An artificial intelligence system and a method for preparing a non-toxic potassium oxide fire extinguishing agent

Abstract

An artificial intelligence system and a method for preparing a non-toxic potassium oxide fire extinguishing agent belong to the field of artificial intelligence technology. In the system, a fire extinguishing model is configured to input the weights #imgabs0#, #imgabs1#, #imgabs2# and #imgabs3# of an oxidant, a reductant, a curing stabilizer and a binder into an input layer, and an output layer outputs a fire extinguishing volume v and a fire extinguishing time t; a calculation module is configured to generate a loss function #imgabs7# according to the fire extinguishing volume v, the fire extinguishing time t and the fire extinguishing target volume #imgabs5# and the fire extinguishing target time #imgabs6# of a finished fire extinguishing agent with a weight of #imgabs4#; a weight percentage adjustment module is configured to calculate the gradients according to #imgabs8#, and update #imgabs9#, #imgabs10#, #imgabs11# and #imgabs12# using the gradients: a judgment module is configured to judge whether #imgabs13# is the smallest, and the corresponding weight percentage when L1 is the smallest is the weight percentage of the pure raw material for preparing the fire extinguishing agent. The present invention does not require repeated experiments, saving costs and manpower.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence system and a method for preparing a non-toxic potassium oxide fire extinguishing agent. Background Art

[0002] For example, the Chinese invention patent application with publication number CN118436947A discloses a method for preparing a safe and non-toxic potassium oxide fire extinguishing agent, which is composed of the following raw materials in parts by weight: 40-45 parts of potassium nitrate, 30-35 parts of dicyandiamide, 20-25 parts of natural resin glue, and 0.8-1.2 parts of a multi-polymer.

[0003] However, the potassium oxide fire extinguishing agent generated in the prior art to construct the potassium oxide fire extinguishing agent has a numerical floating space for each weight percentage corresponding to each of its constituent materials. More importantly, the constituent materials are relatively complex, and constituent materials with different values ​​have different effects on the performance of the final product, which makes it difficult to obtain the best performance of the potassium oxide fire extinguishing agent prepared under the same preparation process. For example, potassium nitrate, dicyandiamide, resin glue and other materials are usually used to prepare potassium oxide fire extinguishing agents. These materials affect each other and the weight percentage of each material has a large numerical floating space, which hinders the exploration of the preparation parameters of the potassium oxide fire extinguishing agent with the best performance, that is, the specific weight percentages corresponding to the various materials.

[0004] When exploring the specific weight percentage combination of multiple materials corresponding to the best performance product, the general pattern is to put a large number of specific weight percentage combinations into a large number of actual tests one by one to obtain a large number of different potassium oxide fire extinguishing agents, and then ignite each portion of potassium oxide fire extinguishing agent to test the fire extinguishing volume and fire extinguishing time, obtain the fire extinguishing volume and fire disaster time of each portion of potassium oxide fire extinguishing agent, and finally evaluate the performance of each portion of potassium oxide fire extinguishing agent, and use the specific weight percentage combination corresponding to the potassium oxide fire extinguishing agent with the best performance as the optimized preparation parameter.

[0005] Obviously, this exploration mode has two major drawbacks: first, the number of experiments is large, each experimental process is complicated, and the entire exploration process is very time-consuming; second, the weight percentage fluctuation and change gradient of each material are limited, resulting in a large number of specific weight percentage combinations obtained that cannot effectively represent all numerical combinations, and thus makes it difficult for the optimal performance of the potassium oxide fire extinguishing agent obtained to be close to the actual optimal performance. Summary of the invention

[0006] In order to solve the above-mentioned drawbacks, the present invention provides an artificial intelligence system and a method for preparing a non-toxic potassium oxide fire extinguishing agent, which simulates the relationship between the fire extinguishing volume and fire extinguishing time of a set weight of the fire extinguishing agent and the weight percentage of the prepared pure raw materials through a neural network, without the need for repeated experiments.

