Safety early warning method and system for coking coal detection chamber

Abstract

The invention belongs to the technical field of safety early warning, and particularly relates to a safety early warning method and system for a coking coal detection room, and the method comprises the steps: constructing the system state transition potential energy at the current moment according to a multi-dimensional state vector and a risk evolution rate vector at the current moment, and calculating the risk level certainty weight at the current moment, correcting the decision function value of the multidimensional state vector at the current moment input by each dichotomy support vector machine of the support vector machine model through the risk level deterministic weight, and taking the risk level corresponding to the dichotomy support vector machine with the maximum corrected decision value as a risk level predicted value at the current moment; and triggering a corresponding safety early warning measure according to the risk level prediction value at the current moment. According to the method, the problem of class imbalance is solved, the early warning sensitivity to high-risk events is improved, meanwhile, the decision fuzziness of multi-class division is effectively eliminated, and the certainty and reliability of early warning are enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of safety early warning, and more particularly to a safety early warning method and system for a coking coal detection chamber. Background Art

[0002] The coking coal testing room is a special closed or semi-closed space that brings together high-risk substances (coal samples, coal gas, coal dust) and precision testing equipment, and there is a potential risk of coal dust explosion. Therefore, it is crucial to monitor multiple environmental parameters in the testing room, such as gas concentration, temperature, humidity, etc. in real time and provide reliable safety warnings.

[0003] The concentrations of methane and carbon monoxide in the detection room were within the limit, but their specific combination with air temperature and humidity (such as high temperature and low humidity) would significantly amplify the comprehensive risk of dust explosion or human poisoning. Therefore, it is necessary to establish a dynamic risk assessment model based on the coupling of multi-dimensional parameters (multiple gas concentrations, temperature, humidity, and air pressure).

[0004] Support Vector Machines (SVMs) have unique advantages in handling small sample, high-dimensional, and nonlinear problems. Safety incidents in coking coal testing rooms, especially high-risk samples, are usually sparse, which is very suitable for the small sample learning characteristics of SVMs. Through kernel functions, SVMs can effectively handle the nonlinear relationships between various parameters, demarcate clear safety boundaries for different complex risk states, and achieve classification warnings with strong robustness and good generalization capabilities.

[0005] Support vector machines can effectively solve multi-classification problems by training multiple binary support vector machines through a one-to-many strategy. However, there are difficulties in processing inseparable regions and imbalanced categories, resulting in poor recognition of the small but critical "warning" or "danger" states and an inability to clearly determine the risk level. Summary of the Invention

[0006] In order to solve the technical problems of "indivisible areas" and "category imbalance" in the above-mentioned multi-category risk division, the present invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides a safety warning method for a coking coal detection room, comprising: collecting environmental data of the coking coal detection room for constructing a multidimensional state vector at each moment; using the multidimensional state vectors and risk level labels at historical moments as training data sets to train a support vector machine model, wherein the support vector machine model is composed of multiple binary support vector machines; constructing a risk evolution rate vector at the current moment according to the changing trend of the multidimensional state vector at the current moment over time; constructing a system state transfer potential at the current moment according to the multidimensional state vector and the risk evolution rate vector at the current moment; calculating a risk level deterministic weight at the current moment according to the system state transfer potential; calculating a corrected decision value of each binary support vector machine for the input multidimensional state vector at the current moment according to the decision function value of each binary support vector machine of the support vector machine model for the input multidimensional state vector at the current moment and the risk level deterministic weight at the current moment; taking the risk level corresponding to the binary support vector machine with the largest corrected decision value as the risk level prediction value at the current moment; triggering corresponding safety warning measures according to the risk level prediction value at the current moment.

[0008] By introducing state transition potential and deterministic weights, the present invention gives additional decision weights to data points with dangerous evolution trends. Even if such high-risk samples are very rare in the training set, the model can effectively identify and amplify them in actual predictions, greatly reducing the omission rate of serious safety accidents; further, when the data points are located in the fuzzy boundary area of ​​multiple binary support vector machines, the adaptive weighting mechanism of the present invention can provide a clear decision-making guide that tends to a higher safety level based on the dynamic evolution trend of the risk, eliminating the uncertainty of the early warning system at critical moments and ensuring the decisiveness and correctness of the early warning decision.

