according to the preset correspondence table of the harm relationship between carbon dioxide volume fraction and exposure time.

[0017] The present invention also discloses a building fire warning and control system based on the Internet of Things platform, including:

[0018] A conversion module for converting the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of the social physical information system, where the parameter field of fluid mechanics includes a velocity field, a temperature field, and a carbon dioxide concentration field;

[0019] A fusion module for constructing a multi-source fusion model X = F[C1, C2, C3], where F is an information fusion device, and generating smoke fusion state data by using Kalman filter to fuse the representation data value C1, the data-driven prediction value C2, and the theoretical estimation value C3; where the data-driven prediction value C2 of the smoke prediction neural network trained by using the smoke diffusion data set obtained through pre-experiment is used to predict the smoke state at future moments;

[0020] An evacuation decision module for decomposing the three-dimensional smoke parameter field into three orthogonal two-dimensional fields by horizontal and vertical plane projections, extracting the carbon dioxide concentration and temperature parameters on the evacuation path based on the two-dimensional fields, calculating the personnel evacuation speed according to the preset calculation model of personnel evacuation speed, and constructing an evacuation decision model, where the evacuation decision model is used to determine the suitable current evacuation decision according to the personnel evacuation speed and the smoke diffusion speed;

[0021] A control implementation module for dividing the grid according to the characteristic size of the current building component, expanding the two-dimensional field into a square matrix, predicting the parameter field of each moment of the square matrix by using the trained decoder and encoder models, outputting the decision information of each stage of the fire based on the evacuation decision model and the parameter field of each moment, and accessing the Internet of Things platform for real-time control.

[0022] The present invention also discloses a building fire warning and control device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of any of the foregoing methods are implemented.

[0023] The present invention also discloses a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of any of the foregoing methods are implemented.

[0024] The building fire warning and control method and system based on the Internet of Things platform disclosed by the present invention convert the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of the social physical information system. The parameter field of fluid mechanics includes a velocity field, a temperature field, and a carbon dioxide concentration field. Then, a multi-source fusion model is constructed to generate smoke fusion state data. By decomposing the three-dimensional smoke parameter field into three orthogonal two-dimensional fields according to horizontal and vertical projections, the carbon dioxide concentration and temperature parameters on the escape path are extracted based on the two-dimensional fields. The escape speed of personnel is calculated according to a preset personnel escape speed calculation model, and an escape decision model is constructed. Finally, grids are divided according to the characteristic dimensions of the current building components, the two-dimensional field is extended into a square matrix, and the decoder and encoder models trained are used to predict the parameter fields at each moment of the square matrix. Based on the escape decision model and the parameter fields at each moment, decision information at each stage of the fire is output and connected to the Internet of Things platform for real-time control. Thus, scientific and reasonable escape suggestions can be provided to trapped personnel in real time according to the smoke diffusion state, and the current building fire warning and control effect can be improved.

[0025] The additional aspects and advantages of the present invention will be partially given in the following description, partially become obvious from the following description, or be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The drawings described herein are used to provide a further understanding of the present invention, and constitute a part of this application. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0027] Figure 1 is a flowchart showing the process of the building fire warning and control method based on the Internet of Things platform disclosed in an embodiment of the present invention.

[0028] Figure 2 is a schematic structural diagram of the building fire warning and control system based on the Internet of Things platform disclosed in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the described embodiments of the present invention without creative efforts shall fall within the scope of protection of the present invention.

[0030] In the present invention, unless otherwise clearly defined and limited, the terms "installation", "connection", "connection", "fixation" and other terms should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0031] In the present invention, unless otherwise clearly defined and limited, the first feature being "above" or "below" the second feature may include the direct contact between the first and second features, or may include the situation where the first and second features are not in direct contact but in contact through other features therebetween. Moreover, the first feature being "above", "over" and "on" the second feature includes that the first feature is directly above and obliquely above the second feature, or merely means that the horizontal height of the first feature is higher than that of the second feature. The first feature being "under", "below" and "beneath" the second feature includes that the first feature is directly below and obliquely below the second feature, or merely means that the horizontal height of the first feature is lower than that of the second feature.

[0032] Unless otherwise defined, the technical terms or scientific terms used herein shall have the ordinary meanings understood by those of ordinary skill in the art to which the present invention pertains. The "first", "second" and similar terms used in the specification and claims of this patent application of the present invention do not denote any order, quantity or importance, but are only used to distinguish different components. Similarly, the terms such as "a" or "one" do not denote a quantity limitation, but mean that there is at least one.