[0007] To achieve the above-mentioned object, the present invention provides an artificial intelligence system, which comprises:

[0008] Potassium oxide fire extinguishing agent fire extinguishing model, first loss function calculation module, judgment module and weight percentage adjustment module, wherein the input layer of the potassium oxide fire extinguishing agent fire extinguishing agent inputs the weight of pure oxidant , weight of pure reducing agent , Weight of pure curing stabilizer and the weight of pure binder , the output layer outputs the fire extinguishing volume v and fire extinguishing time t, where, is the current weight percentage of the oxidant to the pure feed weight G, is the current weight percentage of the reducing agent to the pure raw material weight G, is the current weight percentage of the curing stabilizer to the pure raw material weight G, is the current weight percentage of the binder to the pure raw material weight G, and satisfies the following conditions ; The pure raw material weight G is the total weight of the pure oxidant, pure reducing agent, pure curing stabilizer and pure binder contained in the potassium oxide fire extinguishing agent;

[0009] The first loss function calculation module is configured to calculate the fire extinguishing volume v, fire extinguishing time t and weight according to the output layer of the potassium oxide fire extinguishing agent fire extinguishing model. The target volume of ionized potassium oxide generated by the finished fire extinguishing agent after ignition , Fire extinguishing target time Generate the first loss function , , is the impurity content;

[0010] The weight percentage adjustment module is configured to adjust the weight percentage according to the first loss function Calculate the gradient of the weight percentage a of the oxidant to the weight G of the pure raw material , the gradient of the weight percentage b of the reducing agent to the weight G of the pure raw material and the gradient of the weight percentage c of the curing stabilizer to the weight G of the pure raw material ; Update the weight percentage of the oxidant to the pure raw material weight G by the following formula respectively for , the weight percentage of reducing agent in the weight of pure raw material G for , the weight percentage of curing stabilizer in the weight of pure raw material G for , the weight percentage of binder in the pure raw material weight G for :

[0011] ;

[0012] ;

[0013] ;

[0014] ,

[0015] , and is the learning coefficient;

[0016] The judgment module is configured to judge the first loss function L 1 Is it the minimum? If so, make the first loss function L 1 The minimum weight percentage of the oxidant in the pure raw material weight G , the weight percentage of reducing agent in the weight of pure raw material G , the weight percentage of curing stabilizer in the weight of pure raw material G , the weight percentage of binder in the pure raw material weight G As the weight percentage of each component of the potassium oxide fire extinguishing agent, if not, the weight percentage of the oxidant in the pure raw material weight G after updating Assign to , the weight percentage of the reducing agent in the pure raw material weight G after update Assign to , the weight percentage of the curing stabilizer to the weight of the pure raw material G after updating Assign to , the weight percentage of the binder in the pure raw material weight G after the change Assign to , and re-input it into the potassium oxide fire extinguishing agent fire extinguishing model for processing.

[0017] In order to achieve the above-mentioned object of the invention, the present invention provides a method for preparing a non-toxic potassium oxide fire extinguishing agent, which comprises the following steps:

[0018] S3-01: Weight percentage of oxidant according to the output of the artificial intelligence system , weight percentage of reducing agent , weight percentage of curing stabilizer , Binder weight percentage Calculate the weight of the oxidant raw materials separately , weight of reducing agent raw material , weight of curing stabilizer raw materials and the weight of the binder raw materials ; is the purity of the oxidant, is the purity of the reducing agent, is the purity of the curing stabilizer, is the purity of the binder, It is the total weight of pure oxidant, pure reducing agent, pure curing stabilizer and pure binder contained in potassium oxide fire extinguishing agent;

[0019] S3-02: Set the weight to The oxidant raw material and the weight of The reducing agent raw materials are broken and mixed under sterile conditions and below -20°C to obtain a crude mixed material;

[0020] S3-03: filtering the coarse mixture through a filter sieve to obtain a fine mixture;