[0009] Preferably, the environmental data includes: methane concentration, carbon monoxide concentration, temperature, humidity, and air pressure.

[0010] Preferably, there are four risk levels, representing "safety", "attention", "warning" and "danger", which are respectively recorded as the first risk level, the second risk level, the third risk level and the fourth risk level; when the risk level is the first risk level, the risk level label is "1"; when the risk level is the second risk level, the risk level label is "2"; when the risk level is the third risk level, the risk level label is "3"; when the risk level is the fourth risk level, the risk level label is "4".

[0011] Preferably, the binary support vector machine is trained to distinguish a specific risk level from all other risk levels, and each trained binary support vector machine corresponds to a risk level.

[0012] The present invention trains multiple binary classification support vector machines through a one-to-many strategy, which can effectively solve the multi-classification problem and realize the prediction of risk level by combining the multidimensional characteristics of environmental data.

[0013] Preferably, the binary support vector machine training process includes: Risk level, , equalize the risk level label to The samples with risk level labels equal to The samples are marked as negative; according to the positive samples and negative samples, a two-class support vector machine is trained to distinguish the Risk level and For all risk levels other than the risk level, the goal of training the binary support vector machine is to maximize the interval between positive samples and negative samples and obtain the hyperplane that separates positive samples from negative samples with the largest interval.

[0014] Preferably, the constructing of the risk evolution rate vector at the current moment includes: Where, For the current moment The risk evolution rate vector; and The current moment and time The multidimensional state vector of Indicates the preset time step.

[0015] The present invention captures the dynamic change trend of the risk state by constructing a risk evolution rate vector, providing a data basis for the system state transfer potential at the subsequent construction moment.

[0016] Preferably, the step of constructing the system state transition potential at the current moment includes: Where, For the current moment The system state transition potential energy; For the current moment The multidimensional state vector Dimensions; It is Risk contribution coefficient of each environmental parameter; It is the current moment The multidimensional state vector No. Quantity It is the current moment The risk evolution rate vector No. Quantity is the sign function. When the rate is positive, the sign function , when the rate is negative, the sign function .

[0017] The present invention constructs a model that quantifies the system's tendency to transition to a higher risk level by constructing the system state transition potential at the current moment, which is used to adjust the weight of the binary support vector machine for decision-making.

[0018] Preferably, the calculation of the risk level certainty weight at the current moment includes: Where, For the current moment The certainty weight of the risk level; For the current moment The system state transition potential energy; represents the natural exponential function.

[0019] Preferably, the calculation of the modified decision value of each binary support vector machine for the input multi-dimensional state vector at the current moment includes: Where, For the current moment The multidimensional state vector of For the A binary support vector machine classifies the multidimensional state vector of the current moment into The revised decision value of Indicates the A binary support vector machine classifies the input The decision function value refers to the multidimensional state vector at the current moment Distance The distance between the hyperplanes of the two-class support vector machine; represents the risk bias function, Represents the risk bias function for The risk bias value of the risk level, and ; The deterministic weight of the risk level at the current moment.

[0020] The present invention corrects the decision function value through the risk bias value of the risk level and the certainty weight of the risk level at the current moment, so that when the model faces ambiguous boundary situations, it will decisively tend to be safer and have a higher risk level early warning judgment, effectively compensating for the model bias caused by the scarcity of training samples.

[0021] In a second aspect, the present invention provides a safety warning system for a coking coal detection room, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned safety warning method for a coking coal detection room is implemented.

[0022] By adopting the above technical solution, the above-mentioned safety warning method for a coking coal detection room is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby making a terminal device based on the memory and the processor for easy use.