[0033] The present invention discloses a method for building fire warning and control based on the Internet of Things platform, as shown in the attached Figure 1 figure, and specifically includes the following steps.

[0034] Step S1, converting the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of the social physical information system, where the parameter field of fluid mechanics includes a velocity field, a temperature field and a carbon dioxide concentration field.

[0035] In this embodiment, step S1 may specifically include: transforming the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding X := [P # E # S # A] of the social physical information system, where the parameter field of fluid mechanics includes the velocity field, temperature field, and carbon dioxide concentration field; where X is the knowledge ontology encoding of the parameter in the digital twin platform, P is the physical attribute encoding, E is the engineering attribute encoding, S is the social attribute encoding, A is the cognitive attribute encoding, the symbol := is the definition formula for the projection of the parameter onto the social physical information system, and the symbol # refers to the generalized interaction to reflect the weaving effect among the fields of P, E, S, and A.

[0036] In the prior art, it is difficult to assign values to the known domain P, that is, it is difficult to obtain the parameter field based on fluid mechanics. Therefore, if we want to keep the position of X unchanged in the overall knowledge structure, the relaxation of P can be achieved through the tight binding balance of E, S, and A. For E, the applicable range of the model for engineering objects can be reduced, making the damaged simplified theory no longer universal. For S, the social needs can be changed, such as degenerating from the requirement to solve the carbon dioxide concentration in a certain space to directly judging the possibility of casualties. For A, the way of risk perception can be corrected, and the balance can be adjusted between the rational perception and decision-making autonomous process that requires quantitative evidence and intuition. Through such a transformation at the domain level, that is, transforming the problem of solving the parameter field of fluid mechanics into the problem of parameter assignment in CPSS. The assignment method no longer seeks to optimize only the theory, model, or algorithm in the P domain; but mainly realizes it by tightly binding the domains E, S, and A of X.

[0037] Step S2, construct a multi-source fusion model X = F[C1, C2, C3], where F is the information fusion device, and generate the smoke fusion state data by using the Kalman filter to fuse the data value C1, the data-driven prediction value C2, and the theoretical estimation value C3; where the data-driven prediction value C2 of the smoke state at a future moment is predicted by using the smoke diffusion data set obtained through pre-experiments to train the smoke prediction neural network.

[0038] In this embodiment, step S2 may specifically include the following content.

[0039] Step S21, train a smoke prediction neural network NN through high-cost simulation pre-experiments, which can predict the smoke state at time t based on the smoke state at time t - 1 Predict the smoke state C at time t 2|t ,

[0040] Step S22, by calculating C at time t 3|t Obtain the output estimate of the fire smoke field parameters at the target time t:

[0041] Where the fusion coefficient K = Pt|t-1 H T (HP t|t-1 H T +R) -1 , P t|t-1 is the prediction covariance, R is C 3|t The error covariance of the true state of the field parameters at time t, H is the identity matrix.

[0042] In this embodiment, parameter X is any node between entities. When it is called by an algorithm model, it needs to be assigned a value to obtain the corresponding calculation result. Ideally, accurate measurement, reasoning about dynamic relationships under time-space slices, and equivalent calculations based on prior theories can obtain mutually consistent results. The estimation effect of these three relatively independent methods on X can be recorded as X=F[C1, C2, C3]. Among them, C1, C2, and C3 represent the estimation of parameter X in the three directions of representation, data, and theory, respectively, and F[] represents an information fusion device that fuses C1, C2, and C3 into an optimal output result. The significance of F[] lies in the fact that there are errors in C1, C2, and C3 in practice, which makes the three inconsistent with each other. For example, taking the classical state estimation theory of automation as an example, it is assumed that the theoretical model has no deviation, thereby simplifying the optimal estimation problem of X to an optimization problem of the covariance of additive Gaussian noise of C1 and C2. At this time, F[] degenerates into Kalman filtering. However, the design of F[] is not limited to one approach, and it is also possible to directly and adaptively adjust the fusion weights of C1, C2, and C3 using methods based on neural networks.