[0021] S3-04: drying the fine mixture so that the moisture content of the fine mixture is less than 1% to obtain a dry mixture;

[0022] S3-05: Dry mix and The curing stabilizer raw material is added to the weight of The binder raw material is stirred at 40° C. to 50° C. to obtain a resin mixture;

[0023] S3-06: leaving the resin mixture to stand at room temperature until it is initially cured to obtain an initially cured mixture;

[0024] S3-07: Freeze the initially solidified mixture to below -20°C and extrude it at a pressure of 7 MPa -7.9 MPa to obtain a molded product;

[0025] S3-08: drying the molded product at a temperature of 40°C to 50°C to obtain a dried molded product;

[0026] S3-09: Spray a layer of natural resin coating on the entire surface or both ends of the dried molded product to obtain potassium oxide fire extinguishing agent.

[0027] Compared with the prior art, the present invention has the following beneficial effects: the present invention simulates the relationship between the fire extinguishing volume, the fire extinguishing time and the weight percentage of the prepared raw materials through a neural network, and does not require repeated experiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The embodiments of the present invention will be described below with reference to the accompanying drawings, wherein:

[0029] Figure 1 It is a block diagram of the composition of the artificial intelligence system provided by the present invention. DETAILED DESCRIPTION

[0030] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0031] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. The meaning of "several" is one or more, unless otherwise clearly and specifically defined.

[0032] Figure 1 It is a block diagram of the composition of the artificial intelligence system provided by the present invention, such as Figure 1 As shown, the artificial intelligence system provided by the present invention includes:

[0033] Potassium oxide fire extinguishing agent fire extinguishing model, first loss function calculation module, judgment module and weight percentage adjustment module, wherein the input layer of the potassium oxide fire extinguishing agent fire extinguishing agent inputs the weight of pure oxidant , weight of pure reducing agent , Weight of pure curing stabilizer and the weight of pure binder , the output layer outputs the fire extinguishing volume v and fire extinguishing time t, where, is the current weight percentage of the oxidant to the pure feed weight G, is the current weight percentage of the reducing agent to the pure raw material weight G, is the current weight percentage of the curing stabilizer to the pure raw material weight G, is the current weight percentage of the binder to the pure raw material weight G, and satisfies the following conditions ; The weight of pure raw materials G is the total weight of pure oxidant, pure reducing agent, pure curing stabilizer and pure binder contained in potassium oxide fire extinguishing agent; the weight of pure oxidant , weight of pure reducing agent , Weight of pure curing stabilizer and the weight of pure binder The current weight percentage , Current weight percentage , Current weight percentage and current weight percentage It is obtained by multiplying the pure raw material weight G through a multiplier;

[0034] The first loss function calculation module is configured to calculate the fire extinguishing volume v, fire extinguishing time t and weight according to the output layer of the potassium oxide fire extinguishing agent fire extinguishing model. The target volume of ionized potassium oxide generated by the finished fire extinguishing agent after ignition , Fire extinguishing target time Generate the first loss function L 1 , , is the impurity content;

[0035] The weight percentage adjustment module is configured to adjust the weight percentage according to the first loss function Calculate the gradient of the weight percentage a of the oxidant to the weight G of the pure raw material , the gradient of the weight percentage b of the reducing agent to the weight G of the pure raw material and the gradient of the weight percentage c of the curing stabilizer to the weight G of the pure raw material ; Update the weight percentage of the oxidant to the pure raw material weight G by the following formula respectively for , the weight percentage of reducing agent in the weight of pure raw material G for , the weight percentage of curing stabilizer in the weight of pure raw material G for , the weight percentage of binder in the pure raw material weight G for :

[0036] ;

[0037] ;

[0038] ;

[0039] ,

[0040] , and is the learning coefficient;