[0023] The beneficial effects of the present invention are: By introducing state transition potential and deterministic weights, the present invention gives additional decision weights to data points with dangerous evolution trends. Even if such high-risk samples are very rare in the training set, the model can effectively identify and amplify them in actual predictions, greatly reducing the omission rate of serious safety accidents; further, when the data points are located in the fuzzy boundary area of ​​multiple binary support vector machines, the adaptive weighting mechanism of the present invention can provide a clear decision-making guide that tends to a higher safety level based on the dynamic evolution trend of the risk, eliminating the uncertainty of the early warning system at critical moments and ensuring the decisiveness and correctness of the early warning decision. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart schematically illustrating a safety early warning method for a coking coal detection chamber in the present invention; Figure 2 is a flowchart schematically showing step S3. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0026] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0027] The embodiment of the present invention discloses a safety early warning method for a coking coal detection room, referring to Figure 1 , including steps S1 to S5: S1. Collect environmental data from the coking coal testing room to construct a multi-dimensional state vector at each moment.

[0028] Specifically, multiple sensors are deployed in the coking coal detection room to collect data on various environmental parameters in the coking coal detection room in real time as environmental data. The environmental parameters include but are not limited to methane concentration, carbon monoxide concentration, temperature, humidity, and air pressure. All environmental data are normalized and mapped to the [0,1] interval to eliminate dimensional effects.

[0029] Among them, the methane concentration is collected by a methane concentration sensor, and the types of methane concentration sensors include but are not limited to semiconductor gas sensors and infrared absorption methane sensors; the carbon monoxide concentration is collected by a carbon monoxide concentration sensor, and the types of carbon monoxide concentration sensors include but are not limited to electrochemical carbon monoxide sensors; the temperature is collected by a temperature sensor, and the types of temperature sensors include but are not limited to thermistors, NTC (negative temperature coefficient) sensors and digital temperature sensors; the humidity is collected by a humidity sensor, and the types of humidity sensors include but are not limited to capacitive humidity sensors; the air pressure is collected by an air pressure sensor, and the types of air pressure sensors include but are not limited to digital air pressure sensors.

[0030] Furthermore, the environmental data of multiple dimensions at each moment are taken as components to form a multidimensional state vector at each moment, which is used to comprehensively reflect the system state. The methane concentration, carbon monoxide concentration, temperature, humidity, and air pressure are the 1st, 2nd, 3rd, 4th, and 5th components of the multidimensional state vector respectively.

[0031] S2. Use the multi-dimensional state vectors and risk level labels of historical moments as training data sets to train a support vector machine model.

[0032] It should be noted that through a one-to-many strategy, training multiple binary support vector machines can effectively solve the multi-classification problem. In the risk level prediction scenario of the coking coal detection room, the powerful classification ability of the support vector machine can be fully utilized, and the multidimensional characteristics of the environmental data can be combined to realize the prediction of the risk level.

[0033] Therefore, the specific training process of the support vector machine model is: 1. Combining historical accident cases and safety drill data, experts evaluate the multidimensional state vectors at historical moments and assign a clear risk level label to the multidimensional state vectors at each historical moment.

[0034] Specifically, there are 4 risk levels, representing "safety", "attention", "warning" and "danger", and the 4 risk levels are respectively recorded as the first risk level, the second risk level, the third risk level and the fourth risk level; therefore, there are 4 risk level labels.

[0035] Furthermore, when the risk level is "safe", that is, the first risk level, the risk level label is "1"; when the risk level is "attention", that is, the second risk level, the risk level label is "2"; when the risk level is "warning", that is, the third risk level, the risk level label is "3"; when the risk level is "danger", that is, the fourth risk level, the risk level label is "4".

[0036] 2. Take the multidimensional state vector and risk level label of each historical moment as a sample, and form a complete, labeled training dataset based on the multidimensional state vectors and risk level labels of all historical moments.

[0037] 3. Using the training data set, multiple binary support vector machines are trained to form a support vector machine model. Each binary support vector machine is trained to distinguish a specific risk level from all other risk levels. Therefore, each trained binary support vector machine corresponds to a risk level.

[0038] It should be noted that the kernel function, namely the Gaussian radial basis function, is introduced into the support vector machine to map the original, linearly inseparable data into a higher-dimensional feature space. In the high-dimensional feature space, the data points of different risk levels that were originally entangled can now be separated by a hyperplane.

[0039] Specifically, for the Risk level, , equalize the risk level label to The samples with risk level labels equal to The samples are marked as negative; according to the positive samples and negative samples, a two-class support vector machine is trained to distinguish the Risk level and For all risk levels other than the risk level, the goal of training the binary support vector machine is to maximize the interval between positive samples and negative samples and obtain the hyperplane that separates positive samples from negative samples with the largest interval.

