A gas state evaluation and risk discrimination method based on pressure time sequence

Abstract

The application discloses a kind of gas state evaluation and risk discrimination method based on pressure timing, comprising: using single pressure sensor to collect pressure value, obtains pressure timing data according to different sampling frequency;Continuous gradient analysis pressure timing, generate four kinds of phase fragments and two-dimensional phase characteristics;Pressure phase hypergraph is constructed, and the hypergraph path formed by hypernode and hyperedge is formed;Build and update operating condition template library, and generate two-dimensional phase residual signal by matching current hypergraph path;Based on two-dimensional phase residual, build hierarchical state machine, drive state migration and output multi-state gas state;According to multi-state gas state and residual signal, generate risk level, execute valve and record event.The application collects pressure timing data by single pressure sensor, combines two-dimensional phase characteristic analysis and multi-layer state machine modeling, realizes the accurate identification and risk discrimination of multi-state gas pipeline.

Description

Technical Field

[0001] This invention relates to the field of gas safety monitoring technology, and in particular to a method for gas status assessment and risk identification based on pressure time series. Background Technology

[0002] With the continuous growth of gas usage in residential and commercial sectors, gas pipeline safety monitoring has become a crucial issue that must be addressed in gas applications. Existing gas safety valves generally employ single-judgment methods such as pressure threshold triggering, flow threshold triggering, or gas concentration triggering. In terms of hardware structure, they typically rely on a single pressure sensor or a combination of multiple pressure / concentration sensors, determining the gas status through fixed thresholds or simple state conditions. However, these single-point numerical judgment-based solutions can only identify abnormal pressure changes exceeding the threshold, failing to accurately distinguish between ignition transient pressure drops, steady-state combustion fluctuations, backpressure changes after ignition shutdown, and complex situations involving various types of leaks, easily leading to false alarms or missed alarms. Especially under single-sensor conditions, traditional methods cannot extract effective morphological information from the structural trends of pressure changes, thus limiting their state recognition capabilities.

[0003] Existing technologies typically rely on a single pressure change amplitude for one-dimensional judgment, using whether the pressure exceeds a fixed threshold as the leakage trigger condition. These methods fail to consider the coupling relationship between pressure and time, nor do they analyze the continuous morphological characteristics of pressure changes. They cannot identify the rapid pressure drop-rebound characteristic curve during ignition, nor can they distinguish the slow downward trend in micro-leak scenarios from the slight fluctuations in normal combustion. Because fixed thresholds are difficult to adapt to different users' appliance types, pipe diameters, pipe lengths, and usage environments, existing methods often exhibit instability and lack adaptability in various household environments.

[0004] Existing gas safety monitoring methods generally lack structured processing mechanisms for pressure time-series data, and have not established pressure behavior models that can express the relationship between phase structure and amplitude changes, thus lacking the ability to reason about complex operating conditions. Traditional technologies do not have an analysis method that divides pressure time series into multiple phase segments, nor have they formed a graph structure or hypergraph structure to describe pressure behavior sequences, and they do not combine trend, phase combination, and residual signals to drive multi-level state machines to perform state transitions.

[0005] Therefore, how to provide a gas state assessment and risk identification method based on pressure time sequence is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a gas state assessment and risk identification method based on pressure time series. This invention fully utilizes the pressure time series acquisition capability of a single pressure sensor, and identifies various gas states during use, such as ignition, combustion, ignition shutdown, and micro-leakage and large-leakage, by constructing two-dimensional phase features, a pressure phase hypergraph structure, and a three-layer hierarchical state machine model. This invention achieves dynamic migration of gas states and risk level judgment by structurally modeling the pressure change trend over time, combining phase duration with pressure amplitude changes to form two-dimensional phase features, constructing a pressure phase hypergraph using multi-phase combination relationships, and then combining this with a hierarchical state machine driven by phase structure, phase trend, and residual changes. This invention can accurately identify complex gas conditions without adding any additional sensors, and has the advantages of high identification accuracy, strong dynamic adaptability, good adaptability to different household piping environments, and high reliability in risk identification.

[0007] A method for assessing the state of gas and determining risk based on pressure time sequence according to an embodiment of the present invention includes:

[0008] The pressure value of the gas pipeline is collected using a single pressure sensor, and different sampling frequencies are set to obtain time-series data of pressure changes over time.

[0009] Continuous gradient analysis is performed on the pressure time series data in the time dimension. The pressure time series is divided into four phase segments according to the direction and rate of pressure change. Two-dimensional phase features consisting of phase duration and pressure amplitude change are extracted for each phase segment.

[0010] A pressure phase hypergraph is constructed based on multiple two-dimensional phase features. Each phase segment is used as a hypernode, and the sequential relationship between phases and the amplitude difference between phases are used as hyperedges. The current pressure time series is mapped as a hypergraph path composed of hypernodes and hyperedges.

[0011] A working condition template library is constructed using the pressure phase hypergraphs of normal ignition, normal combustion, and normal ignition shutdown processes. The working condition template library is updated online to form an adaptive baseline. The current pressure phase hypergraph path is structurally matched with the working condition template library to generate a two-dimensional phase residual signal.

[0012] A hierarchical state machine model is constructed based on two-dimensional phase residual signals. The valve state machine and the gas state machine are independent levels. The hierarchical state machine model is driven to perform state transitions by using the two-dimensional phase characteristic change trend, phase combination structure and two-dimensional phase residual signals, and outputs multi-state gas state.

[0013] Risk levels are generated based on the multi-state gas state and two-dimensional phase residual signals. When the risk level reaches the preset trigger condition, valve closing action is performed. The two-dimensional phase characteristics, pressure phase hypergraph path, residual signal and risk level are written into the historical event buffer to form a traceable event record.

[0014] Optionally, the step of setting different sampling frequencies to obtain pressure time-series data that changes over time includes: setting a pressure sampling frequency of no more than 1 Hz for the idle state, a pressure sampling frequency of no less than 20 Hz for the ignition detection state, a pressure sampling frequency of 3 Hz to 10 Hz for the steady-state combustion state, and a pressure sampling frequency of no less than 12 Hz for the suspected abnormal state, and continuously acquiring pressure sampling values ​​at the sampling frequency, and obtaining pressure time-series data that changes over time based on the pressure sampling values.

[0015] Optionally, the extraction of two-dimensional phase features consisting of phase duration and pressure amplitude changes for each phase segment includes:

[0016] Each pressure sampling point is read sequentially according to the sampling time order, and the sampling time information and corresponding sampling frequency information are associated with each pressure sampling point to form a pressure sampling sequence.