[0041] The judgment module is configured to judge the first loss function L 1 Is it the minimum? If so, record the first loss function L 1 The minimum weight percentage of the oxidant in the pure raw material weight G , the weight percentage of reducing agent in the weight of pure raw material G , the weight percentage of curing stabilizer in the weight of pure raw material G , the weight percentage of binder in the pure raw material weight G As the weight percentage of each component of the potassium oxide fire extinguishing agent, if not, the weight percentage of the oxidant in the pure raw material weight G after updating Assign to , the weight percentage of the reducing agent in the pure raw material weight G after update Assign to , the weight percentage of the curing stabilizer to the weight of the pure raw material G after updating Assign to , the weight percentage of the binder in the pure raw material weight G after the change Assign to , and re-input it into the potassium oxide fire extinguishing agent fire extinguishing model for processing.

[0042] The fire extinguishing volume v and the fire extinguishing time t output by the output layer of the potassium oxide fire extinguishing agent fire extinguishing model in the present invention are respectively expressed by the following functions: , , are neural network parameters;

[0043] Assume the first loss function is: ;

[0044] therefore, , , , .

[0045] In the present invention, the weight percentage of the oxidant in the pure raw material weight G is , the weight percentage of reducing agent in the weight of pure raw material G , the weight percentage of curing stabilizer in the pure raw material weight G , the weight percentage of binder to the weight of pure raw material G The initial values ​​are: 50%, 20%, 0.1%, and 30% respectively.

[0046] In the present invention, the potassium oxide fire extinguishing agent fire extinguishing model is trained by a neural network, and the training steps include:

[0047] S2-01: Obtain N sets of training data and N sets of test data, where N is an integer greater than or equal to 1; each set of training data includes a set weight of The weight percentage of oxidant contained in the pure raw material , weight percentage of reducing agent , weight percentage of curing stabilizer and weight percentage of binder ; The test data is by weight Oxidant, weight Reducing agent, weight of curing stabilizer and weight The binder is made into potassium oxide fire extinguishing agent through the preparation process and ignited to generate ionic potassium oxide fire extinguishing volume and fire extinguishing time ,in, is the purity of the oxidant, is the purity of the reducing agent, is the purity of the curing stabilizer, is the purity of the binder, i=1,…,N, let ;

[0048] S2-02: Take the i-th group of data from N groups of training data and N groups of test data;

[0049] S2-03: The weight of the oxidant , weight of reducing agent , weight of curing stabilizer and the weight of the binder The composed vector is input to the input layer of the neural network, and the output of the neural network output layer is the fire extinguishing estimated volume and estimated time to extinguish the fire , the current parameters of the neural network are ;

[0050] S2-04: Based on fire extinguishing volume , Fire extinguishing time , Fire extinguishing estimated volume and estimated time to extinguish the fire Generate the second loss function , according to the second loss function Calculate the parameters of the neural network Gradient ;

[0051] S2-05: Using the second loss function L 2 Gradients with respect to neural network parameters Update the current parameters of Shen Network for :

[0052] ,

[0053] In the formula, is the learning rate;

[0054] S2-06: Determine whether the second loss function value is the minimum when the i-th group of data is used for training. If so, the training is completed using the i-th group of data, and then step S2-07 is executed; if not, the new parameter Assign to , and return to S2-03;

[0055] S2-07: , and determine whether i is greater than N. If not, return to step S2-02. If yes, end the training.

[0056] In the present invention, the second loss function is assumed to be: ,therefore,

[0057] .

