Intelligent detection system and method for harmful gas of shield tunneling machine

Abstract

The invention relates to the technical field of shield tunneling machine gas detection, and discloses a shield tunneling machine harmful gas intelligent detection system and method, which monitors the harmful gas concentration and environmental parameters in a shield tunneling machine in real time, and transmits data to an early warning terminal by using a communication module. On the basis, an AI technology is introduced to dynamically adjust an initial alarm threshold value, and accurate correction is performed according to environmental parameters (such as temperature and humidity), so that the alarm accuracy is improved. In addition, a forward LSTM model is adopted to analyze environment state coding features, a threshold adjustment coefficient is calculated, and finer risk assessment is realized. Therefore, not only is the intelligent level of harmful gas monitoring improved, but also a foundation is laid for constructing a comprehensive safety protection system.

Description

Technical Field

[0001] This application relates to the technical field of gas detection for shield machines, and more specifically, to an intelligent harmful gas detection system and method for shield machines. Background Art

[0002] During the current shield construction process, the complex underground geological environment may cause the leakage of various toxic and harmful gases such as hydrogen sulfide and methane, posing a serious threat to construction safety. Existing technical solutions, such as an intelligent toxic and harmful gas detection system for shield machines with the publication number CN214847059U, although able to provide real-time monitoring and early warning functions to a certain extent, still have some deficiencies and need further improvement.

[0003] Specifically, this existing system mainly relies on fixed alarm thresholds to determine whether a dangerous situation occurs. This static threshold setting method fails to fully consider the complex and variable factors in the actual construction environment. For example, changes in environmental parameters such as temperature and humidity may significantly affect the measurement results of gas concentration. Therefore, in some cases, the fixed threshold may not accurately reflect the real dangerous situation, resulting in false alarms or missed alarms, thus affecting construction safety and efficiency.

[0004] Therefore, an optimized intelligent harmful gas detection solution for shield machines is expected. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide an intelligent harmful gas detection system and method for shield machines, which not only improve the intelligent level of harmful gas monitoring but also lay a foundation for building a comprehensive safety protection system.

[0006] According to one aspect of this application, an intelligent harmful gas detection method for shield machines is provided, including: real-time monitoring of the harmful gas concentration inside the shield machine and real-time monitoring of environmental parameters to obtain harmful gas concentration data and environmental parameter data; transmitting the harmful gas concentration data and the environmental parameter data to an early warning terminal through a communication module; extracting an initial alarm threshold and dynamically adjusting the initial alarm threshold based on the environmental parameter data to obtain a dynamically adjusted threshold, wherein the initial alarm threshold is dynamically adjusted based on the time-series depth collaborative analysis between the environmental temperature value and the environmental humidity value in the environmental parameter data; and judging whether the harmful gas concentration is abnormal based on the comparison between the harmful gas concentration data and the dynamically adjusted threshold.

[0007] In the above intelligent harmful gas detection method for shield machines, the environmental parameters include an environmental temperature value and an environmental humidity value.

[0008] In the above intelligent detection method for harmful gases of a shield machine, an initial alarm threshold is extracted, and the initial alarm threshold is dynamically adjusted based on the environmental parameter data to obtain a dynamically adjusted threshold, including: sorting the environmental parameter data based on timestamps to obtain a time queue of environmental temperature values and a time series of environmental humidity values; performing sequence encoding on the time queue of environmental temperature values and the time series of environmental humidity values based on an environmental parameter monitoring time window to obtain a time queue of moving average values of environmental temperature and a time queue of moving average values of environmental humidity; calculating a threshold adjustment coefficient based on the time queue of moving average values of environmental temperature and the time queue of moving average values of environmental humidity; and dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient to obtain the dynamically adjusted threshold.

[0009] In the above intelligent detection method for harmful gases of a shield machine, performing sequence encoding on the time queue of environmental temperature values and the time series of environmental humidity values based on an environmental parameter monitoring time window to obtain a time queue of moving average values of environmental temperature and a time queue of moving average values of environmental humidity, including: performing sequence segmentation on the time queue of environmental temperature values based on the environmental parameter monitoring time window to obtain a sequence of environmental temperature monitoring time windows; calculating the mean value of environmental temperature values within each environmental temperature monitoring time window in the sequence of environmental temperature monitoring time windows to obtain the time queue of moving average values of environmental temperature; performing sequence segmentation on the time queue of environmental humidity values based on the environmental parameter monitoring time window to obtain a sequence of environmental humidity monitoring time windows; and calculating the mean value of environmental humidity values within each environmental humidity monitoring time window in the sequence of environmental humidity monitoring time windows to obtain the time queue of moving average values of environmental humidity.

[0010] In the above intelligent detection method for harmful gases of a shield machine, calculating a threshold adjustment coefficient based on the time queue of moving average values of environmental temperature and the time queue of moving average values of environmental humidity, including: performing sequence encoding on the time queue of moving average values of environmental temperature based on a forward LSTM model to obtain environmental temperature time series fluctuation features; performing sequence encoding on the time queue of moving average values of environmental humidity based on a forward LSTM model to obtain environmental humidity time series fluctuation features; performing environmental parameter time series collaborative encoding on the environmental temperature time series fluctuation features and the environmental humidity time series fluctuation features to obtain environmental state encoding features; and performing environmental compensation decoding based on a decoder on the environmental state encoding features to obtain the threshold adjustment coefficient.

[0011] In the above intelligent detection method for harmful gases of a shield machine, dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient to obtain the dynamically adjusted threshold includes: dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient according to the following formula: ; where is the initial alarm threshold, is the threshold adjustment coefficient, is the dynamically adjusted threshold.