[0017] For each pressure sampling point in the pressure sampling sequence except for the first sampling point, calculate the pressure change between the pressure sampling point and the previous sampling point, as well as the rate of change of the pressure change over the actual time interval. Use the direction and rate of change of pressure change as the local change features of the pressure sampling point.

[0018] Based on the local variation characteristics of each pressure sampling point, the pressure sampling points are classified and labeled:

[0019] Based on the local change characteristics of each pressure sampling point, sampling points whose pressure change direction is decreasing and whose change rate is greater than the first gradient boundary parameter are marked as fast descent candidate points;

[0020] Sampling points where the pressure change direction is upward and the rate of change is greater than the second gradient boundary parameter are marked as candidate points for rapid recovery.

[0021] Sampling points whose absolute rate of change in the direction of pressure change is less than the third gradient boundary parameter are marked as steady-state candidate points;

[0022] Sampling points where the pressure changes in a downward direction and the rate of change is between the first gradient boundary parameter and the third gradient boundary parameter are marked as slow-decreasing candidate points.

[0023] In each type of candidate point, adjacent candidate points of the same type are merged into a continuous phase segment according to the time sequence. Short phase segments with a duration less than the preset minimum duration are merged or eliminated. The start and end positions of the phase segments are corrected according to the preset timing constraints of fast falling phase, fast rising phase, steady-state phase and slow falling phase.

[0024] For each phase segment obtained after correction, the difference between the start time and the end time of the phase segment is determined as the phase duration, and the difference between the start pressure value and the end pressure value of the phase segment is determined as the pressure amplitude change. The phase duration and the pressure amplitude change are combined to form a two-dimensional phase feature.

[0025] Optionally, the construction of the pressure phase hypermap based on multiple two-dimensional phase features includes:

[0026] For each phase segment, the phase type, phase duration, and pressure amplitude change are read separately. The phase duration and pressure amplitude change are combined to form a two-dimensional phase feature set, while maintaining the temporal order of the two-dimensional phase features and the corresponding phase segments.

[0027] Based on the numerical range of the duration and pressure amplitude changes of each phase in the two-dimensional phase feature set, the phase features are clustered according to the phase type and the similarity of the two-dimensional features. Multiple phase segments belonging to the same cluster result are grouped into the same phase cluster. Each phase cluster is abstracted into a super node, and each super node carries the representative phase type, representative phase duration and representative pressure amplitude change of its respective phase cluster.

[0028] On the time axis, according to the actual occurrence order of phase segments, adjacent phase combinations containing multiple phase segments are selected sequentially in a sliding time window manner. A hyperedge connection relationship is established between multiple supernodes that appear successively in the same time window and whose representative pressure amplitude changes meet the preset amplitude difference conditions. The time sequence relationship and pressure amplitude difference are used as constraints for establishing the hyperedge.

[0029] The hyperedge connections obtained within all time windows are merged, and hyperedges belonging to the same type of phase combination pattern are summarized as pattern hyperedges. The temporal sequence of phase segments, the phase type sequence, and the combination pattern of phase duration and pressure amplitude change are used as the attribute information of the pattern hyperedges, forming a pressure phase hypergraph structure composed of multiple hypernodes and multiple pattern hyperedges.

[0030] According to the time sequence of a single gas usage process, the supernodes involved in the usage process are connected in series according to their connection relationship in the pressure phase supergraph to form a supergraph path. The supergraph path serves as a structured representation of the pressure time sequence in the gas usage process at the phase and amplitude levels.

[0031] Optionally, generating the two-dimensional phase residual signal includes:

[0032] From the historical event buffer, only gas usage processes that are judged to be normal ignition conditions, normal combustion conditions, and normal flameout conditions and continuously meet the preset acceptance criteria are selected, and the corresponding pressure phase hypergraph paths are used as candidate condition samples.

[0033] The pressure phase hypergraph path of the candidate working condition sample is standardized, including phase type alignment, interval labeling of phase duration, interval labeling of pressure amplitude change and time sequence consistency check, to generate template unit;

[0034] Based on the template unit, a working condition template library is constructed. Global templates and user-level sub-templates are established for ignition, combustion, and shutdown working conditions, respectively. Each template consists of a set of supernode representatives and a set of mode superedges. A phase topology index and an amplitude range index are established.

[0035] The working condition template library is updated online. When a newly generated normal working condition sample passes the acceptance criteria, the corresponding supernode representative set is merged into the supernode representative set of the same type of template. The count information and the most recent update time of the mode superedge set are updated. The value range of the phase duration interval and pressure amplitude interval are corrected according to the preset weight decay strategy. When an abnormal sample is triggered in the working condition template, the template contamination protection strategy is activated and the acceptance and merging operation of the corresponding working condition template is suspended.

[0036] During the current gas usage process, after the corresponding pressure phase hypergraph path is generated, a working condition template consistent with the current working condition type is selected from the working condition template library. First, a phase topology index-driven structural matching is performed to determine the phase type sequence correspondence. Then, an amplitude interval index-driven refined matching is performed to determine the deviation between the pressure amplitude interval and the phase duration interval of each phase. The two-dimensional phase residual signal sequence generated by each phase segment is output in chronological order as the two-dimensional phase residual signal of the current gas usage process.

[0037] Optionally, the step of correcting the value range of the phase duration interval and the pressure amplitude interval according to the preset weighted attenuation strategy includes:

[0038] A first weight is set to characterize the influence of the new sample, and the value of the first weight is between 0.6 and 0.8;

[0039] A second weight is set to characterize the retention degree of existing sample intervals in the working condition template library. The value of the second weight is complementary to the first weight, that is, the second weight is equal to 1 minus the first weight.

[0040] Multiply the phase duration and pressure amplitude changes corresponding to the normal operating condition sample by the first weight, multiply the existing phase duration interval and pressure amplitude interval in the operating condition template library by the second weight, and add the two together as the upper and lower boundaries of the updated interval range.

[0041] Optionally, the output multi-state gas state includes:

[0042] A three-layer hierarchical state machine model is established, with the valve state machine, gas state machine, and risk state machine as the upper, middle, and lower layers, respectively. The valve state machine includes open, closed, and fault states; the gas state machine includes idle, ignition, combustion, flameout, micro-leakage, and large-leakage states; and the risk state machine includes normal, attention, warning, and emergency states. A state combination table in which the three states can occur simultaneously is recorded in the hierarchical state machine model.

[0043] During each round of state update, the control command of the valve actuator and the valve position detection result are read. The current state of the valve state machine is updated according to the correspondence between the control command and the detection result. The updated current state of the valve state machine is written as the upper-level state into the corresponding record in the state combination table.