[0058] The present invention also provides a method for preparing a non-toxic potassium oxide fire extinguishing agent, the method comprising the following steps:

[0059] S3-01: According to the weight percentage of oxidant output by the above artificial intelligence system , weight percentage of reducing agent , weight percentage of curing stabilizer , Binder weight percentage Calculate the weight of the oxidant raw materials separately , weight of reducing agent raw material , weight of curing stabilizer raw materials and the weight of the binder raw materials ; is the purity of the oxidant, is the purity of the reducing agent, is the purity of the curing stabilizer, is the purity of the binder, It is the total weight of pure oxidant, pure reducing agent, pure curing stabilizer and pure binder contained in potassium oxide fire extinguishing agent;

[0060] S3-02: Set the weight to The oxidant raw material and the weight of The reducing agent raw material is mixed by rapid wall breaking in a sterile environment at a temperature below -20°C to obtain a coarse mixture. In the present invention, the oxidant raw material and the reducing agent raw material are mixed by rapid wall breaking in a stirrer with a rotation speed of 2000r / min to 4000r / min; the temperature below -20°C is from -65°C to -20°C, and can specifically be -65°C, -60°C, -55°C, -50°C, -45°C, -40°C, -35°C, -30°C, -25°C, -20°C and any value therebetween; the rotation speed can be 2000r / min, 2500r / min, 3000r / min, 3500r / min, 4000r / min and any value therebetween;

[0061] S3-03: Filter the coarse mixture through a filter screen to obtain a fine mixture. The filter screen used in the present invention has a density of 150 mesh to 300 mesh, and the density can be 150 mesh, 160 mesh, 170 mesh, 180 mesh, 190 mesh, 200 mesh, 210 mesh, 220 mesh, 230 mesh, 240 mesh, 250 mesh, 260 mesh, 270 mesh, 280 mesh, 290 mesh, 300 mesh, and any value between them;

[0062] S3-04: drying the fine mixture so that the moisture content of the fine mixture is less than 1% to obtain a dry mixture;

[0063] S3-05: Dry mix and The curing stabilizer raw material is added to the weight of The binder raw material is slowly stirred at 40° C. to 50° C. to obtain a resin mixture; in the present invention, the slow stirring is carried out by a stirrer with a rotation speed of 40 r / min to 100 r / min, and the rotation speed can be 40 r / min, 50 r / min, 60 r / min, 70 r / min, 80 r / min, 90 r / min, 100 r / min and any value between them;

[0064] S3-06: leaving the resin mixture to stand at room temperature until it is initially cured to obtain an initially cured mixture;

[0065] S3-07: Freeze the initially solidified mixture to below -20°C and extrude it at a pressure of 7 MPa to 7.9 MPa to obtain a molded product, wherein the temperature below -20°C is between -65°C and -20°C, and specifically can be -65°C, -60°C, -55°C, -50°C, -45°C, -40°C, -35°C, -30°C, -25°C, -20°C and any value therebetween; the pressure can be 7.0 MPa, 7.2 MPa, 7.4 MPa, 7.6 MPa, 7.8 MPa, 7.9 MPa and any value therebetween;

[0066] S3-08: Drying the molded product at a temperature of 40°C to 50°C to obtain a dried molded product. The temperature may be 40°C, 42°C, 44°C, 46°C, 48°C, 50°C and any value therebetween;

[0067] S3-09: Spray a layer of natural resin coating on the entire surface or both ends of the dried molded product to obtain potassium oxide fire extinguishing agent.

[0068] In the present invention, the oxidant is one of the following potassium salts: potassium nitrate, potassium chlorate, potassium perchlorate, potassium magnesium cyanide and potassium dichromate, and the potassium nitrate is food grade potassium nitrate.

[0069] In the present invention, the reducing agent is one of the following materials: dicyandiamide, cellulose, Teflon and cellulose acetate.

[0070] In the present invention, the binder is one of the following materials: pure natural resin glue, phenol resin, melamine resin, unsaturated polyester resin, epoxy resin, various polymers, paint, gelatin, guar gum, agarose, camphor and paraffin.

[0071] The present invention mixes the oxidant and the reductant at low temperature, and extrude the initially solidified mixed material under high pressure at low temperature, thereby reducing the activity of the oxidant and improving the safety of production.