[0012] In the above intelligent detection method for harmful gases of a shield machine, performing environmental parameter time-series collaborative coding on the time-series fluctuation characteristics of the environmental temperature and the time-series fluctuation characteristics of the environmental humidity to obtain environmental state coding features includes: performing non-linear feature mapping reconstruction on the time-series fluctuation characteristics of the environmental temperature and the time-series fluctuation characteristics of the environmental humidity to obtain a set of environmental temperature local feature coding vectors and a set of environmental humidity local feature coding vectors; calculating a potential association feature joint coding matrix between each corresponding environmental temperature local feature coding vector and environmental humidity local feature coding vector in the set of environmental temperature local feature coding vectors and the set of environmental humidity local feature coding vectors to obtain a set of environmental temperature-environmental humidity potential association feature joint coding matrices; calculating the sparse dimension control weights of each environmental temperature-environmental humidity potential association feature joint coding matrix in the set of environmental temperature-environmental humidity potential association feature joint coding matrices to obtain a set of environmental temperature-environmental humidity sparse dimension control weights; and based on the set of environmental temperature-environmental humidity sparse dimension control weights, performing dynamic weight fusion on the set of environmental temperature-environmental humidity potential association feature joint coding matrices to obtain the environmental state coding features.

[0013] In the above intelligent detection method for harmful gases of a shield machine, calculating a set of potential association feature joint coding matrices between each corresponding environmental temperature local feature coding vector and environmental humidity local feature coding vector in the set of environmental temperature local feature coding vectors and the set of environmental humidity local feature coding vectors to obtain a set of environmental temperature - environmental humidity potential association feature joint coding matrices includes: performing association coding on each corresponding environmental temperature local feature coding vector and environmental humidity local feature coding vector in the set of environmental temperature local feature coding vectors and the set of environmental humidity local feature coding vectors to obtain a set of initial environmental temperature - environmental humidity potential association feature joint coding matrices; constructing a non - linear solution matrix based on each corresponding environmental temperature local feature coding vector and environmental humidity local feature coding vector; calculating a local linear norm - preserving constraint matrix of the initial environmental temperature - environmental humidity potential association feature joint coding matrix; and performing projection smoothing optimization on the initial environmental temperature - environmental humidity potential association feature joint coding matrix based on the non - linear solution matrix and the local linear norm - preserving constraint matrix to obtain the set of environmental temperature - environmental humidity potential association feature joint coding matrices.

[0014] In the above intelligent detection method for harmful gases of a shield machine, calculating a set of sparse dimension regulation weights for each environmental temperature - environmental humidity potential association feature joint coding matrix in the set of environmental temperature - environmental humidity potential association feature joint coding matrices to obtain a set of environmental temperature - environmental humidity sparse dimension regulation weights includes: calculating the square of the Frobenius norm of the environmental temperature - environmental humidity potential association feature joint coding matrix to obtain an environmental temperature - environmental humidity energy scaling factor; and processing the environmental temperature - environmental humidity energy scaling factor through an activation function to obtain the environmental temperature - environmental humidity sparse dimension regulation weights.

[0015] According to another aspect of the present application, there is also provided an intelligent detection system for harmful gases of a shield machine, including: a shield machine data acquisition module for real - time monitoring of the harmful gas concentration inside the shield machine and real - time monitoring of environmental parameters to obtain harmful gas concentration data and environmental parameter data; a shield machine data transmission module for transmitting the harmful gas concentration data and the environmental parameter data to a warning terminal through a communication module; a dynamic adjustment threshold determination module for extracting an initial alarm threshold and dynamically adjusting the initial alarm threshold based on the environmental parameter data to obtain a dynamically adjusted threshold, wherein the initial alarm threshold is dynamically adjusted based on a time - series depth collaborative analysis between the environmental temperature value and the environmental humidity value in the environmental parameter data; and a harmful gas concentration monitoring module for judging whether the harmful gas concentration is abnormal based on a comparison between the harmful gas concentration data and the dynamically adjusted threshold.

[0016] Compared with the prior art, the intelligent detection system and method for harmful gases of the shield machine provided by this application monitor the concentration of harmful gases and environmental parameters inside the shield machine in real time, and transmit the data to the warning terminal through the communication module. On this basis, AI technology is introduced to dynamically adjust the initial alarm threshold and make precise corrections according to environmental parameters (such as temperature and humidity), thereby improving the alarm accuracy. In addition, the forward LSTM model is used to analyze the environmental state coding features and calculate the threshold adjustment coefficient to achieve a more refined risk assessment. In this way, not only the intelligent level of harmful gas monitoring is improved, but also a foundation is laid for building a comprehensive safety protection system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 It is a schematic flow chart of the intelligent detection method for harmful gases of the shield machine according to the embodiment of the present application.

[0019] Figure 2 It is a schematic diagram of the data flow of the intelligent detection method for harmful gases of the shield machine according to the embodiment of the present application.

[0020] Figure 3 It is a schematic flow chart of step S3 in the intelligent detection method for harmful gases of the shield machine according to the embodiment of the present application.

[0021] Figure 4 It is a schematic flow chart of step S32 in the intelligent detection method for harmful gases of the shield machine according to the embodiment of the present application.

[0022] Figure 5 It is a schematic flow chart of step S33 in the intelligent detection method for harmful gases of the shield machine according to the embodiment of the present application.

[0023] Figure 6 It is a schematic flow chart of step S333 in the intelligent detection method for harmful gases of the shield machine according to the embodiment of the present application.