[0040] Among them, maximizing the margin can make the trained binary classification support vector machine have better robustness and generalization ability for the classification of unknown data, and is not prone to misjudgment due to slight perturbations of the data.

[0041] For example, for the first risk level, that is, the risk level is "safe", the samples with the risk level label equal to 1 are marked as positive, and the samples with the risk level label not equal to 1, that is, the risk level labels equal to 2, 3 and 4 are marked as negative, that is, the samples with risk levels of "attention", "warning" and "danger" are marked as negative; for the third risk level, that is, the risk level is "warning", the samples with the risk level label equal to 3 are marked as positive, and the samples with the risk level label not equal to 3, that is, the risk level labels equal to 1, 2 and 4 are marked as negative, that is, the samples with risk levels of "safe", "attention" and "danger" are marked as negative.

[0042] S3. Calculate the certainty weight of the risk level at the current moment.

[0043] It should be noted that the one-to-many strategy will construct multiple binary support vector machines; however, at the intersection of the boundaries of multiple binary support vector machines, "inseparable areas" or "fuzzy areas" are likely to occur, that is, a new data point may be accepted or rejected by multiple binary support vector machines at the same time, resulting in an inability to clearly determine its risk level; in addition, in safety warning scenarios, the vast majority of historical data belongs to the "safe" or "attention" status, while the data samples of the "warning" or "danger" status are very few, which will lead to serious class imbalance problems, and the standard support vector machine model will tend to favor categories with large sample sizes during training, resulting in poor recognition of the small but critical "warning" or "danger" status.

[0044] It should be further explained that in order to solve the processing difficulties of "indivisible areas" and "category imbalance" in multi-category risk division, this embodiment introduces a quantitative assessment of the dynamic evolution trend of risks and generates adaptive weights to adjust the decision-making process of the support vector machine model, thereby improving the recognition sensitivity of high-risk states and eliminating decision ambiguity. Therefore, it is necessary to first construct a model that can quantify the tendency of the system to transition to a higher risk level, that is, the state transition potential model, and adjust the decision weight of the support vector machine model based on this.

[0045] See the flowchart of step S3 Figure 2 , including steps S301 to S303, specifically: S301. Construct a risk evolution rate vector at the current moment based on the changing trend of the multi-dimensional state vector at the current moment over time.

[0046] Specifically, by comparing the current moment and time The multi-dimensional state vector of the current moment is calculated The risk evolution rate vector can capture the dynamic change trend of multi-dimensional risk status; at the current moment The calculation formula of the risk evolution rate vector is: ; Where, For the current moment The risk evolution rate vector, whose components represent the speed and direction of change of each environmental parameter; and The current moment and time The multidimensional state vector of Indicates the preset time step.

[0047] Among them, for any environmental parameter, such as methane concentration, a rapidly rising methane concentration carries a higher immediate risk than a slowly rising or stable methane concentration. Therefore, by constructing the risk evolution rate vector to characterize this dynamic trend.

[0048] Among them, the specific value of the preset time step can be set according to the actual application scenario and requirements, and the value range of the preset time step is [0.5, 3], the unit is minutes, and the present invention sets the preset time step to 1 minute.

[0049] S302: Construct the system state transfer potential energy at the current moment based on the multi-dimensional state vector and risk evolution rate vector at the current moment.

[0050] It should be noted that, from a physics perspective, an unstable system has a higher potential energy and tends to transition to a more stable low-energy state. Therefore, a dangerous coking coal detection room environment can be regarded as an unstable state with high-risk potential energy, and the system state transition potential energy is jointly determined by various environmental parameters. Therefore, the system state transition potential energy at the current moment is constructed based on the multi-dimensional state vector and risk evolution rate vector at the current moment.

[0051] Specifically, according to the multi-dimensional state vector and risk evolution rate vector at the current moment, the system state transfer potential energy at the current moment is constructed, then The calculation formula of the system state transfer potential energy at time is: ; Where, For the current moment The system state transition potential energy; For the current moment The multidimensional state vector Dimensions; It is The risk contribution coefficient of each environmental parameter reflects the contribution of different parameters to the total risk. The sum of the risk contribution coefficients of all environmental parameters is 1, that is, ,and ; It is the current moment The multidimensional state vector No. Quantity It is the current moment The risk evolution rate vector No. Quantity is a sign function. When the rate is positive, that is, the environmental data shows an upward trend, the sign function , when the rate is negative, that is, the environmental data shows a downward trend, the sign function .