[0072] Comparative experiment

[0073] 3 liters of n-heptane were put into a 315×315×100 mm experimental fuel tray (about 0.1 square meters) for combustion, and potassium oxide fire extinguishing agent was prepared using the formula generated by the artificial intelligence system provided by the present invention for fire extinguishing experiments. For different maximum fire extinguishing coverage volumes and fire extinguishing times, one experiment was conducted respectively, and the formulas and experimental data of the pure raw materials for the final target products with set weights are shown in Example 1, Example 2 and Example 3 in Table 1 below. The formula was adjusted using the traditional method to prepare potassium oxide fire extinguishing agent target products for different maximum fire extinguishing coverage volumes and fire extinguishing times, and 8 experiments, 10 experiments and 15 experiments were conducted respectively. The formulas and experimental data of the final target products are shown in Comparative Examples 1, Comparative Example 2 and Comparative Example 3 in Table 1 below.

[0074] Table 1

[0075] .

[0076] From the experiments of Example 1 and Comparative Example 1 above, it can be seen that 45g of fire extinguishing agent wants to achieve a maximum coverage volume of about 5.0 cubic meters and a fire extinguishing time of about 25 seconds. The weight percentage of the raw materials provided by the artificial intelligence system provided by the present invention is obtained in only one experiment. The weight percentages of the raw materials for preparing the non-toxic potassium oxide fire extinguishing agent are: food grade potassium nitrate: 45.0wt%; dicyandiamide: 29.9wt%; stable curing agent: 0.1wt%; pure natural resin glue: 25.0wt%. By traditional methods, it took 8 experiments to obtain the weight percentages of the raw materials for preparing the non-toxic potassium oxide fire extinguishing agent, which are: food grade potassium nitrate: 44.1wt%; dicyandiamide: 30.8wt%; stable curing agent: 0.2wt%; pure natural resin glue: 24.9wt%.

[0077] From the experiments of Example 2 and Comparative Example 2, it can be seen that 45g of fire extinguishing agent wants to achieve a maximum coverage volume of about 3.0 cubic meters and a fire extinguishing time of about 31 seconds. The weight percentage of the raw materials provided by the artificial intelligence system provided by the present invention is obtained in only one experiment. The weight percentages of the raw materials for preparing the non-toxic potassium oxide fire extinguishing agent are: food-grade potassium nitrate: 43.0wt%; dicyandiamide: 32.0wt%; stable curing agent: 2.0wt%; pure natural resin glue: 23.0wt%. By traditional methods, it took 10 experiments to obtain the weight percentages of the raw materials for preparing the non-toxic potassium oxide fire extinguishing agent, which are: food-grade potassium nitrate: 37.4wt%; dicyandiamide: 35.5wt%; stable curing agent: 2.6wt%; pure natural resin glue: 24.5wt%.

[0078] From the experiments of Example 3 and Comparative Example 3, it can be seen that 45g of fire extinguishing agent wants to achieve a maximum coverage volume of about 1.5 cubic meters and a fire extinguishing time of about 40 seconds. The weight percentage of the raw materials provided by the artificial intelligence system provided by the present invention is obtained in only one experiment. The weight percentages of the raw materials for preparing the non-toxic potassium oxide fire extinguishing agent are: food-grade potassium nitrate: 29.0wt%; dicyandiamide: 38.0wt%; stable curing agent: 6.0wt%; pure natural resin glue: 27.0wt%. By the traditional method, it took 15 experiments to obtain the weight percentages of the raw materials for preparing the non-toxic potassium oxide fire extinguishing agent: food-grade potassium nitrate: 31.8wt%; dicyandiamide: 7.3wt%; stable curing agent: 5.4wt%; pure natural resin glue: 25.5wt%.

[0079] The present invention obtains multiple weight percentages of preparation materials for preparing potassium oxide fire extinguishing agent by correcting the weight percentages of preparation components through an artificial intelligence system, thereby reducing the number of actual tests, improving the speed of parameter verification, reducing the labor cost, time cost and economic cost of parameter verification, and obtaining reliable and effective preferred preparation parameters.