[0024] Figure 7 It is a schematic block diagram of the intelligent detection system for harmful gases of the shield machine according to the embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0026] Figure 1 It is a schematic flowchart of an intelligent detection method for harmful gases of a shield machine according to an embodiment of the present application. Figure 2 It is a schematic diagram of data flow of an intelligent detection method for harmful gases of a shield machine according to an embodiment of the present application. As Figure 1 and Figure 2 shown, the intelligent detection method for harmful gases of the shield machine includes: S1, real-time monitoring of the concentration of harmful gases in the shield machine and real-time monitoring of environmental parameters to obtain harmful gas concentration data and environmental parameter data; S2, transmitting the harmful gas concentration data and the environmental parameter data to a warning terminal through a communication module; S3, extracting an initial alarm threshold and dynamically adjusting the initial alarm threshold based on the environmental parameter data to obtain a dynamically adjusted threshold; S4, based on the comparison between the harmful gas concentration data and the dynamically adjusted threshold, determining whether the concentration of harmful gases is abnormal.

[0027] Specifically, in step S1, the concentration of harmful gases in the shield machine is real-time monitored and the environmental parameters are real-time monitored to obtain harmful gas concentration data and environmental parameter data. In one embodiment, the environmental parameters include an environmental temperature value and an environmental humidity value. It should be understood that various geological conditions may be encountered during the underground tunneling process of the shield machine, and these conditions may cause the leakage of toxic and harmful gases, such as hydrogen sulfide, methane, etc. These gases not only threaten the lives of construction workers but may also cause damage to equipment. Therefore, real-time monitoring of the concentration of these gases is the basis for ensuring construction safety. At the same time, changes in environmental parameters such as temperature and humidity will also indirectly affect the diffusion rate and concentration change of gases, further affecting the accuracy of safety assessment. For example, in a high-temperature and high-humidity environment, the diffusion rate of certain gases may increase, resulting in a rapid rise in the gas concentration in a local area, increasing the risk of explosion or poisoning. Therefore, by comprehensively analyzing the concentration of harmful gases and environmental parameters, the safety status of the current construction environment can be judged more accurately, potential risks can be discovered in time, and corresponding measures can be taken.

[0028] In a specific embodiment, a sensor network is deployed inside the shield machine. These sensors can accurately measure the concentration of different types of harmful gases and record parameters such as the temperature and humidity of the surrounding environment. For example, various types of detectors can be installed at key positions of the shield machine to detect oxygen, hydrogen sulfide, methane, carbon monoxide, nitrogen oxides, and dust, etc. At the same time, dedicated temperature and humidity sensors are equipped to capture changes in environmental parameters.

[0029] Specifically, in step S2, the harmful gas concentration data and the environmental parameter data are transmitted to the early warning terminal through the communication module. It should be understood that transmitting the data to the early warning terminal through the communication module can achieve centralized processing and analysis of the data, enabling the staff to timely obtain the safety status of the current construction environment and take corresponding measures according to the situation. Specifically, transmitting the harmful gas concentration data and the environmental parameter data to the early warning terminal enables the system to perform in-depth analysis using advanced algorithms to identify potential risks.

[0030] In a specific embodiment, the data collected by the sensor is first sent to the local data collection unit, which is responsible for preliminary data processing and integration. Then, the processed data is transmitted to the early warning terminal located in the shield machine control room through a dedicated communication network (such as ZigBee, LoRa, or Ethernet). In this process, to ensure the reliability and security of data transmission, encryption technology is usually used to protect the data from being tampered with or leaked. At the same time, considering that there may be signal interference problems in the underground construction environment, the communication module needs to have strong anti-interference capabilities.

[0031] Specifically, in step S3, the initial alarm threshold is extracted, and the initial alarm threshold is dynamically adjusted based on the environmental parameter data to obtain a dynamically adjusted threshold. It should be understood that the traditional fixed alarm threshold cannot adapt to the complex underground construction environment and is prone to false alarms or missed alarms. By dynamically adjusting the alarm threshold, more accurate judgments can be made according to the changes in the real-time monitored environmental parameters. For example, in a high-temperature and high-humidity environment, the diffusion rate of some harmful gases may increase, resulting in a rapid rise in the gas concentration in a local area. If the alarm is still triggered according to the fixed threshold, the best response time may be missed. By dynamically adjusting the threshold through the above method, not only can the accuracy of the alarm be improved, but also potential risks can be identified in a timely manner, and corresponding preventive measures can be taken to maximize the safety of construction workers. In addition, this method can also combine historical data and current data for prediction, helping managers better plan the construction progress, optimize resource allocation, and improve the overall work efficiency.

[0032] In an embodiment, such as Figure 3As shown, the initial alarm threshold is extracted, and the initial alarm threshold is dynamically adjusted based on the environmental parameter data to obtain a dynamically adjusted threshold, including: S31, sorting the environmental parameter data based on the time stamp to obtain a time queue of environmental temperature values and a time series of environmental humidity values; S32, performing sequence encoding on the time queue of environmental temperature values and the time series of environmental humidity values based on an environmental parameter monitoring time window to obtain a time queue of moving average values of environmental temperature and a time queue of moving average values of environmental humidity; S33, calculating a threshold adjustment coefficient based on the time queue of moving average values of environmental temperature and the time queue of moving average values of environmental humidity; S34, dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient to obtain the dynamically adjusted threshold.

[0033] Specifically, in step S31, the environmental parameter data is sorted based on the time stamp to obtain a time queue of environmental temperature values and a time series of environmental humidity values. It should be understood that the environmental parameter data needs to be preliminarily sorted first. For example, multiple sensor nodes are installed inside the shield machine, and these nodes can monitor the changes in environmental temperature and humidity in real time and mark the collected data with time stamps. The purpose of this is to ensure that each piece of data has clear time information, which is convenient for subsequent time series analysis. The environmental temperature values and environmental humidity values in the environmental parameter data are respectively arranged according to the time stamp to obtain a time queue of environmental temperature values and a time series of environmental humidity values.