[0043] In this embodiment, C1 can only be used as a posteriori or local evaluation in the fusion sequence. C2 can obtain a large number of high-quality data set samples related to the smoke diffusion process and train the neural network through high-cost simulation pre-experiments, so that when the fire material actually exists, it can play a role in generating the real-time evolution of the parameter field according to the input conditions. In this embodiment, C1 and C2 will have an indirect auxiliary optimization effect on the final result, but cannot directly solve the expected goal of how to obtain the parameter field in real time. Therefore, a space must be found from C3 so that the elements in the space can be numerically equal to the three-dimensional tensor of the target physical field through transformation.

[0044] Therefore, in a preferred embodiment, the process of assigning parameters C1, C2, and C3 to X may include the following content.

[0045] Through expensive simulation experiments, a prediction neural network NN is trained to predict the smoke state at time t-1. Predict the smoke state C at time t 2|t ,

[0046] By calculating C at time t 3|tObtain the output estimate of the fire smoke field parameters at the target time t:

[0047]

[0048] Among them, H is the observation matrix. At this time, we assume that the observed value is the state variable itself, so H is the identity matrix.

[0049] Obtain the fusion coefficient K, K = P t|t-1 H T (HP t|t-1 H T +R) -1 , where R is the error covariance between C 3|t and the true state of the field parameters at time t.

[0050] Obtain the predicted covariance P t|t-1 , P t|t-1 = λ k (NN)P t-1 (NN) T +Q; among them, Q is the error covariance between C 2|t and the true state of the field parameters at time t. The error covariance P t-1 is obtained from the calculation formula of P t , but related parameters such as K are continuously updated with the changes of the state and environment at different times. The initial covariance P0 can be set in various ways, including setting based on prior knowledge or experience, using simple default values, and initializing with observed data, etc. In this embodiment, the method is not specifically limited and described in detail:

[0051] P t =(I - KH)P t|t-1 (I - KH) T +KRK T .

[0052] In this embodiment, it also includes the step of optimizing the time-varying fading factor λ k .

[0053] The role of the fading factor λ k is to timely sense that when the fire smoke develops or encounters certain mutations in the environment, various situations like this may cause NN to become inaccurate. At this time, it is necessary to intervene in the predicted covariance P t|t-1 , and reduce the influence of C2 on the estimation result. There are already various approximation methods for the initial value λ0 in the research of strong tracking filtering, and it is not specifically limited and described in detail in this embodiment. The optimization adjustment action a k for λ t adopts a deterministic strategy, that is, directly outputs the specific action in a given state, rather than the probability distribution of the action:

[0054] a t = argmax Q w (λ k , a). Where Q w (λ k , a) is the action-value function, representing the long-term expected reward for executing action a in state λ w . argmax means to find the a that maximizes the Q k value, and the result of the selection is a w . t .

[0055] In this embodiment, it also includes updating using the Q-learning value function. Executing the update action a k will obtain a reward r t . At the same time, the smoke also spreads to the t+1 moment, corresponding to the real state X t+1 . (λ t+1 , a k , r t , λ t+1 ) will form a set of experience tuples D. Sampling batch data from D to achieve the update of Q k+1 : w

[0056] Q w (λ k , a t ) ← Q w (λ k , a t ) + α[r t+1 + γ max Q w (λ k+1 , a′) - Q w (λ k , a t )].

[0057] Among them, α is the learning rate, which is obtained in advance through the experience of early algorithm debugging. γ is the discount factor, used to adjust the influence of the innovation on the Q w value. γ → 0 tends to consider mutations as a kind of measurement outliers, and γ → 1 tends to follow the small changes in the environment. r t+1 , as the reward function, needs to be designed specifically according to the problem, including but not limited to the following forms:

[0058] Define Then r t+1 = k1d t + k2(d t-1 - d t ); where k1 < 0 is the penalty coefficient, k2 > 0 is the reward coefficient, and the values are obtained by debugging according to the experience of the early algorithm. ​

[0059] In this embodiment, the problem of smoke propagation is simplified to a two-dimensional problem to reduce the computational cost in all aspects. The same applies to the three-dimensional tensor, except that the calculation steps are more complex. This part is to give a specific method and steps on how C1 and C2 optimize and enhance C3. However, even if there are no conditions to obtain C1 and C2, as long as the output of C3 is ensured, the basic effect can still be achieved.