[0044] Read the phase type change, phase duration change and pressure amplitude change in the pressure phase hypergraph path and two-dimensional phase residual signal. Generate candidate states of the gas state machine based on the state transition conditions recorded in the gas state machine and the current state of the gas state machine at the previous moment. Select the target state of the gas state machine from the candidate states according to the state combination table record that is consistent with the current valve state machine state.

[0045] The current state of the valve state machine, the target state of the gas state machine, and the two-dimensional phase residual signal are input into the risk state machine. The target state of the risk state machine is determined according to the state determination rules recorded in the risk state machine. When the target state of the risk state machine is an emergency state, a valve closing control command is generated and sent to the valve actuator. The target state of the gas state machine is adjusted to a state related to the termination of gas use.

[0046] After completing the current state update of the valve state machine, gas state machine, and risk state machine, the current state of the three state machines is recorded, and the current state of the gas state machine is output as the polymorphic gas state.

[0047] Optionally, the generation of risk levels based on multi-state gas conditions and two-dimensional phase residual signals includes:

[0048] Read the polymorphic gas state, two-dimensional phase residual signal, and current state of the risk state machine obtained from the current round of state update, associate the polymorphic gas state with the two-dimensional phase residual signal, and generate risk assessment input data for the current round of gas usage process;

[0049] Based on the type of polymorphic gas state, the magnitude of the two-dimensional phase residual signal, and the duration of the residual signal in time in the risk assessment input data, the corresponding risk index value is calculated and compared with the risk level classification conditions recorded in the control unit to obtain the risk level code.

[0050] The risk level code is compared with the valve closing trigger condition recorded in the control unit. When the risk level code meets the valve closing trigger condition, a valve closing control command is generated in the control unit and output to the valve actuator. The time when the risk level code meets the trigger condition in this round is recorded.

[0051] When the risk level code does not meet the valve closing trigger condition, the current valve opening and closing state remains unchanged. The risk level code is associated with the corresponding polymorphic gas state and two-dimensional phase residual signal, and output as the risk assessment result of this round to the upper monitoring or alarm module.

[0052] After completing the risk level calculation and valve closing command judgment, the two-dimensional phase characteristics, pressure phase hypergraph path, two-dimensional phase residual signal, multi-state gas state, current state of risk state machine, risk level code, and valve closing command execution results generated during this round of gas use are written into the historical event buffer in chronological order to form event record entries.

[0053] The beneficial effects of this invention are:

[0054] This invention upgrades the traditional one-dimensional judgment method, which relies solely on pressure thresholds, to a two-dimensional phase analysis method based on time-varying trends and pressure amplitude characteristics by structuring the pressure time series collected by a single pressure sensor. This allows the pressure change process to be divided into a quantifiable structure containing phase duration and pressure amplitude changes. By constructing a pressure phase hypergraph, this invention can extract phase combination relationships and change patterns from continuous pressure time series, thus fully expressing the structural characteristics of the pressure change process. This effectively overcomes the shortcomings of existing technologies, such as the inability to identify ignition transients, distinguish between normal combustion fluctuations and micro-leakage, and the difficulty in capturing slow changing trends, achieving higher accuracy in gas state determination.

[0055] This invention constructs a three-layer hierarchical state machine model comprising a valve state machine, a gas state machine, and a risk state machine. This model enables pressure phase characteristics, phase combination structures, and two-dimensional phase residual signals to directly drive multi-layer state transitions. State determination relies not only on pressure amplitude but also on a complete phase structure and phase trend logic. The collaborative update mechanism among the three state machines provides stronger logical constraints and completeness in gas state determination, maintaining the reliability of state transitions under complex operating conditions. This reduces misjudgments, missed judgments, and abnormal state transitions caused by simple threshold triggering in traditional methods, thereby improving the accuracy of gas system state identification.

[0056] This invention, in terms of risk assessment, introduces two-dimensional phase residual signals and hierarchical state transmission results, unifying pressure behavior trends, structural deviations, and hierarchical state characteristics into a risk assessment model. It establishes a complete risk level generation mechanism and achieves structured recording of risk events through a historical event buffer. Compared to existing technologies that only rely on single-point pressure values ​​for alarms, this invention can determine risk levels based on the continuous change patterns of pressure behavior over time, enabling earlier identification of progressive hazards such as micro-leaks. This improves the timeliness and effectiveness of risk response, thereby achieving a comprehensive improvement in the accuracy, intelligence, adaptability, and traceability of gas safety monitoring. Attached Figure Description

[0057] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0058] Figure 1 This is a flowchart of a gas state assessment and risk identification method based on pressure time sequence proposed in this invention;

[0059] Figure 2 This is a schematic diagram illustrating the composition of the working condition template library and the template update process of a gas state assessment and risk identification method based on pressure time sequence proposed in this invention.

[0060] Figure 3 This is a schematic diagram of a three-layer hierarchical state machine model and state transition relationship for a gas state assessment and risk discrimination method based on pressure time sequence proposed in this invention. Detailed Implementation

[0061] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0062] refer to Figure 1 , Figure 2 and Figure 3A method for gas state assessment and risk identification based on pressure time series includes:

[0063] The pressure value of the gas pipeline is collected using a single pressure sensor, and different sampling frequencies are set to obtain time-series data of pressure changes over time.

[0064] Continuous gradient analysis is performed on the pressure time series data in the time dimension. The pressure time series is divided into four phase segments according to the direction and rate of pressure change. Two-dimensional phase features consisting of phase duration and pressure amplitude change are extracted for each phase segment.

[0065] A pressure phase hypergraph is constructed based on multiple two-dimensional phase features. Each phase segment is used as a hypernode, and the sequential relationship between phases and the amplitude difference between phases are used as hyperedges. The current pressure time series is mapped as a hypergraph path composed of hypernodes and hyperedges.

[0066] A working condition template library is constructed using the pressure phase hypergraphs of normal ignition, normal combustion, and normal ignition shutdown processes. The working condition template library is updated online to form an adaptive baseline. The current pressure phase hypergraph path is structurally matched with the working condition template library to generate a two-dimensional phase residual signal.

[0067] A hierarchical state machine model is constructed based on two-dimensional phase residual signals. The valve state machine and the gas state machine are independent levels. The hierarchical state machine model is driven to perform state transitions by using the two-dimensional phase characteristic change trend, phase combination structure and two-dimensional phase residual signals, and outputs multi-state gas state.

[0068] Risk levels are generated based on the multi-state gas state and two-dimensional phase residual signals. When the risk level reaches the preset trigger condition, valve closing action is performed. The two-dimensional phase characteristics, pressure phase hypergraph path, residual signal and risk level are written into the historical event buffer to form a traceable event record.