[0052] Among them, the system state transition potential energy Determined by various environmental parameters: For any environmental parameter, the value The larger the growth rate The faster the system state transfer potential This shows that when the environmental parameters themselves are already at a high level and are still deteriorating rapidly, the system is in an extremely unstable high potential energy state and is very likely to transition to a higher risk level.

[0053] Among them, methane is a flammable gas that can easily cause explosion accidents. It is one of the most critical risk factors in the coking coal testing room and should be assigned a higher risk contribution coefficient. Therefore, the risk contribution coefficient of methane concentration ranges from [0.4, 0.5]. Carbon monoxide is a toxic gas that has direct harm to human health. Although carbon monoxide is more harmful, its concentration change usually lags behind that of methane. Therefore, its risk contribution coefficient is slightly lower than that of methane. Therefore, the risk contribution coefficient of carbon monoxide concentration ranges from [0.2, 0.3]. Temperature increases may accelerate the chemical reactions of methane or other flammable gases, increasing For explosion risk, temperature changes usually have an indirect impact on risk, so their weight is lower than that of methane and carbon monoxide. The range of the risk contribution coefficient of temperature is [0.15, 0.2]. Low humidity may cause dust to fly and increase the risk of fire. The impact of humidity on safety risk is relatively indirect, so its weight is lower. The range of the risk contribution coefficient of humidity is [0.1, 0.15]. Abnormal air pressure may affect the efficiency of the ventilation system and indirectly affect other environmental parameters. The impact of air pressure on safety risk is the smallest, so its weight is the lowest. The range of the risk contribution coefficient of air pressure is [0.05, 0.1].

[0054] In summary, the present invention uses the risk contribution coefficients of methane concentration, carbon monoxide concentration, temperature, humidity, and air pressure as the 、 、 、 、 Set them to 0.45, 0.23, 0.16, 0.1, and 0.06 respectively.

[0055] S303. Calculate the risk level certainty weight at the current moment based on the system state transfer potential at the current moment.

[0056] Specifically, based on the system state transfer potential at the current moment, the deterministic weight of the risk level at the current moment is obtained through a nonlinear mapping function; wherein the nonlinear mapping function refers to an S-shaped function, and the Sigmoid function is a common S-shaped function.

[0057] The current moment The calculation formula of the risk level certainty weight is: ; Where, For the current moment The risk level deterministic weight is in the range of (0,1); For the current moment The system state transition potential energy; represents the natural exponential function.

[0058] Among them, when the system state transfer potential energy When it is lower, Approaches 0; as the system state changes, the potential energy The increase, will grow rapidly and approach 1; therefore, In essence, it is the confidence level in the judgment that "the system is evolving towards a high-risk state". The higher the potential energy of the system state transition, the more certain it is that the system is becoming dangerous.

[0059] S4. Calculate the revised decision value of each binary support vector machine for the multidimensional state vector at the current moment based on the decision function value of each binary support vector machine of the support vector machine model for the multidimensional state vector at the current moment, and the risk level deterministic weight at the current moment; and use the risk level corresponding to the binary support vector machine with the largest revised decision value as the risk level prediction value at the current moment.

[0060] It should be noted that the deterministic weight of the risk level at the current moment obtained in step S3 is integrated into the decision-making process of multiple binary classification support vector machines of the support vector machine model.

[0061] Specifically, the multidimensional state vector at the current moment is input into each binary support vector machine of the support vector machine model obtained in step S2, thereby obtaining the decision function value of each binary support vector machine for the input multidimensional state vector at the current moment.

[0062] Furthermore, based on the decision function value of each binary support vector machine of the support vector machine model for the multidimensional state vector input at the current moment, and the risk level certainty weight at the current moment, the corrected decision value of each binary support vector machine for the multidimensional state vector input at the current moment is calculated, and the risk level corresponding to the binary support vector machine with the largest corrected decision value is used as the risk level prediction value at the current moment.