[0060] Step S3: Decompose the three-dimensional smoke parameter field into three orthogonal two-dimensional fields by projecting onto the horizontal plane and the vertical plane. Based on the two-dimensional fields, extract the carbon dioxide concentration and temperature parameters on the escape route, calculate the personnel escape speed according to the preset personnel escape speed calculation model, and construct an escape decision model. The escape decision model is used to determine the appropriate escape decision for the current situation based on the personnel escape speed and the smoke diffusion speed.

[0061] Specifically, the personnel escape speed v can be calculated according to the preset personnel escape speed calculation model. By comparing it with the smoke diffusion speed v soot an escape decision model is constructed. If the Decision value is 0, information suggesting staying in place and waiting for rescue is generated. If the Decision value is 1, information suggesting escaping is generated.

[0062] In this embodiment, the tight binding of the social attribute S is to convert the carbon dioxide concentration field into a personnel escape decision threshold, and modify the personnel escape speed calculation model according to the preset correspondence table between the carbon dioxide volume fraction and the hazard relationship of the exposure time.

[0063] Specifically, the tight binding S can be achieved in the following way: The problem of a person running along an escape route in a certain posture (standing or crawling), which makes the solution of N 3 values of the smoke field X, can be simplified to the solution of N xy values of the two-dimensional fields X xz 、X yz 、X 2 respectively, which are the projections of the three-dimensional field parameters onto three orthogonal planes:

[0064] X→(X xy ,X xz ,X yz )

[0065] X xy =∑ z X,X xz =∑ y X,X yz =∑ x X

[0066] Assume that z is the direction of gravity and xy is the horizontal plane of the building. Objectively, the escape speed of a person is determined by the height of the escape line of sight, the CO2 concentration in the air, where in this embodiment, all harmful gases are converted into CO2 for simplicity, and the exposure time at temperatures above 333K. The impact of the CO2 gas concentration on the human body can be adjusted according to the corresponding table of the harm relationship between the volume fraction of carbon dioxide and the exposure time obtained in advance through experiments and other means.

[0067] Since the escape paths in the building are limited, at the selected initial escape moment t0, it is first necessary to determine whether there is an opportunity to escape. A boundary case is when the speed v at which the smoke spreads along the escape path soot is greater than the escape speed, then stay and wait for rescue on the spot; otherwise, choose to escape. The escape speed is related to the selected escape path L i . Extract an escape path visible layer (l, z l ) from X. Then the escape speed v(L i , t0) is determined by the average CO2 concentration l on z and the average temperature . The above escape decision-making model can be described as:

[0068]

[0069] Assume that the decomposition l of L i is always set to x. Then z l and its corresponding and other parameters can be obtained by processing the estimator X xz . That is, through the tight bundle S, the simplification problem of 3N 3 to 3N 2 is realized.

[0070] Step S4: Divide the grid according to the characteristic size of the current building component, expand the two-dimensional field into a square matrix, use the trained decoder and encoder models to predict the parameter fields at each moment of the square matrix, output the decision-making information at each stage of the fire based on the escape decision-making model and the parameter fields at each moment, and connect to the Internet of Things platform for real-time control.

[0071] In this embodiment, step S4 specifically includes: dividing the grid according to the characteristic size of the current building component, expanding the two-dimensional smoke field into a square matrix X N*N , where the grid values at non-spatial entities of the building component are set to zero. Decompose the fire process into a multi-stage time series X N*N(ti), the trained encoder model is used to extract the principal component features, and the decoder model restores the details; the transfer matrix is used to replace the traditional similarity criterion, and the parameter field of the smoke at each moment is quickly restored under the simplified physical conditions through maximum likelihood estimation. The generation process of the transfer matrix includes: encoding and dimension reduction of the fire time series images through a convolutional neural network, and supplementing the detailed features of the flow field during decoding to achieve the rapid restoration of the parameter field.

[0072] In this embodiment, the tight bundle E can be obtained in the following manner: according to engineering concepts such as the building fire space L i The decomposition l of must necessarily correspond to a basic spatial component. On its characteristic dimensions (x, z), the selection of the grid density is related to the morphology of the smoke diffusion region and the spatial gradients of CO2 concentration and temperature. Select a suitable grid shape and size (△x, △z), take the maximum value N = max(x / △x, z / △z) of the number of grid cells in the x and z directions, and expand the parameter field X xz into a square matrix X NXN . The grid at the non-spatial entity of the building component is constantly taken as 0.