[0069] In this embodiment, the step of setting different sampling frequencies to obtain pressure time-series data that changes over time includes: setting a pressure sampling frequency of no more than 1 Hz for the idle state, setting a pressure sampling frequency of no less than 20 Hz for the ignition detection state, setting a pressure sampling frequency of 3 Hz to 10 Hz for the steady-state combustion state, and setting a pressure sampling frequency of no less than 12 Hz for the suspected abnormal state, and continuously acquiring pressure sampling values ​​at the sampling frequency, and obtaining pressure time-series data that changes over time based on the pressure sampling values.

[0070] In this embodiment, the extraction of two-dimensional phase features composed of phase duration and pressure amplitude changes for each phase segment includes:

[0071] Each pressure sampling point is read sequentially according to the sampling time order, and the sampling time information and corresponding sampling frequency information are associated with each pressure sampling point to form a pressure sampling sequence.

[0072] For each pressure sampling point in the pressure sampling sequence, excluding the first sampling point, calculate the pressure change between the pressure sampling point and the previous sampling point, as well as the rate of change of the pressure change over the actual time interval. Use the direction and rate of pressure change as local change features of the pressure sampling point. The calculation of the pressure change between the pressure sampling point and the previous sampling point, and the rate of change of the pressure change over the actual time interval, includes:

[0073] Read the pressure value of the current pressure sampling point and the pressure value of the previous pressure sampling point, and use the difference between the two values ​​as the pressure change.

[0074] Read the sampling time of the current pressure sampling point and the sampling time of the previous pressure sampling point, and use the time difference between the two as the actual time interval;

[0075] The pressure change is compared with the corresponding actual time interval. The direction of pressure change at the current pressure sampling point is determined based on the numerical relationship between the two. The rate of pressure change at the current pressure sampling point is determined based on the degree of change of the pressure change within the actual time interval.

[0076] Based on the local variation characteristics of each pressure sampling point, the pressure sampling points are classified and labeled:

[0077] Based on the local change characteristics of each pressure sampling point, sampling points whose pressure change direction is decreasing and whose change rate is greater than the first gradient boundary parameter are marked as fast descent candidate points;

[0078] Sampling points where the pressure change direction is upward and the rate of change is greater than the second gradient boundary parameter are marked as candidate points for rapid recovery.

[0079] Sampling points whose absolute rate of change in the direction of pressure change is less than the third gradient boundary parameter are marked as steady-state candidate points;

[0080] Sampling points where the pressure changes in a downward direction and the rate of change is between the first gradient boundary parameter and the third gradient boundary parameter are marked as slow-decreasing candidate points.

[0081] Wherein are the first gradient boundary parameter, the second gradient boundary parameter, and the third gradient boundary parameter:

[0082] For each pressure sampling point in the historical normal ignition process, calculate the rate of pressure change. Arrange all the rates of change in the downward direction in descending order of value, and select the first 10% of the rates of change in the arrangement as the first gradient boundary parameter.

[0083] For each pressure sampling point during the recovery process after a successful ignition, the pressure change rate is calculated. All the rate of change in the upward direction are arranged in descending order of value. The first 10% of the rate of change in the arrangement is selected as the second gradient boundary parameter.

[0084] For each pressure sampling point in the historical normal idle state and normal steady-state combustion state, calculate the rate of pressure change. Arrange all the absolute values ​​of the rate of change in ascending order of value, and select the absolute values ​​of the first 10% of the rate of change in the arrangement as the third gradient boundary parameter.

[0085] After determining the first gradient boundary parameter, the second gradient boundary parameter, and the third gradient boundary parameter, the three types of gradient boundary parameters are proportionally converted according to the scaling factor between the current sampling frequency and the historical sample sampling frequency to obtain a gradient boundary parameter set consistent with the current sampling frequency.

[0086] In each type of candidate point, adjacent candidate points of the same type are merged into a continuous phase segment according to the time sequence. Short phase segments with a duration less than the preset minimum duration are merged or eliminated. The start and end positions of the phase segments are corrected according to the preset timing constraints of fast falling phase, fast rising phase, steady-state phase and slow falling phase.

[0087] For each phase segment obtained after correction, the difference between the start time and the end time of the phase segment is determined as the phase duration, and the difference between the start pressure value and the end pressure value of the phase segment is determined as the pressure amplitude change. The phase duration and the pressure amplitude change are combined to form a two-dimensional phase feature.

[0088] In this embodiment, the construction of the pressure phase hypermap based on multiple two-dimensional phase features includes:

[0089] For each phase segment, the phase type, phase duration, and pressure amplitude change are read separately. The phase duration and pressure amplitude change are combined to form a two-dimensional phase feature set, while maintaining the temporal order of the two-dimensional phase features and the corresponding phase segments.

[0090] Based on the numerical range of the duration and pressure amplitude changes of each phase in the two-dimensional phase feature set, the phase features are clustered according to the phase type and the similarity of the two-dimensional features. Multiple phase segments belonging to the same cluster result are grouped into the same phase cluster. Each phase cluster is abstracted into a super node, and each super node carries the representative phase type, representative phase duration and representative pressure amplitude change of its respective phase cluster.

[0091] On the time axis, according to the actual occurrence order of phase segments, adjacent phase combinations containing multiple phase segments are selected sequentially in a sliding time window manner. A hyperedge connection relationship is established between multiple supernodes that appear successively in the same time window and whose representative pressure amplitude changes meet the preset amplitude difference conditions. The time sequence relationship and pressure amplitude difference are used as constraints for establishing the hyperedge.

[0092] The hyperedge connections obtained within all time windows are merged, and hyperedges belonging to the same type of phase combination pattern are summarized as pattern hyperedges. The temporal sequence of phase segments, the phase type sequence, and the combination pattern of phase duration and pressure amplitude change are used as the attribute information of the pattern hyperedges, forming a pressure phase hypergraph structure composed of multiple hypernodes and multiple pattern hyperedges.

[0093] According to the time sequence of a single gas usage process, the supernodes involved in the usage process are connected in series according to their connection relationship in the pressure phase supergraph to form a supergraph path. The supergraph path serves as a structured representation of the pressure time sequence in the gas usage process at the phase and amplitude levels.

[0094] In this embodiment, generating the two-dimensional phase residual signal includes:

[0095] From the historical event buffer, only gas usage processes that are judged as normal ignition conditions, normal combustion conditions, and normal ignition conditions and continuously meet the preset acceptance criteria are selected, and the corresponding pressure phase hypergraph paths are used as candidate condition samples. The acceptance criteria include: the hierarchical state machine output is normal, the two-dimensional phase residual signal is lower than the preset range and continuously meets the preset number of times.