[0034] Specifically, as Figure 4 shown, in step S32, performing sequence encoding on the time queue of environmental temperature values and the time series of environmental humidity values based on an environmental parameter monitoring time window to obtain a time queue of moving average values of environmental temperature and a time queue of moving average values of environmental humidity includes: S321, performing sequence segmentation on the time queue of environmental temperature values based on the environmental parameter monitoring time window to obtain a sequence of environmental temperature monitoring time windows; S322, calculating the mean value of environmental temperature values within each environmental temperature monitoring time window in the sequence of environmental temperature monitoring time windows to obtain the time queue of moving average values of environmental temperature; S323, performing sequence segmentation on the time queue of environmental humidity values based on the environmental parameter monitoring time window to obtain a sequence of environmental humidity monitoring time windows; S324, calculating the mean value of environmental humidity values within each environmental humidity monitoring time window in the sequence of environmental humidity monitoring time windows to obtain the time queue of moving average values of environmental humidity.

[0035] It should be understood that, based on a pre-set environmental parameter monitoring time window (for example, every 5 minutes as a time window, which can be adjusted according to the actual situation, and this embodiment does not make specific limitations), the time queues of the environmental temperature value and the humidity value are segmented, and the average value within each time window is calculated respectively, so as to form a time queue of the moving average of the environmental temperature and a time queue of the moving average of the environmental humidity. This process helps to eliminate the noise impact caused by short-term fluctuations, making the data smoother and more representative.

[0036] Specifically, as Figure 5 shown, in step S33, based on the time queue of the moving average of the environmental temperature and the time queue of the moving average of the environmental humidity, calculate a threshold adjustment coefficient, including: S331, perform sequence encoding on the time queue of the moving average of the environmental temperature based on a forward LSTM model to obtain environmental temperature time series fluctuation features; S332, perform sequence encoding on the time queue of the moving average of the environmental humidity based on a forward LSTM model to obtain environmental humidity time series fluctuation features; S333, perform environmental parameter time series collaborative encoding on the environmental temperature time series fluctuation features and the environmental humidity time series fluctuation features to obtain environmental state encoding features; S334, perform environmental compensation decoding based on a decoder on the environmental state encoding features to obtain the threshold adjustment coefficient.

[0037] In step S331 and step 332, a forward long short-term memory (LSTM) model is used to perform sequence encoding on the time queues of the moving average of the environmental temperature and the moving average of the environmental humidity. LSTM is a special recurrent neural network (RNN), which is particularly suitable for processing time series data and can capture long-term dependencies. By training the LSTM model, the time series fluctuation features of the environmental temperature and humidity can be extracted from the time queues. Specifically, input the time queue of the moving average of the environmental temperature into the LSTM model, and the output is the environmental temperature time series fluctuation feature; similarly, input the time queue of the moving average of the environmental humidity into another LSTM model, and the output is the environmental humidity time series fluctuation feature. These two features jointly reflect the trend and pattern of the environmental parameters changing over time.

[0038] In step S333, considering that the time series data of environmental temperature and humidity are processed by the forward LSTM model respectively, the independent temporal fluctuation characteristics can be captured. However, analyzing only a single environmental factor cannot comprehensively reflect the real environmental state and its impact on the diffusion of harmful gases. Therefore, it is necessary to further perform collaborative encoding on the temporal fluctuation characteristics of environmental temperature and humidity to construct an environmental state encoding feature that can comprehensively reflect the interaction relationship between the two. This approach takes into account the potential correlation between environmental temperature and humidity and their combined effect on the behavior of harmful gases, thereby making the threshold adjustment coefficient calculated based on this encoding feature more accurately reflect the complex dynamic changes in the actual construction environment. This helps to improve the sensitivity and accuracy of the alarm system, avoid false alarms or missed alarms caused by fixed thresholds, and thus provide more reliable protection for construction safety.

[0039] In one embodiment, as Figure 6 shown, performing environmental parameter temporal collaborative encoding on the environmental temperature temporal fluctuation feature and the environmental humidity temporal fluctuation feature to obtain an environmental state encoding feature includes: S3331, performing non-linear feature mapping reconstruction on the environmental temperature temporal fluctuation feature and the environmental humidity temporal fluctuation feature to obtain a set of environmental temperature local feature encoding vectors and a set of environmental humidity local feature encoding vectors; S3332, calculating a potential correlation feature joint encoding matrix between each corresponding environmental temperature local feature encoding vector and environmental humidity local feature encoding vector in the set of environmental temperature local feature encoding vectors and the set of environmental humidity local feature encoding vectors to obtain a set of environmental temperature - environmental humidity potential correlation feature joint encoding matrices; S3333, calculating the sparse dimension regulation weights of each environmental temperature - environmental humidity potential correlation feature joint encoding matrix in the set of environmental temperature - environmental humidity potential correlation feature joint encoding matrices to obtain a set of environmental temperature - environmental humidity sparse dimension regulation weights; S3334, based on the set of environmental temperature - environmental humidity sparse dimension regulation weights, performing dynamic weight fusion on the set of environmental temperature - environmental humidity potential correlation feature joint encoding matrices to obtain the environmental state encoding feature.