[0073] In engineering, the square matrix X NXN is decomposed into a finite number of time series X NXN (ti) at certain time scales. Regarding X NXN (ti) as a series of images, the principle of image processing algorithms is adopted, and the generation of the field parameters is realized through two key steps of an encoder and a decoder. The role of the encoder is to achieve a processing effect similar to principal component analysis (PCA) by training a convolutional neural network (CNN). It is equivalent to reducing an arbitrary square matrix to an approximate diagonal matrix through PCA. The decoder is used to generate a new image, that is, the reproduction or restoration of the original image in this embodiment, by adding details and removing noise and other complex processes to the feature image to be processed, that is, the approximate diagonal matrix. The encoder and the decoder are repeatedly trained with real representative data to achieve approximate reversibility.

[0074] X NXN (ti) at different time slices ti, the approximate diagonal matrix obtained through the encoder has an eigenvalue vector Therefore, X NXN (t) can obtain the transfer matrix through the encoder and its post-processing Thus, the traditional similarity criterion with the equality of dimensionless numbers as the similarity invariant is replaced by the transfer matrix T as the similarity invariant to ensure transitivity.

[0075] Through the above steps, an algebraic-based equivalent structure is constructed. It only needs to find a maximum likelihood PL, that is, the generalized maximum likelihood estimation, that satisfies T under a relatively easy-to-implement physical condition to complete the relaxation of P in step S1.

[0076] By defining the prototype problem as P0, it is defaulted through the above steps that PL is a parallel physical generator that generates approximately linear features of P0. Its optimality can be continuously improved through the aforementioned process, and this optimization process can be foreseen to be related to the mathematical properties of the analytical solution space of P0. If the mathematical properties of the P0 solution space are poor, it will be observed that it is difficult to find a stable PL as the single likelihood of the transfer matrix T, and the likelihood is likely to be carried out separately for At this time, PL degrades into a group of PLs - PL(ti).

[0077] Through the above steps, when a building detects a fire or the fire source again. T can be quickly calculated on PL(ti), and based on this T, the encoder, and the necessary steps of its data preprocessing and postprocessing, the entire diffusion process C3 of the fire smoke can be restored. This avoids the high-cost repeated calculations on P0 and is not completely data-driven.

[0078] The method for building fire warning and control based on the Internet of Things platform disclosed in this embodiment transforms the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of the social physical information system. The parameter field of fluid mechanics includes the velocity field, temperature field, and carbon dioxide concentration field. Then, a multi-source fusion model is constructed to generate smoke fusion state data. By decomposing the three-dimensional smoke parameter field into three orthogonal two-dimensional fields by horizontal and vertical plane projections, the carbon dioxide concentration and temperature parameters on the escape path are extracted based on the two-dimensional fields, and the personnel escape speed is calculated according to the preset personnel escape speed calculation model, and an escape decision model is constructed; finally, the grid is divided according to the characteristic size of the current building component, the two-dimensional field is extended into a square matrix, and the parameter fields of each moment of the square matrix are predicted using the trained decoder and encoder models. Based on the escape decision model and the parameter fields of each moment, the decision information at each stage of the fire is output and connected to the Internet of Things platform for real-time control. Thus, it can provide scientific and reasonable escape suggestions for trapped personnel in real time according to the smoke diffusion state, and improve the current building fire warning and control effect.

[0079] In another embodiment, a building fire warning and control system based on the Internet of Things platform is also disclosed, as shown in the appendix Figure 2As shown in the figure, the building fire warning and control system includes a conversion module 1, a fusion module 2, an escape decision-making module 3, and a control implementation module 4. Among them, the conversion module 1 is used to convert the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of the social physical information system. The parameter field of fluid mechanics includes a velocity field, a temperature field, and a carbon dioxide concentration field. The fusion module 2 is used to construct a multi-source fusion model X = F[C1, C2, C3], where F is an information fusion device, and by adopting Kalman filter fusion to represent the data value C1, the data-driven prediction value C2, and the theoretical estimated value C3, to generate the smoke fusion state data; among them, the data-driven prediction value C2 of the smoke state at a future moment is predicted by using the smoke diffusion data set obtained through pre-experiments to train a smoke prediction neural network. The escape decision-making module 3 is used to decompose the three-dimensional smoke parameter field into three orthogonal two-dimensional fields by horizontal and vertical plane projections, extract the carbon dioxide concentration and temperature parameters on the escape path based on the two-dimensional fields, calculate the personnel escape speed according to the preset personnel escape speed calculation model, and construct an escape decision-making model. The escape decision-making model is used to determine the suitable current escape decision according to the personnel escape speed and the smoke diffusion speed. The control implementation module 4 is used to divide the grid according to the characteristic size of the current building component, expand the two-dimensional field into a square matrix, use the trained decoder and encoder models to predict the parameter fields at each moment of the square matrix, output the decision information at each stage of the fire based on the escape decision-making model and the parameter fields at each moment, and connect to the Internet of Things platform for real-time control.