[0096] The pressure phase hypergraph path of the candidate working condition sample is standardized, including phase type alignment, interval labeling of phase duration, interval labeling of pressure amplitude change and time sequence consistency check, to generate template unit;

[0097] Based on the template unit, a working condition template library is constructed. Global templates and user-level sub-templates are established for ignition, combustion, and shutdown working conditions, respectively. Each template consists of a set of supernode representatives and a set of mode superedges. A phase topology index and an amplitude range index are established.

[0098] Online updates are performed on the working condition template library. When a newly generated normal working condition sample passes the acceptance criteria, the corresponding supernode representative set is merged into the supernode representative set of the same type of template. The count information and most recent update time of the mode superedge set are updated. The value range of the phase duration interval and pressure amplitude interval are corrected according to a preset weight decay strategy. When an abnormal sample is triggered in the working condition template, a template contamination protection strategy is activated, and the acceptance and merging operation of the corresponding working condition template is suspended. Specifically, the activation of the template contamination protection strategy when an abnormal sample is triggered in the working condition template is as follows:

[0099] When the current pressure phase hypermap path and the structural matching result of the corresponding working condition template are inconsistent, record the inconsistency event and mark the inconsistency event as an abnormal sample trigger event.

[0100] After the abnormal sample triggering event is recorded, the acceptance status of the corresponding working condition template is set to the paused state, and the incorporation of new supernode representative sets and pattern superedge sets into the working condition template is stopped.

[0101] During the pause of template acceptance, newly generated normal operating condition samples are stored independently, and after the pause is lifted, a judgment is made on whether to merge the independently stored operating condition samples according to the recovery conditions.

[0102] During the current gas usage process, after the corresponding pressure phase hypergraph path is generated, a working condition template consistent with the current working condition type is selected from the working condition template library. First, a phase topology index-driven structural matching is performed to determine the phase type sequence correspondence. Then, an amplitude interval index-driven refined matching is performed to determine the deviation between the pressure amplitude interval and the phase duration interval of each phase. The two-dimensional phase residual signal sequence generated by each phase segment is output in chronological order as the two-dimensional phase residual signal for the current gas usage process, where:

[0103] First, perform phase topology index-driven structure matching to determine the correspondence of phase type sequences, specifically:

[0104] Read the sequence of supernodes arranged in chronological order in the current pressure phase supermap path, and compare the phase type labels of the supernode sequences one by one with the phase type labels recorded in the selected working condition template.

[0105] During the comparison process, when the phase type label of the current supernode matches the phase type label at the same position in the selected working condition template, the correspondence between the positions is recorded.

[0106] When a phase type label is inconsistent, the phase type label that matches the current supernode is searched from the adjacent phase types according to the recording order of the phase topology index in the working condition template, and the index position corresponding to the phase type is recorded to form a phase type sequence correspondence between the current supergraph path and the working condition template.

[0107] Then, a refined matching driven by the amplitude range index is performed to determine the deviation between the pressure amplitude range and the phase duration range for each phase, specifically:

[0108] Read the pressure amplitude variation range and phase duration corresponding to each phase segment in the current pressure phase hypermap path, and compare the pressure amplitude variation range and phase duration with the pressure amplitude range and phase duration range of the same phase type in the working condition template.

[0109] When the pressure amplitude change range exceeds the pressure amplitude interval boundary recorded in the working condition template, the pressure amplitude deviation will be recorded as the amplitude deviation of the phase segment.

[0110] When the phase duration exceeds the boundary of the phase duration interval recorded in the working condition template, the duration deviation is recorded as the duration deviation of the phase segment.

[0111] The deviations are output sequentially according to the time order of the phase in the pressure phase hypergraph path, forming a two-dimensional phase residual signal sequence.

[0112] In this embodiment, the step of correcting the value range of the phase duration interval and the pressure amplitude interval according to the preset weighted attenuation strategy includes:

[0113] A first weight is set to characterize the influence of the new sample, and the value of the first weight is between 0.6 and 0.8;

[0114] A second weight is set to characterize the retention degree of existing sample intervals in the working condition template library. The value of the second weight is complementary to the first weight, that is, the second weight is equal to 1 minus the first weight.

[0115] Multiply the phase duration and pressure amplitude changes corresponding to the normal operating condition sample by the first weight, multiply the existing phase duration interval and pressure amplitude interval in the operating condition template library by the second weight, and add the two together as the upper and lower boundaries of the updated interval range.

[0116] In this embodiment, the output of multi-state gas states includes:

[0117] A three-layer hierarchical state machine model is established, with the valve state machine, gas state machine, and risk state machine as the upper, middle, and lower layers, respectively. The valve state machine includes open, closed, and fault states; the gas state machine includes idle, ignition, combustion, flameout, micro-leakage, and large-leakage states; and the risk state machine includes normal, attention, warning, and emergency states. A state combination table in which the three states can occur simultaneously is recorded in the hierarchical state machine model.

[0118] During each round of state update, the control command of the valve actuator and the valve position detection result are read. The current state of the valve state machine is updated according to the correspondence between the control command and the detection result. The updated current state of the valve state machine is written as the upper-level state into the corresponding record in the state combination table.

[0119] Read the phase type change, phase duration change and pressure amplitude change in the pressure phase hypergraph path and two-dimensional phase residual signal. Generate candidate states of the gas state machine based on the state transition conditions recorded in the gas state machine and the current state of the gas state machine at the previous moment. Select the target state of the gas state machine from the candidate states according to the state combination table record that is consistent with the current valve state machine state.

[0120] The current state of the valve state machine, the target state of the gas state machine, and the two-dimensional phase residual signal are input into the risk state machine. The target state of the risk state machine is determined according to the state determination rules recorded in the risk state machine. When the target state of the risk state machine is an emergency state, a valve closing control command is generated and sent to the valve actuator. The target state of the gas state machine is then adjusted to a state related to terminating gas usage. Specifically, determining the target state of the risk state machine according to the state determination rules recorded in the risk state machine involves:

[0121] When the current state of the valve state machine is open and the target state of the gas state machine is micro-leakage, the amplitude conversion value of the two-dimensional phase residual signal is compared with the attention level threshold recorded in the risk state machine. When the amplitude conversion value is greater than the attention level threshold, the target state of the risk state machine is set to the attention state.

[0122] When the target state of the gas state machine is an abnormal state and the duration conversion value of the two-dimensional phase residual signal is compared with the warning level threshold recorded in the risk state machine, when the duration conversion value is greater than the warning level threshold, the target state of the risk state machine is set to a warning state.