[0080] The foregoing description of the exemplary embodiments of the present invention is provided for the purpose of illustration and description. It is not intended to be exhaustive or to limit the present invention to the exact form disclosed. Obviously, many modifications and variations are obvious to those skilled in the art. The exemplary embodiments are selected and described in order to best illustrate the principles of the present invention and its practical application, so that other persons skilled in the art can understand the various embodiments and various variations of the present invention applicable to the specific purpose contemplated. It is intended that the scope of the present invention be limited by the appended claims and their equivalents.

Claims

1. An artificial intelligence system, characterized in that: include: Potassium oxide fire extinguishing agent fire extinguishing model, first loss function calculation module, judgment module and weight percentage adjustment module, wherein the input layer of the potassium oxide fire extinguishing agent fire extinguishing agent inputs the weight of pure oxidant , weight of pure reducing agent , Weight of pure curing stabilizer and the weight of pure binder , the output layer outputs the fire extinguishing volume v and fire extinguishing time t, where, is the current weight percentage of the oxidant to the pure feed weight G, is the current weight percentage of the reducing agent to the pure raw material weight G, is the current weight percentage of the curing stabilizer to the pure raw material weight G, is the current weight percentage of the binder to the pure raw material weight G, and satisfies the following conditions ; The pure raw material weight G is the total weight of the pure oxidant, pure reducing agent, pure curing stabilizer and pure binder contained in the potassium oxide fire extinguishing agent; The first loss function calculation module is configured to calculate the fire extinguishing volume v, fire extinguishing time t and weight according to the output layer of the potassium oxide fire extinguishing agent fire extinguishing model. The target volume of ionized potassium oxide generated by the finished fire extinguishing agent after ignition , Fire extinguishing target time Generate the first loss function , , is the impurity content; The weight percentage adjustment module is configured to adjust the weight percentage according to the first loss function Calculate the gradient of the weight percentage a of the oxidant to the weight G of the pure raw material , the gradient of the weight percentage b of the reducing agent to the weight G of the pure raw material and the gradient of the weight percentage c of the curing stabilizer to the weight G of the pure raw material ; Update the weight percentage of the oxidant to the pure raw material weight G by the following formula respectively for , the weight percentage of reducing agent in the weight of pure raw material G for , the weight percentage of curing stabilizer in the weight of pure raw material G for , the weight percentage of binder in the pure raw material weight G for : ; ; ; , , and is the learning coefficient; The judgment module is configured to judge whether the first loss function L1 is the minimum, and if so, the weight percentage of the oxidant that minimizes the first loss function L1 to the weight of the pure raw material G , the weight percentage of reducing agent in the weight of pure raw material G , the weight percentage of curing stabilizer in the weight of pure raw material G , the weight percentage of binder in the pure raw material weight G As the weight percentage of each component of the potassium oxide fire extinguishing agent, if not, the weight percentage of the oxidant in the pure raw material weight G after updating Assign to , the weight percentage of the reducing agent in the pure raw material weight G after update Assign to , the weight percentage of the curing stabilizer to the weight of the pure raw material G after updating Assign to , the weight percentage of the binder in the pure raw material weight G after the change Assign to , and re-input it into the potassium oxide fire extinguishing agent fire extinguishing model for processing.