[0080] In this embodiment, the fusion module 2 is specifically further configured to train a smoke prediction neural network NN through high-cost simulation pre-experiments, which can predict the smoke state at time t-1 and predict the smoke state C at time t 2|t : By calculating C at time t 3|t the output estimate of the parameters of the fire smoke field at the target time t is obtained: where the fusion coefficient K = P t|t-1 H T (HP t|t-1 H T +R) -1 , P t|t-1 is the prediction covariance, R is the error covariance of the true state of the field parameters of C 3|t at time t, and H is the identity matrix.

[0081] The specific functions of the above building fire warning and control system based on the Internet of Things platform correspond one by one to those of the building fire warning and control method based on the Internet of Things platform disclosed in the previous embodiments. Therefore, they will not be described in detail herein. For details, reference can be made to the respective embodiments of the building fire warning and control method based on the Internet of Things platform disclosed above. It should be noted that the various embodiments in this specification are described in a progressive manner, and the key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other.

[0082] In some other embodiments, a building fire warning and control device is also provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements each step of the building fire warning and control method based on the Internet of Things platform described in the above embodiments.

[0083] The building fire warning and control device may include, but is not limited to, a processor and a memory. The server may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the schematic diagram is only an example of the server and does not constitute a limitation on the server device. It may include more or fewer components than shown in the figure, or combine some components, or different components. For example, the server device may also include input / output devices, network access devices, buses, etc.

[0084] The so-called processor may be a central processing unit, or may also be other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or the processor may also be any conventional processor, etc. The processor is the control center of the server device, and connects various parts of the entire server device through various interfaces and lines. The memory can be used to store the computer program and / or modules. The processor realizes various functions of the server device by running or executing the computer program and / or modules stored in the memory, and by calling the data stored in the memory. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory.

[0085] When the building fire warning and control method based on the Internet of Things platform is implemented in the form of software functional units and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-mentioned embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention.

[0087] In summary, the above are only the preferred embodiments of the present invention, and all equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope covered by the patent of the present invention.

Claims

1. A building fire warning and control method based on the Internet of Things platform, characterized in that, It includes the following steps: S1. Convert the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of the social physical information system, where the parameter field of fluid mechanics includes the velocity field, temperature field, and carbon dioxide concentration field; S2. Construct a multi-source fusion model X = F[C1, C2, C3], where F is an information fusion device, and generate smoke fusion state data by using the Kalman filter to fuse the data value C1, data-driven predicted value C2, and theoretical estimated value C3; among them, the data-driven predicted value C2 of the smoke state at a future moment is predicted by using the smoke diffusion data set obtained through pre-experiments to train the smoke prediction neural network; S3. Decompose the three-dimensional smoke parameter field into three orthogonal two-dimensional fields by projecting on the horizontal and vertical planes, extract the carbon dioxide concentration and temperature parameters on the escape path based on the two-dimensional fields, calculate the personnel escape speed according to the preset personnel escape speed calculation model, and construct an escape decision-making model, where the escape decision-making model is used to determine the suitable current escape decision according to the personnel escape speed and smoke diffusion speed; S4. Divide the grid according to the characteristic size of the current building component, expand the two-dimensional field into a square matrix, use the trained decoder and encoder models to predict the parameter fields at each moment of the square matrix, output the decision-making information at each stage of the fire based on the escape decision-making model and the parameter fields at each moment, and connect to the Internet of Things platform for real-time control.

2. The method for building fire warning and control based on the Internet of Things platform according to claim 1, characterized in that, The specific content of step S3 includes: Calculate the personnel evacuation speed v according to the preset personnel evacuation speed calculation model, and construct an evacuation decision-making model by comparing it with the smoke diffusion speed v soot Compare to construct an evacuation decision-making model If the Decision value is 0, generate information suggesting staying in place for assistance. If the Decision value is 1, generate information suggesting evacuation.