[0123] When the target state of the gas state machine is a large leakage state or the amplitude conversion value of the two-dimensional phase residual signal is compared with the emergency level threshold recorded in the risk state machine and meets the emergency level threshold, the target state of the risk state machine is set to an emergency state.

[0124] After completing the current state update of the valve state machine, gas state machine, and risk state machine, the current state of the three state machines is recorded, and the current state of the gas state machine is output as the polymorphic gas state.

[0125] In this embodiment, the step of generating a risk level based on the multi-state gas state and the two-dimensional phase residual signal includes:

[0126] Read the polymorphic gas state, two-dimensional phase residual signal, and current state of the risk state machine obtained from the current round of state update, associate the polymorphic gas state with the two-dimensional phase residual signal, and generate risk assessment input data for the current round of gas usage process;

[0127] Based on the type of polymorphic gas state, the amplitude of the two-dimensional phase residual signal, and the duration of the residual signal over time in the risk assessment input data, the corresponding risk index value is calculated. This value is then compared with the risk level classification conditions recorded in the control unit to obtain the risk level code. Specifically, the calculation of the corresponding risk index value involves:

[0128] Read the state type corresponding to the current multi-state gas state, use the risk weight preset in the control unit as the state weight, and record the state weight;

[0129] Read the amplitude value of the current two-dimensional phase residual signal, perform linear conversion on the amplitude value according to the amplitude coefficient recorded in the control unit, obtain the amplitude conversion value, and record the amplitude conversion value;

[0130] Read the duration of the current two-dimensional phase residual signal, convert the duration according to the time coefficient recorded in the control unit, obtain the duration conversion value, and record the duration conversion value;

[0131] The state weight, amplitude conversion value and duration conversion value are summed, and the summed result is used as the risk indicator value corresponding to the current pressure time series.

[0132] The risk level code is compared with the valve closing trigger condition recorded in the control unit. When the risk level code meets the valve closing trigger condition, a valve closing control command is generated in the control unit and output to the valve actuator. The time when the risk level code meets the trigger condition in this round is recorded.

[0133] When the risk level code does not meet the valve closing trigger condition, the current valve opening and closing state remains unchanged. The risk level code is associated with the corresponding polymorphic gas state and two-dimensional phase residual signal, and output as the risk assessment result of this round to the upper monitoring or alarm module.

[0134] After completing the risk level calculation and valve closing command judgment, the two-dimensional phase characteristics, pressure phase hypergraph path, two-dimensional phase residual signal, multi-state gas state, current state of risk state machine, risk level code, and valve closing command execution results generated during this round of gas use are written into the historical event buffer in chronological order to form event record entries.

[0135] Example 1:

[0136] To verify the feasibility of this invention in practice, it was applied to a resident's home in a certain community. The resident's kitchen uses natural gas, and two single-burner stoves are connected to a gas branch line approximately 2.8 meters long with a pipe diameter of 20 mm and an inlet static pressure of approximately 1950 Pa. To verify the pressure-sequential gas state assessment and risk identification method of this invention, without changing the original gas structure in the kitchen or using any additional sensors, a single pressure sensor as described in this invention was installed only on the gas valve outlet side.

[0137] In the first phase of testing in this embodiment, the user conducted 12 consecutive real cooking processes. The system first established an idle pressure baseline with a sampling frequency not exceeding 1Hz. Under idle conditions, the pressure fluctuation range was maintained between 1935Pa and 1960Pa. Subsequently, the user turned on the stove to ignite it, and the system automatically switched to a high-frequency sampling mode of not less than 20Hz. Approximately 0.3 seconds after the ignition knob was turned, the pressure dropped from 1950Pa to 1780Pa, forming a rapid drop phase; then, it rapidly rebounded to 1885Pa within 0.25 seconds and entered the steady-state combustion stage. The system divided the drop, rebound, and steady-state phases into different phase segments and recorded the duration and pressure amplitude changes of each phase, forming a two-dimensional phase feature. These phase features were converted into supernodes, and the temporal relationships and pressure amplitude differences between multiple nodes formed hyperedges, thereby constructing a complete pressure phase hypergraph path.

[0138] In the first phase of continuous testing, every ignition-combustion-off process could find a matching path in the operating condition template library, indicating that the structured modeling method of the template library can accurately reflect the real characteristics of the household gas system. Taking a cooking process as an example, the steady-state combustion pressure was maintained at 1860Pa ​​to 1880Pa, and the pressure rose back to 1930Pa within 1 second when the gas was turned off. The system matched the phase hypergraph of this usage process with the template library and generated a two-dimensional phase residual signal based on the difference. The results showed that the residual change remained within a stable range and no abnormalities occurred.

[0139] The second phase of testing was used to verify the micro-leakage detection capability. On-site testers created a micro-leakage hole with a diameter of approximately 0.10 mm on the flexible hose at the front of the stove. With the stove in the off position, the pressure dropped only about 1.2 Pa per second from 1945 Pa. The pressure change was small in magnitude and long in duration, typical of a "slowly decreasing phase." Traditional threshold methods fail to trigger alarms because the pressure did not significantly deviate from the baseline. However, this invention, based on the formed two-dimensional phase structure and hypergraph path, can identify a continuously decreasing trend structure. Simultaneously, it uses a three-layered state machine to transition the gas state from idle to micro-leakage state. This test lasted 180 seconds, during which the system detected 178 micro-leakage states without any false alarms or false triggers for normal fluctuations.

[0140] To further verify the risk assessment capability of this invention, a large-leakage safety simulation was conducted in the third phase of testing. The leak hole was enlarged to 1.0 mm, at which point the pressure dropped sharply from 1948 Pa to 1602 Pa within 0.25 seconds. This was recorded as a very short, sharp drop segment using two-dimensional phase characteristics, exhibiting a significant anomalous deviation in the hypergraph structure. The system generated a high-amplitude two-dimensional phase residual signal during hypergraph matching, triggering an anomalous transition of the hierarchical state machine. In this test, the risk state machine switched to an "emergency state," outputting a valve-closing action. The entire process, from the sudden pressure change to complete valve closure, took only approximately 1.8 seconds, preventing the potential danger from escalating.

[0141] Table 1. Statistical Table of Pressure Phase Characteristics and Identification Results During Gas Use

[0142]

[0143] Table 1 shows that the present invention can accurately characterize the pressure timing features of ignition, combustion, and shut-off during normal cooking. The ignition phase exhibits a typical rapid decrease-rapid recovery structure; the combustion phase pressure fluctuates stably within the range of 1860–1885 Pa; and the shut-off pressure briefly recovers to 1930 Pa. All these features are completely captured by the two-dimensional phase feature and pressure phase hypergraph. The identification results are completely consistent with the actual process in the embodiments, indicating that the present invention can accurately establish a normal operating condition template and achieve stable identification of normal gas usage behavior.