2. The artificial intelligence system according to claim 1, characterized in that: The potassium oxide fire extinguishing agent fire extinguishing model is trained by a neural network. The training steps include: S2-01: Obtain N sets of training data and N sets of test data, where N is an integer greater than or equal to 1; each set of training data includes a set weight of Weight percentage of oxidant contained in pure raw material , weight percentage of reducing agent , weight percentage of curing stabilizer and weight percentage of binder ; The test data is by weight Oxidant, weight Reducing agent, weight of curing stabilizer and weight The binder is made into potassium oxide fire extinguishing agent through the preparation process, and the ionized potassium oxide generated by ignition has an extinguishing volume of and fire extinguishing time ,in, is the purity of the oxidant, is the purity of the reducing agent, is the purity of the curing stabilizer, is the purity of the binder, i=1,…,N, let ; S2-02: Take the i-th group of data from N groups of training data and N groups of test data; S2-03: The weight of the oxidant , weight of reducing agent , weight of curing stabilizer and the weight of the binder The composed vector is input to the input layer of the neural network, and the output of the neural network output layer is the fire extinguishing estimated volume and estimated time to extinguish the fire , the current parameters of the neural network are ; S2-04: Based on fire extinguishing volume , Fire extinguishing time , Fire extinguishing estimated volume and estimated time to extinguish the fire Generate the second loss function , according to the second loss function Calculate the parameters of the neural network Gradient ; S2-05: Using the second loss function Gradients with respect to neural network parameters Update the current parameters of the neural network for : ， In the formula, is the learning rate; S2-06: Determine whether the second loss function L2 is the minimum when the i-th group of data is used for training. If so, the training is completed using the i-th group of data, and then step S2-07 is executed; if not, the new parameter Assign to , and return to S2-03; S2-07: , and determine whether i is greater than N. If not, return to step S2-02. If yes, end the training.

3. A method for preparing a non-toxic potassium oxide fire extinguishing agent, characterized in that: The method comprises the following steps: S3-01: The weight percentage of the oxidant output by the artificial intelligence system according to any one of claims 1-2 , weight percentage of reducing agent , weight percentage of curing stabilizer , Binder weight percentage Calculate the weight of the oxidant raw materials separately , weight of reducing agent raw material , weight of curing stabilizer raw materials and the weight of the binder raw materials ; is the purity of the oxidant, is the purity of the reducing agent, is the purity of the curing stabilizer, is the purity of the binder, It is the total weight of pure oxidant, pure reducing agent, pure curing stabilizer and pure binder contained in potassium oxide fire extinguishing agent; S3-02: Set the weight to The oxidant raw material and the weight of The reducing agent raw materials are broken and mixed under sterile conditions and below -20°C to obtain a crude mixed material; S3-03: filtering the coarse mixture through a filter sieve to obtain a fine mixture; S3-04: drying the fine mixture so that the moisture content of the fine mixture is less than 1% to obtain a dry mixture; S3-05: Dry mix and The curing stabilizer raw material is added to the weight of The binder raw material is stirred at 40° C. to 50° C. to obtain a resin mixture; S3-06: leaving the resin mixture to stand at room temperature until it is initially cured to obtain an initially cured mixture; S3-07: Freeze the initially solidified mixture to below -20°C and extrude it at a pressure of 7 MPa -7.9 MPa to obtain a molded product; S3-08: drying the molded product at a temperature of 40°C to 50°C to obtain a dried molded product; S3-09: Spray a layer of natural resin coating on the entire surface or both ends of the dried molded product to obtain potassium oxide fire extinguishing agent.

4. The method for preparing the non-toxic potassium oxide fire extinguishing agent according to claim 3, characterized in that: The -20°C or below refers to -65°C to -20°C.

5. The method for preparing the non-toxic potassium oxide fire extinguishing agent according to claim 4, characterized in that: The oxidizing agent is one of the following potassium salts: potassium chlorate, potassium nitrate, potassium perchlorate, potassium magnesium cyanide and potassium dichromate.

6. The method for preparing the non-toxic potassium oxide fire extinguishing agent according to claim 5, characterized in that: The reducing agent is one of the following materials: melamine, Teflon, dicyandiamide and cellulose acetate.

7. The method for preparing the non-toxic potassium oxide fire extinguishing agent according to claim 6, characterized in that: The binder is one of the following materials: phenol resin, unsaturated polyester resin, epoxy resin, pure natural resin glue, lacquer, gelatin, guar gum, agarose, camphor and paraffin.

Images