3. The method for building fire warning and control based on the Internet of Things platform according to claim 2, wherein The specific content of step S2 includes: S21. Train a smoke prediction neural network NN through high-cost simulation pre-experiments, which can predict the data-driven prediction value C of the smoke state at time t based on the smoke state at time t-1. Predict the data-driven prediction value C of the smoke state at time t. 2|t , S22, obtaining the theoretical estimated value C of the smoke state at time t 3|t to get the output estimation of the fire smoke field parameters at the target time t: where the fusion coefficient K = P t|t-1 H T (HP t|t-1 H T +R) -1 , P t|t-1 is the prediction covariance, R is the error covariance of the true state of the field parameters of C 3|t at time t, and H is the identity matrix.

4. The method for building fire warning and control based on the Internet of Things platform according to claim 3, characterized in that The specific content of step S1 includes: Convert the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding X := [P#E#S#A] of the social physical information system, where the parameter field of fluid mechanics includes the velocity field, temperature field, and carbon dioxide concentration field; where X is the knowledge ontology encoding of the parameter in the digital twin platform, P is the physical attribute encoding, E is the engineering attribute encoding, S is the social attribute encoding, A is the cognitive attribute encoding, the symbol := is the definition formula for the projection of the parameter onto the social physical information system, and the symbol # refers to the generalized interaction to reflect the weaving effect among the fields of P, E, S, and A.

5. The method for building fire warning and control based on the Internet of Things platform according to claim 4, characterized in that, The specific content of step S4 includes: Divide the grid according to the characteristic dimensions of the current building component, and expand the two-dimensional smoke field into a square matrix X N*N , where the grid value at the non-spatial entity of the building component is set to zero; Decompose the fire process into a multi-stage time series X N*N (ti), use the trained encoder model to extract the principal component features, and the decoder model to restore the details; use the transfer matrix to replace the traditional similarity criterion, and quickly restore the parameter field of the smoke at each moment under simplified physical conditions through maximum likelihood estimation.

6. The method for building fire warning and control based on the Internet of Things platform according to claim 5, wherein, Among them, the generation process of the transfer matrix includes: encoding and dimension reduction of the fire time series image through a convolutional neural network, and supplementing the detailed features of the flow field during decoding to achieve the rapid restoration of the parameter field.

7. The method for building fire warning and control based on the Internet of Things platform according to claim 6, characterized in that: It also includes the tight binding of the social attribute S, converting the carbon dioxide concentration field into the personnel escape decision threshold, and correcting the personnel escape speed calculation model according to the preset corresponding table of the harm relationship between carbon dioxide volume fraction and exposure time.

8. An architectural fire warning and control system based on an Internet of Things platform, characterized in that, It includes: A conversion module for converting the parameter field of fluid mechanics related to building fires into the knowledge ontology encoding of the social physical information system, where the parameter field of fluid mechanics includes the velocity field, temperature field, and carbon dioxide concentration field; The fusion module is used to construct a multi-source fusion model X = F[C1, C2, C3], where F is an information fusion device, which generates smoke fusion state data by using Kalman filter to fuse the characterization data value C1, the data-driven prediction value C2 and the theoretical estimation value C3; wherein the data-driven prediction value C2 of the smoke state at a future moment is predicted by using a smoke prediction neural network trained with a smoke diffusion data set obtained through pre-experiments. The escape decision-making module is used to decompose the three-dimensional smoke parameter field into three orthogonal two-dimensional fields by horizontal and vertical plane projections, extract the carbon dioxide concentration and temperature parameters on the escape path based on the two-dimensional fields, calculate the personnel escape speed according to a preset personnel escape speed calculation model, and construct an escape decision-making model, and the escape decision-making model is used to determine a suitable current escape decision according to the personnel escape speed and the smoke diffusion speed. The control and implementation module is used to divide a grid according to the characteristic dimensions of the current building components, expand the two-dimensional field into a square matrix, predict the parameter fields of each moment of the square matrix by using the trained decoder and encoder models, output decision-making information for each stage of the fire based on the escape decision-making model and the parameter fields of each moment, and connect to the Internet of Things platform for real-time control.

9. An architectural fire warning and control device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1-7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, the steps of the method according to any one of claims 1-7 are implemented.

Images