[0063] The specific calculation formula for the modified decision value of each binary support vector machine for the multi-dimensional state vector input at the current moment is: ; Where, For the current moment The multidimensional state vector of For the A binary support vector machine classifies the multidimensional state vector of the current moment into The revised decision value of Indicates the A binary support vector machine classifies the input The decision function value refers to the multidimensional state vector at the current moment Distance The distance between the hyperplanes of the two-class support vector machine; Represents the risk bias function, which is a risk bias function that is positively correlated with the input risk level. Represents the risk bias function for The risk bias value of the risk level, and , which can ensure that the higher the risk level of the category, the greater the weight bonus it obtains; The deterministic weight of the risk level at the current moment.

[0064] Among them, when making the final decision, the decision function value of each category will be passed Make corrections; when the system shows a strong tendency to evolve towards a dangerous state, It is close to 1. The higher the risk level, the higher the risk level. The larger the value, the greater the decision function value bonus obtained; this allows the model to decisively lean towards safer, higher-risk warning judgments when faced with ambiguous boundary situations; at the same time, since the decision function value of high-risk categories is dynamically improved, it effectively compensates for the model bias caused by scarce training samples.

[0065] It should be noted that this embodiment solves the problem of category imbalance and improves the early warning sensitivity for high-risk events. At the same time, it effectively eliminates the decision ambiguity of multi-category division and enhances the certainty and reliability of early warning.

[0066] S5. Trigger corresponding safety warning measures based on the risk level prediction value at the current moment.

[0067] Specifically, in the coking coal testing room, appropriate safety warning measures are taken according to the different risk levels to ensure personnel safety and normal operation of equipment. The following are specific response measures for each risk level: 1. If the current risk level prediction value is the first risk level, that is, the risk level is "safe", it means that the environmental data are normal. Therefore, continue to monitor the environmental data to ensure the normal operation of the system. At the same time, record the current environmental parameters for subsequent analysis.

[0068] 2. If the current risk level prediction value is the second risk level, that is, the risk level is "Caution", it means that some environmental parameters are close to critical values ​​and there may be a slight risk. Therefore, shorten the data collection interval, track the change trend in real time, and make local adjustments, that is, check whether the ventilation system is operating normally and increase ventilation appropriately.

[0069] 3. If the current risk level prediction value is the third risk level, that is, the risk level is "warning", it means that certain environmental parameters have exceeded the critical value and there is an obvious risk. Therefore, it is necessary to activate the sound and light alarm device to alert all personnel and immediately start the emergency ventilation system to reduce the concentration of harmful gases. During this period, unrelated personnel are prohibited from entering the detection room. Subsequently, professional personnel will be organized to conduct a comprehensive inspection of the detection room, locate the source of the problem, and record the abnormal data and processing process in detail for subsequent analysis.

[0070] 4. If the current risk level prediction value is the fourth risk level, that is, the risk level is "dangerous", it means that the environmental parameters are seriously exceeded, and there is a risk of direct threat to personal safety and equipment damage. Therefore, notify all personnel to evacuate the testing room quickly, turn off all non-essential power supplies to prevent electrical sparks from causing explosions, and follow the pre-established emergency plan, including evacuating personnel, isolating dangerous areas, etc. At the same time, contact professional rescue teams (such as the fire brigade) for on-site disposal. After the danger is eliminated, conduct an in-depth analysis of the cause of the incident and improve safety measures.

[0071] It should be noted that this embodiment not only analyzes the static parameters of the current environment, but more importantly, quantifies its dynamic evolution trend. This enables the early warning model to capture the process of risk accumulation and development, and thus issue early warnings when various environmental parameters have not yet reached the traditional fixed thresholds but have already shown dangerous combinations and rapid deterioration trends, thereby buying valuable time for taking preventive measures.

[0072] An embodiment of the present invention further discloses a safety warning system for a coking coal detection room, comprising a processor and a memory, wherein the memory stores computer program instructions. When the computer program instructions are executed by the processor, a safety warning method for a coking coal detection room according to the present invention is implemented.