[0040] In step S3331, by adopting the method of reconstructing with non - linear feature mapping, the original temporal fluctuation features of environmental temperature and temporal fluctuation features of environmental humidity can be converted into a set of more abstract and information - rich local feature encoding vectors. This method is not simply extracting local features, but by simulating observations under different "delays", it excavates the deep - level structural information hidden in the time series. In specific operations, one - dimensional convolution plays a key role. It captures the change patterns of features at different time delays by implementing local perception in the form of a sliding window on the time - series data of environmental parameters. This way is similar to the idea of phase - space reconstruction in dynamic system theory, regarding the static feature vector as a potential trajectory in the abstract feature space, and then revealing the internal structural information and dimensional correlation of the features. Each convolution kernel is like a feature detector that can identify local features of specific patterns, and multiple groups of convolution kernels working in parallel provide the possibility of observing data from multiple perspectives, forming rich expressions of the inherent mode components of the signal under different patterns. Specifically, this process can be expressed by the formula: ; where and are the temporal fluctuation features of the environmental temperature and the temporal fluctuation features of the environmental humidity respectively, is the non - linear feature mapping reconstruction based on one - dimensional convolution encoding, and are the set of local feature encoding vectors of environmental temperature and the set of local feature encoding vectors of environmental humidity respectively, are the first, second, th, and th local feature encoding vectors of environmental temperature in the set of local feature encoding vectors of environmental temperature respectively, are the first, second, th, and th local feature encoding vectors of environmental humidity in the set of local feature encoding vectors of environmental humidity respectively.

[0041] In step S3332, by calculating the potential correlation feature joint encoding matrix between each group of corresponding environmental temperature local feature encoding vectors and environmental humidity local feature encoding vectors, a set of environmental temperature - environmental humidity potential correlation feature joint encoding matrices can be constructed, thereby revealing the complex interaction patterns between the two. Specifically, after feature encoding, a specific algorithm is used to explore the potential correlation between the environmental temperature and humidity local feature encoding vectors. Here, each potential correlation feature joint encoding matrix serves as a bridge connecting two seemingly independent but actually interacting feature sets, and on this basis, reveals the deep - level interaction relationship between them. This processing method goes beyond the traditional single - variable analysis method, can capture the subtle changes generated by the combined action of environmental temperature and humidity, and thus provides more accurate data support for adjusting the alarm threshold. In the process of constructing this potential correlation feature space, the predefined fixed interaction pattern is abandoned, and instead, a data - driven approach is adopted to learn the most effective interaction pattern from the actual monitoring data. This approach gives the model greater flexibility and adaptability, enabling it to automatically adjust parameters according to different environmental conditions and be more in line with the actual situation. For example, in a high - temperature and high - humidity environment, the diffusion rate of some harmful gases may increase significantly, and through this method, these changes can be more accurately identified and quantified, thereby improving the response speed and accuracy of the early warning system.

[0042] In one embodiment, calculating the potential correlation feature joint encoding matrix between each group of corresponding environmental temperature local feature encoding vectors and environmental humidity local feature encoding vectors in the set of environmental temperature local feature encoding vectors and the set of environmental humidity local feature encoding vectors to obtain a set of environmental temperature - environmental humidity potential correlation feature joint encoding matrices includes: performing correlation encoding on each group of corresponding environmental temperature local feature encoding vectors and environmental humidity local feature encoding vectors in the set of environmental temperature local feature encoding vectors and the set of environmental humidity local feature encoding vectors to obtain a set of initial environmental temperature - environmental humidity potential correlation feature joint encoding matrices; constructing a non - linear solution matrix based on each group of corresponding environmental temperature local feature encoding vectors and environmental humidity local feature encoding vectors; calculating the local linear norm - preserving constraint matrix of the initial environmental temperature - environmental humidity potential correlation feature joint encoding matrix; and performing projection smoothing optimization on the initial environmental temperature - environmental humidity potential correlation feature joint encoding matrix based on the non - linear solution matrix and the local linear norm - preserving constraint matrix to obtain the set of environmental temperature - environmental humidity potential correlation feature joint encoding matrices.

[0043] Specifically, the following formula is used to perform associated coding on each pair of corresponding environmental temperature local feature coding vectors and environmental humidity local feature coding vectors in the set of environmental temperature local feature coding vectors and the set of environmental humidity local feature coding vectors to obtain a set of initial environmental temperature - environmental humidity potential associated feature joint coding matrices, where the formula is: ; where and are respectively the trainable weight matrix of the th environmental humidity local feature coding vector and the trainable weight matrix of the th environmental temperature local feature coding vector, is the scale of the trainable weight matrix of the th environmental temperature local feature coding vector, that is, width times height, are the respective initial environmental temperature - environmental humidity potential associated feature joint coding matrices. In a general embodiment, the initial environmental temperature - environmental humidity potential associated feature joint coding matrix can be used as the environmental temperature - environmental humidity potential associated feature joint coding matrix in the next step.

[0044] In a preferred embodiment, in order to further enhance the trivialization of the implicit coding projection interaction from the non - linear feature mapping reconstruction space to the potential associated feature space, it is desired to impose a smoothness propagation constraint on the local non - linear connection pattern caused by the construction transformation of the non - linear feature mapping reconstruction space. Therefore, first, for the non - linear feature mapping reconstruction space, for the eigenvectors of the non - weight substantial coding source domain vectors used as the initial environmental temperature - environmental humidity potential associated feature joint coding matrix and eigenvector , based on the th eigenvalue in eigenvector and the th eigenvalue in eigenvector , a non - linear solution matrix is obtained through local linear transformation constraints, expressed as: ; where represents the natural constant, represents the value at the position of the non - linear solution matrix obtained by local linear transformation constraints.

[0045] Then, the local linear norm - preserving constraint matrix in the case of the spatial state dynamic mapping transformation of the potential associated feature joint coding matrix is obtained, that is, the local linear norm - preserving constraint matrix Satisfy the following relationship: ; where represents matrix multiplication, represents the transpose of a matrix.