[0144] In the micro-leakage scenario, the multiple tests in the table all showed a continuous and slow downward trend, with a decrease rate of approximately 1.2 Pa / s, which remained consistent under different initial pressures. Although the pressure change was small and without significant abrupt changes, traditional threshold-based methods could not detect it. However, this invention, through the two-dimensional phase characteristics formed by the phase duration and pressure amplitude changes, as well as the stable deviation of the hypergraph path, can continuously identify micro-leakage states and output information of interest, without false alarms or missed detections. This demonstrates that this invention can effectively capture gradual leakage trends and solves the problem of existing technologies being insensitive to minor risks.

[0145] In the large leak test, the pressure plummeted from 1948 Pa to 1602 Pa in less than 0.3 seconds, exhibiting a typical abrupt change. The two-dimensional phase feature constructed in this invention showed significant anomalies in both phase duration and amplitude differences. The pressure phase hypergraph path deviated severely from the template library, and the hierarchical state machine rapidly entered a large leak-emergency state, triggering valve closure. After valve closure, the pressure quickly recovered to the baseline region, verifying the system's rapid and reliable response capability under high-risk conditions. This invention enables high-precision identification and risk control of different operating conditions under single-sensor conditions.

[0146] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A pressure time series based gas state evaluation and risk discrimination method, characterized in that, include: The pressure value of the gas pipeline is collected using a single pressure sensor, and different sampling frequencies are set to obtain time-series data of pressure changes over time. Continuous gradient analysis is performed on the pressure time series data in the time dimension. The pressure time series is divided into four types of phase segments according to the direction and rate of pressure change. Two-dimensional phase features consisting of phase duration and pressure amplitude change are extracted for each phase segment. A pressure phase hypergraph is constructed based on multiple two-dimensional phase features. Multiple phase segments belonging to the same cluster result are grouped into the same type of phase cluster, and each phase cluster is abstracted as a hypernode. The temporal sequence relationship and pressure amplitude difference are used as constraints for establishing hyperedges. Hyperedge connection relationship is established between hypernodes that meet the constraints, and the current pressure time series is mapped as a hypergraph path composed of hypernodes and hyperedges. A working condition template library is constructed using the pressure phase hypergraph of normal ignition, normal combustion, and normal shutdown processes. The working condition template library is updated online to form an adaptive baseline. The current pressure phase hypergraph path is matched with the working condition template library using phase topology index-driven structural matching to determine the correspondence of phase type sequences. Then, a refined matching driven by amplitude interval index is performed to determine the deviation between the pressure amplitude interval and the phase duration interval of each phase. The deviation is output sequentially according to the time order of the phase in the pressure phase hypergraph path to generate a two-dimensional phase residual signal. A hierarchical state machine model is constructed based on two-dimensional phase residual signals. The valve state machine, gas state machine and risk state machine are independent levels. The hierarchical state machine model is driven to perform state transition by the two-dimensional phase characteristic change trend, phase combination structure and two-dimensional phase residual signals, and outputs multi-state gas state. Risk levels are generated based on the multi-state gas state and two-dimensional phase residual signals. When the risk level reaches the preset trigger condition, valve closing action is performed. The two-dimensional phase characteristics, pressure phase hypergraph path, residual signal and risk level are written into the historical event buffer to form a traceable event record.

2. The method according to claim 1, wherein, The step of setting different sampling frequencies to obtain pressure time-series data that changes over time includes: setting a pressure sampling frequency of no more than 1 Hz for the idle state, a pressure sampling frequency of no less than 20 Hz for the ignition detection state, a pressure sampling frequency of 3 Hz to 10 Hz for the steady-state combustion state, and a pressure sampling frequency of no less than 12 Hz for the suspected abnormal state, and continuously acquiring pressure sampling values ​​at the sampling frequency, and obtaining pressure time-series data that changes over time based on the pressure sampling values.

3. The gas state assessment and risk identification method based on pressure time sequence according to claim 1, characterized in that, The extraction of two-dimensional phase features, consisting of phase duration and pressure amplitude changes, for each phase segment includes: Each pressure sampling point is read sequentially according to the sampling time order, and the sampling time information and corresponding sampling frequency information are associated with each pressure sampling point to form a pressure sampling sequence. For each pressure sampling point in the pressure sampling sequence except for the first sampling point, calculate the pressure change between the pressure sampling point and the previous sampling point, as well as the rate of change of the pressure change over the actual time interval. Use the direction and rate of change of pressure change as the local change features of the pressure sampling point. Based on the local variation characteristics of each pressure sampling point, the pressure sampling points are classified and labeled: Based on the local change characteristics of each pressure sampling point, sampling points whose pressure change direction is decreasing and whose change rate is greater than the first gradient boundary parameter are marked as fast descent candidate points; Sampling points where the pressure change direction is upward and the rate of change is greater than the second gradient boundary parameter are marked as candidate points for rapid recovery. Sampling points whose absolute rate of change in the direction of pressure change is less than the third gradient boundary parameter are marked as steady-state candidate points; Sampling points where the pressure changes in a downward direction and the rate of change is between the first gradient boundary parameter and the third gradient boundary parameter are marked as slow-decreasing candidate points. In each type of candidate point, adjacent candidate points of the same type are merged into a continuous phase segment according to the time sequence. Short phase segments with a duration less than the preset minimum duration are merged or eliminated. The start and end positions of the phase segments are corrected according to the preset timing constraints of fast falling phase, fast rising phase, steady-state phase and slow falling phase. For each phase segment obtained after correction, the difference between the start time and the end time of the phase segment is determined as the phase duration, and the difference between the start pressure value and the end pressure value of the phase segment is determined as the pressure amplitude change. The phase duration and the pressure amplitude change are combined to form a two-dimensional phase feature.