[0073] The above system also includes other components well known to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

Claims

1. A safety early warning method for a coking coal detection room, characterized in that: include: Collect environmental data from the coking coal testing room to construct a multi-dimensional state vector at each moment; The multidimensional state vectors and risk level labels of historical moments are used as training data sets to train a support vector machine model, wherein the support vector machine model is composed of multiple binary classification support vector machines; Based on the changing trend of the multi-dimensional state vector at the current moment over time, the risk evolution rate vector at the current moment is constructed; based on the multi-dimensional state vector and the risk evolution rate vector at the current moment, the system state transfer potential energy at the current moment is constructed; Calculate the risk level certainty weight at the current moment based on the system state transfer potential at the current moment; According to the deterministic weight of the risk level at the current moment, the decision function value of each binary support vector machine of the support vector machine model for the multidimensional state vector input at the current moment is corrected to obtain the corrected decision value of each binary support vector machine for the multidimensional state vector input at the current moment; The risk level corresponding to the binary support vector machine with the largest corrected decision value is used as the risk level prediction value at the current moment; According to the predicted value of the risk level at the current moment, corresponding safety warning measures are triggered.

2. A safety early warning method for a coking coal detection room according to claim 1, characterized in that: The environmental data include: methane concentration, carbon monoxide concentration, temperature, humidity, and air pressure.

3. The safety early warning method for a coking coal detection room according to claim 1, characterized in that: There are four risk levels, namely "safety", "attention", "warning" and "danger", which are recorded as the first risk level, the second risk level, the third risk level and the fourth risk level respectively; When the risk level is the first risk level, the risk level label is "1"; When the risk level is "Second Risk Level", the risk level label is "2"; When the risk level is the third risk level, the risk level label is "3"; when the risk level is the fourth risk level, the risk level label is "4".

4. The safety early warning method for a coking coal detection room according to claim 1, characterized in that: The binary support vector machine is trained to distinguish a specific risk level from all other risk levels, and each trained binary support vector machine corresponds to a risk level.

5. The safety early warning method for a coking coal detection room according to claim 1, characterized in that: The binary support vector machine training process includes: For the Risk level, , equalize the risk level label to The samples with risk level labels equal to The samples are marked as negative; according to the positive samples and negative samples, a two-class support vector machine is trained to distinguish the Risk level and For all risk levels other than the risk level, the goal of training the binary support vector machine is to maximize the interval between positive samples and negative samples and obtain the hyperplane that separates positive samples from negative samples with the largest interval.

6. The safety early warning method for a coking coal detection room according to claim 1, characterized in that: The constructing of the risk evolution rate vector at the current moment includes: ； Where, For the current moment The risk evolution rate vector; and The current moment and time The multidimensional state vector of Indicates the preset time step.

7. The safety early warning method for a coking coal detection room according to claim 1, characterized in that: The constructing of the system state transition potential at the current moment includes: ； Where, For the current moment The system state transition potential energy; For the current moment The multidimensional state vector Dimensions; It is Risk contribution coefficient of each environmental parameter; It is the current moment The multidimensional state vector No. Quantity It is the current moment The risk evolution rate vector No. Quantity is the sign function. When the rate is positive, the sign function , when the rate is negative, the sign function .

8. The safety early warning method for a coking coal detection room according to claim 1, characterized in that: The calculation of the risk level certainty weight at the current moment includes: ； Where, For the current moment The certainty weight of the risk level; For the current moment The system state transition potential energy; represents the natural exponential function.

9. The safety early warning method for a coking coal detection room according to claim 1, characterized in that: The calculation of the modified decision value of each binary support vector machine for the input multi-dimensional state vector at the current moment includes: ； Where, For the current moment The multidimensional state vector of For the A binary support vector machine classifies the multidimensional state vector of the current moment into The revised decision value of Indicates the A binary support vector machine classifies the input The decision function value refers to the multidimensional state vector at the current moment Distance The distance between the hyperplanes of the two-class support vector machine; represents the risk bias function, Represents the risk bias function for the The risk bias value of the risk level, and ; The deterministic weight of the risk level at the current moment.

10. A safety warning system for a coking coal detection room, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a safety early warning method for a coking coal detection chamber according to any one of claims 1 to 9 is implemented.

Images