[0046] In this way, the non-linear solution matrix and the local linear norm-preserving constraint matrix can be used to optimize the initial environmental temperature - environmental humidity potential association feature joint encoding matrix, expressed as: ; where represents addition by position, represents the environmental temperature - environmental humidity potential association feature joint encoding matrix.

[0047] Thus, while realizing spatial state transfer, the trivial association mechanism of the non-linear solution matrix is used for projection smoothing, especially reducing the propagation complexity, thereby avoiding redundant complex coupling with non-linear mapping and interaction fusion in the potential implicit interaction representation of the initial environmental temperature - environmental humidity potential association feature joint encoding matrix, and promoting efficient learning of effective feature interaction.

[0048] That is, in this preferred embodiment, calculating the potential association feature joint encoding matrix between each pair of corresponding environmental temperature local feature encoding vectors and environmental humidity local feature encoding vectors in the set of environmental temperature local feature encoding vectors and the set of environmental humidity local feature encoding vectors to obtain a set of environmental temperature - environmental humidity potential association feature joint encoding matrices, including: performing association encoding on each pair of corresponding environmental temperature local feature encoding vectors and environmental humidity local feature encoding vectors in the set of environmental temperature local feature encoding vectors and the set of environmental humidity local feature encoding vectors to obtain a set of initial environmental temperature - environmental humidity potential association feature joint encoding matrices; constructing a non-linear solution matrix based on each pair of corresponding environmental temperature local feature encoding vectors and environmental humidity local feature encoding vectors; calculating the local linear norm-preserving constraint matrix of the initial environmental temperature - environmental humidity potential association feature joint encoding matrix; and performing projection smoothing optimization on the initial environmental temperature - environmental humidity potential association feature joint encoding matrix based on the non-linear solution matrix and the local linear norm-preserving constraint matrix to obtain the set of environmental temperature - environmental humidity potential association feature joint encoding matrices.

[0049] In step S3333, calculating the sparse dimension regulation weights of each environmental temperature-environmental humidity potential correlation feature joint coding matrix in the set of environmental temperature-environmental humidity potential correlation feature joint coding matrices to obtain a set of environmental temperature-environmental humidity sparse dimension regulation weights includes: calculating the square of the Frobenius norm of the environmental temperature-environmental humidity potential correlation feature joint coding matrix to obtain an environmental temperature-environmental humidity energy scaling factor; processing the environmental temperature-environmental humidity energy scaling factor through an activation function to obtain an environmental temperature-environmental humidity sparse dimension regulation weight. Specifically, this process can be expressed by the formula: ; where is the square of the two-norm of the matrix, is each environmental temperature-environmental humidity sparse dimension regulation factor, is the normalization exponential function, is each environmental temperature-environmental humidity sparse dimension regulation weight.

[0050] It should be understood that calculating the sparse dimension regulation weights of each environmental temperature-environmental humidity potential correlation feature joint coding matrix in the set of environmental temperature-environmental humidity potential correlation feature joint coding matrices can determine which interactions between feature dimensions are the most important for the overall analysis. This step is essentially to impose a spatial sparsity constraint on the environmental temperature-environmental humidity potential correlation feature joint coding matrix to ensure that only those most representative and significant interaction patterns are retained, while other redundant or noisy information is suppressed. This processing method enables the model to focus on the most important part among numerous possible interaction relationships, thereby improving its selectivity and robustness. Specifically, in this process, first calculate the square of the Frobenius norm of the environmental temperature-environmental humidity potential correlation feature joint coding matrix, and use this as the energy scaling factor to measure the sparsity of each matrix. Then, use the activation function to process these energy scaling factors to generate environmental temperature-environmental humidity sparse dimension regulation weights.

[0051] In step S3334, by using the environmental temperature-environmental humidity sparse dimension regulation weight as a guide, the system can dynamically adjust its contribution in the final feature representation according to the importance and quality of each environmental temperature-environmental humidity potential association feature joint encoding matrix. This method goes beyond simple averaging or concatenation operations, achieving adaptive aggregation and ensuring the optimal fusion of information. This dynamic weight fusion process essentially ranks and filters the importance of information from different local perspectives, enabling those matrices containing the most critical interaction information to dominate in the final feature vector. For example, in a high-temperature and high-humidity environment, certain specific temperature and humidity combinations may significantly affect the diffusion rate and concentration change of harmful gases. Through dynamic weight fusion, these key factors can be accurately identified and highlighted, providing strong support for subsequent risk assessment. Specifically, this process can be expressed by the formula: ; where is the environmental temperature-environmental humidity semantic interaction fusion matrix, represents feature reshaping, is the environmental state encoded feature.

[0052] In step S334, since factors such as temperature and humidity in the underground construction environment will significantly affect the diffusion rate and concentration change of harmful gases, fixed alarm thresholds may not be able to adapt to this variability, increasing the risk of false alarms or missed alarms. Therefore, by introducing an environmental compensation decoding mechanism, precise correction of these factors can be achieved, improving the sensitivity and accuracy of early warning. Specifically, a decoder network is used to decode the environmental state encoded feature, aiming to map these high-dimensional abstract features back to the threshold adjustment coefficient related to the actual application scenario. This decoder usually consists of several layers of neural networks, and each layer contains a weight matrix and an activation function, which are used to gradually learn how to extract useful threshold adjustment information from the complex environmental state encoded features. Here, it should be known that the weight matrix and bias in the deep learning network are learned from the data through the training process, and the specific training process can be carried out by means of gradient descent and backpropagation.

[0053] Specifically, in step S34, the initial alarm threshold is dynamically adjusted based on the threshold adjustment coefficient to obtain the dynamically adjusted threshold. It should be understood that through the decoder, environmental compensation decoding of the environmental state encoded feature is performed to obtain the threshold adjustment coefficient. This coefficient reflects the deviation degree under the current environmental conditions relative to the standard conditions and can be used to dynamically adjust the initial alarm threshold.

[0054] In one embodiment, dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient to obtain the dynamically adjusted threshold includes: dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient according to the following formula: ; where is the initial alarm threshold, is the threshold adjustment coefficient, is the dynamically adjusted threshold.

[0055] Specifically, in step S4, based on the comparison between the harmful gas concentration data and the dynamically adjusted threshold, it is determined whether the harmful gas concentration is abnormal. Specifically, the latest measured harmful gas concentration data is compared with the adjusted threshold. Once it is found that the concentration of any gas exceeds its corresponding dynamically adjusted threshold, a warning signal is immediately sent to notify relevant personnel to take measures to ensure the safety of the construction site. This method not only improves the intelligent level of harmful gas monitoring but also provides strong support for building a comprehensive safety protection system.

[0056] In summary, the intelligent detection method for harmful gases of a shield machine provided in this application has been clarified. It monitors the harmful gas concentration and environmental parameters inside the shield machine in real time and uses a communication module to transmit the data to a warning terminal. On this basis, AI technology is introduced to dynamically adjust the initial alarm threshold, which is accurately corrected according to environmental parameters (such as temperature and humidity), thereby improving the alarm accuracy. In addition, a forward LSTM model is used to analyze the environmental state coding features and calculate the threshold adjustment coefficient to achieve a more refined risk assessment. In this way, not only is the intelligent level of harmful gas monitoring improved, but also a foundation is laid for building a comprehensive safety protection system.

[0057] This application also provides an intelligent detection system for harmful gases of a shield machine, as Figure 7 shown. The intelligent detection system 700 for harmful gases of a shield machine includes: a shield machine data acquisition module 710, configured to monitor the harmful gas concentration inside the shield machine in real time and monitor the environmental parameters in real time to obtain harmful gas concentration data and environmental parameter data; a shield machine data transmission module 720, configured to transmit the harmful gas concentration data and the environmental parameter data to a warning terminal through a communication module; a dynamically adjusted threshold determination module 730, configured to extract the initial alarm threshold and dynamically adjust the initial alarm threshold based on the environmental parameter data to obtain a dynamically adjusted threshold, wherein the initial alarm threshold is dynamically adjusted based on the time-series depth collaborative analysis between the environmental temperature value and the environmental humidity value in the environmental parameter data; a harmful gas concentration monitoring module 740, configured to determine whether the harmful gas concentration is abnormal based on the comparison between the harmful gas concentration data and the dynamically adjusted threshold.

[0058] An embodiment of the present application also provides a computer-readable storage medium, in which computer program code is stored. When the computer program code runs on a computer, the computer is caused to execute the above-related method steps to implement an intelligent detection method for harmful gases of a shield machine provided in the above embodiment.

[0059] An embodiment of the present application also provides a computer program product. When the computer program product runs on a computer, the computer is caused to execute the above-related steps to implement an intelligent detection method for harmful gases of a shield machine provided in the above embodiment.

[0060] Among them, the system, computer-readable storage medium or computer program product provided in the embodiments of the present application are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be elaborated here.

[0061] It should be noted that the above sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments.

[0062] The processes depicted in the accompanying drawings do not necessarily require the particular order shown or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous. Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

Claims

1. A shield machine harmful gas intelligent detection method, characterized in that: include: Monitor the concentration of harmful gases in the shield machine in real time and monitor the environmental parameters in real time to obtain harmful gas concentration data and environmental parameter data; transmit the harmful gas concentration data and the environmental parameter data to the early warning terminal through the communication module; extract the initial alarm threshold, and dynamically adjust the initial alarm threshold based on the environmental parameter data to obtain a dynamic adjustment threshold, wherein the initial alarm threshold is dynamically adjusted based on the time series deep collaborative analysis between the ambient temperature value and the ambient humidity value in the environmental parameter data; based on the comparison between the harmful gas concentration data and the dynamic adjustment threshold, determine whether the harmful gas concentration is abnormal.

2. The shield machine harmful gas intelligent detection method according to claim 1 is characterized in that: The environmental parameters include an environmental temperature value and an environmental humidity value.

3. The shield machine harmful gas intelligent detection method according to claim 2 is characterized in that: Extracting an initial alarm threshold, and dynamically adjusting the initial alarm threshold based on the environmental parameter data to obtain a dynamically adjusted threshold, including: sorting the environmental parameter data based on a timestamp to obtain a time queue of ambient temperature values ​​and a time series of ambient humidity values; based on an environmental parameter monitoring time window, serially encoding the time queue of ambient temperature values ​​and the time series of ambient humidity values ​​to obtain a time queue of ambient temperature moving average values ​​and a time queue of ambient humidity moving average values; calculating a threshold adjustment coefficient based on the time queue of ambient temperature moving average values ​​and the time queue of ambient humidity moving average values; and dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient to obtain the dynamically adjusted threshold.

4. The shield machine harmful gas intelligent detection method according to claim 3 is characterized in that: Based on the environmental parameter monitoring time window, the time queue of the ambient temperature values ​​and the time series of the ambient humidity values ​​are sequence-encoded to obtain a time queue of an ambient temperature moving average value and a time queue of an ambient humidity moving average value, including: based on the environmental parameter monitoring time window, the time queue of the ambient temperature values ​​is sequence-segmented to obtain a sequence of ambient temperature monitoring time windows; the mean of the ambient temperature values ​​in each ambient temperature monitoring time window in the sequence of the ambient temperature monitoring time windows is calculated to obtain a time queue of the ambient temperature moving average value; based on the environmental parameter monitoring time window, the time queue of the ambient humidity values ​​is sequence-segmented to obtain a sequence of ambient humidity monitoring time windows; the mean of the ambient humidity values ​​in each ambient humidity monitoring time window in the sequence of the ambient humidity monitoring time windows is calculated to obtain a time queue of the ambient humidity moving average value.

5. The shield machine harmful gas intelligent detection method according to claim 4 is characterized in that: Based on the time queue of the ambient temperature moving average and the time queue of the ambient humidity moving average, a threshold adjustment coefficient is calculated, including: performing sequence encoding based on a forward LSTM model on the time queue of the ambient temperature moving average to obtain the ambient temperature time series fluctuation characteristics; performing sequence encoding based on a forward LSTM model on the time queue of the ambient humidity moving average to obtain the ambient humidity time series fluctuation characteristics; performing environmental parameter time series collaborative encoding on the ambient temperature time series fluctuation characteristics and the ambient humidity time series fluctuation characteristics to obtain environmental state coding characteristics; performing decoder-based environmental compensation decoding on the environmental state coding characteristics to obtain the threshold adjustment coefficient.

6. The shield machine harmful gas intelligent detection method according to claim 5 is characterized in that: Dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient to obtain the dynamically adjusted threshold includes: dynamically adjusting the initial alarm threshold based on the threshold adjustment coefficient using the following formula, wherein the formula is: ;in, is the initial alarm threshold, is the threshold adjustment coefficient, To dynamically adjust the threshold.

7. The shield machine harmful gas intelligent detection method according to claim 6 is characterized in that: The environmental parameter time series collaborative encoding is performed on the ambient temperature time series fluctuation characteristics and the ambient humidity time series fluctuation characteristics to obtain the environmental state coding characteristics, including: performing nonlinear feature mapping reconstruction on the ambient temperature time series fluctuation characteristics and the ambient humidity time series fluctuation characteristics to obtain a set of ambient temperature local feature coding vectors and a set of ambient humidity local feature coding vectors; calculating the potential correlation feature joint coding matrix between each group of corresponding ambient temperature local feature coding vectors and ambient humidity local feature coding vectors in the set of ambient temperature local feature coding vectors and the set of ambient humidity local feature coding vectors to obtain a set of ambient temperature-ambient humidity potential correlation feature joint coding matrices; calculating the sparse dimension control weights of each ambient temperature-ambient humidity potential correlation feature joint coding matrix in the set of ambient temperature-ambient humidity potential correlation feature joint coding matrix to obtain a set of ambient temperature-ambient humidity sparse dimension control weights; based on the set of ambient temperature-ambient humidity sparse dimension control weights, dynamically weight fusion is performed on the set of ambient temperature-ambient humidity potential correlation feature joint coding matrices to obtain the environmental state coding characteristics.

8. The shield machine harmful gas intelligent detection method according to claim 7 is characterized in that: The method comprises: performing association coding on each group of corresponding local feature coding vectors of ambient temperature and local feature coding vectors of ambient humidity in the set of local feature coding vectors of ambient temperature and the set of local feature coding vectors of ambient humidity to obtain a set of joint coding matrices of potential association features of ambient temperature and ambient humidity, including: performing association coding on each group of corresponding local feature coding vectors of ambient temperature and local feature coding vectors of ambient humidity in the set of local feature coding vectors of ambient temperature and the set of local feature coding vectors of ambient humidity to obtain a set of joint coding matrices of potential association features of initial ambient temperature and ambient humidity; constructing a nonlinear solution matrix based on each group of corresponding local feature coding vectors of ambient temperature and local feature coding vectors of ambient humidity; calculating a local linear norm-preserving constraint matrix of the initial joint coding matrix of potential association features of ambient temperature and ambient humidity; and performing projection smoothing optimization on the initial joint coding matrix of potential association features of ambient temperature and ambient humidity based on the nonlinear solution matrix and the local linear norm-preserving constraint matrix to obtain a set of joint coding matrices of potential association features of ambient temperature and ambient humidity.

9. The shield machine harmful gas intelligent detection method according to claim 8 is characterized in that: Calculate the sparse dimension control weights of each ambient temperature-ambient humidity potential correlation feature joint coding matrix in the set of ambient temperature-ambient humidity potential correlation feature joint coding matrices to obtain a set of ambient temperature-ambient humidity sparse dimension control weights, including: calculating the square of the Frobenius norm of the ambient temperature-ambient humidity potential correlation feature joint coding matrix to obtain an ambient temperature-ambient humidity energy scaling factor; and processing the ambient temperature-ambient humidity energy scaling factor through an activation function to obtain an ambient temperature-ambient humidity sparse dimension control weight.

10. An intelligent detection system for harmful gases in a shield machine, used to execute the intelligent detection method for harmful gases in a shield machine according to any one of claims 1 to 9, characterized in that: include: A shield machine data acquisition module is used to monitor the concentration of harmful gases in the shield machine in real time and to monitor environmental parameters in real time to obtain harmful gas concentration data and environmental parameter data; a shield machine data transmission module is used to transmit the harmful gas concentration data and the environmental parameter data to the early warning terminal through the communication module; a dynamic adjustment threshold determination module is used to extract the initial alarm threshold and dynamically adjust the initial alarm threshold based on the environmental parameter data to obtain a dynamic adjustment threshold, wherein the initial alarm threshold is dynamically adjusted based on the time series deep collaborative analysis between the ambient temperature value and the ambient humidity value in the environmental parameter data; a harmful gas concentration monitoring module is used to determine whether the harmful gas concentration is abnormal based on the comparison between the harmful gas concentration data and the dynamic adjustment threshold.

Images