4. The method for gas state assessment and risk identification based on pressure time sequence according to claim 1, characterized in that, The construction of the pressure phase hypermap based on multiple two-dimensional phase features includes: For each phase segment, the phase type, phase duration, and pressure amplitude change are read separately. The phase duration and pressure amplitude change are combined to form a two-dimensional phase feature set, while maintaining the temporal order of the two-dimensional phase features and the corresponding phase segments. Based on the numerical range of the duration and pressure amplitude changes of each phase in the two-dimensional phase feature set, the phase features are clustered according to the phase type and the similarity of the two-dimensional features. Multiple phase segments belonging to the same cluster result are grouped into the same phase cluster. Each phase cluster is abstracted into a super node, and each super node carries the representative phase type, representative phase duration and representative pressure amplitude change of its respective phase cluster. On the time axis, according to the actual occurrence order of phase segments, adjacent phase combinations containing multiple phase segments are selected sequentially in a sliding time window manner. A hyperedge connection relationship is established between multiple supernodes that appear successively in the same time window and whose representative pressure amplitude changes meet the preset amplitude difference conditions. The time sequence relationship and pressure amplitude difference are used as constraints for establishing the hyperedge. The hyperedge connections obtained within all time windows are merged, and hyperedges belonging to the same type of phase combination pattern are summarized as pattern hyperedges. The temporal sequence of phase segments, the phase type sequence, and the combination pattern of phase duration and pressure amplitude change are used as the attribute information of the pattern hyperedges, forming a pressure phase hypergraph structure composed of multiple hypernodes and multiple pattern hyperedges. According to the time sequence of a single gas usage process, the supernodes involved in the usage process are connected in series according to their connection relationship in the pressure phase supergraph to form a supergraph path. The supergraph path serves as a structured representation of the pressure time sequence in the gas usage process at the phase and amplitude levels.

5. The gas state assessment and risk identification method based on pressure time sequence according to claim 4, characterized in that, The generation of the two-dimensional phase residual signal includes: From the historical event buffer, only gas usage processes that are judged to be normal ignition conditions, normal combustion conditions, and normal flameout conditions and continuously meet the preset acceptance criteria are selected, and the corresponding pressure phase hypergraph paths are used as candidate condition samples. The pressure phase hypergraph path of the candidate working condition sample is standardized, including phase type alignment, interval labeling of phase duration, interval labeling of pressure amplitude change and time sequence consistency check, to generate template unit; Based on the template unit, a working condition template library is constructed. Global templates and user-level sub-templates are established for ignition, combustion, and shutdown working conditions, respectively. Each template consists of a set of supernode representatives and a set of mode superedges. A phase topology index and an amplitude range index are established. The working condition template library is updated online. When a newly generated normal working condition sample passes the acceptance criteria, the corresponding supernode representative set is merged into the supernode representative set of the same type of template. The count information and the most recent update time of the mode superedge set are updated. The value range of the phase duration interval and pressure amplitude interval are corrected according to the preset weight decay strategy. When an abnormal sample is triggered in the working condition template, the template contamination protection strategy is activated and the acceptance and merging operation of the corresponding working condition template is suspended. During the current gas usage process, after the corresponding pressure phase hypergraph path is generated, a working condition template consistent with the current working condition type is selected from the working condition template library. First, a phase topology index-driven structural matching is performed to determine the phase type sequence correspondence. Then, an amplitude interval index-driven refined matching is performed to determine the deviation between the pressure amplitude interval and the phase duration interval of each phase. The two-dimensional phase residual signal sequence generated by each phase segment is output in chronological order as the two-dimensional phase residual signal of the current gas usage process.

6. The gas state assessment and risk identification method based on pressure time sequence according to claim 5, characterized in that, The method of correcting the value range of the phase duration interval and the pressure amplitude interval according to the preset weight attenuation strategy includes: A first weight is set to characterize the influence of the new sample, and the value of the first weight is between 0.6 and 0.8; A second weight is set to characterize the retention degree of existing sample intervals in the working condition template library. The value of the second weight is complementary to the first weight, that is, the second weight is equal to 1 minus the first weight. Multiply the phase duration and pressure amplitude changes corresponding to the normal operating condition sample by the first weight, multiply the existing phase duration interval and pressure amplitude interval in the operating condition template library by the second weight, and add the two together as the upper and lower boundaries of the updated interval range.

7. The method for gas state assessment and risk identification based on pressure time sequence according to claim 1, characterized in that, The output multi-state gas state includes: A three-layer hierarchical state machine model is established, with the valve state machine, gas state machine, and risk state machine as the upper, middle, and lower layers, respectively. The valve state machine includes open, closed, and fault states; the gas state machine includes idle, ignition, combustion, flameout, micro-leakage, and large-leakage states; and the risk state machine includes normal, attention, warning, and emergency states. A state combination table in which the three states can occur simultaneously is recorded in the hierarchical state machine model. During each round of state update, the control command of the valve actuator and the valve position detection result are read. The current state of the valve state machine is updated according to the correspondence between the control command and the detection result. The updated current state of the valve state machine is written as the upper-level state into the corresponding record in the state combination table. Read the phase type change, phase duration change and pressure amplitude change in the pressure phase hypergraph path and two-dimensional phase residual signal. Generate candidate states of the gas state machine based on the state transition conditions recorded in the gas state machine and the current state of the gas state machine at the previous moment. Select the target state of the gas state machine from the candidate states according to the state combination table record that is consistent with the current valve state machine state. The current state of the valve state machine, the target state of the gas state machine, and the two-dimensional phase residual signal are input into the risk state machine. The target state of the risk state machine is determined according to the state determination rules recorded in the risk state machine. When the target state of the risk state machine is an emergency state, a valve closing control command is generated and sent to the valve actuator. The target state of the gas state machine is adjusted to a state related to the termination of gas use. After completing the current state update of the valve state machine, gas state machine, and risk state machine, the current state of the three state machines is recorded, and the current state of the gas state machine is output as the polymorphic gas state.

8. The gas state assessment and risk identification method based on pressure time sequence according to claim 1, characterized in that, The risk level generation based on the multi-state gas state and two-dimensional phase residual signal includes: Read the polymorphic gas state, two-dimensional phase residual signal, and current state of the risk state machine obtained from the current round of state update, associate the polymorphic gas state with the two-dimensional phase residual signal, and generate risk assessment input data for the current round of gas usage process; Based on the type of polymorphic gas state, the amplitude of the two-dimensional phase residual signal, and the duration of the residual signal over time in the risk assessment input data, the corresponding risk index value is calculated. This value is then compared with the risk level classification conditions recorded in the control unit to obtain the risk level code. The specific calculation of the corresponding risk index value is as follows: Read the state type corresponding to the current multi-state gas state, use the risk weight preset in the control unit as the state weight, and record the state weight; Read the amplitude value of the current two-dimensional phase residual signal, perform linear conversion on the amplitude value according to the amplitude coefficient recorded in the control unit, obtain the amplitude conversion value, and record the amplitude conversion value; Read the duration of the current two-dimensional phase residual signal, convert the duration according to the time coefficient recorded in the control unit, obtain the duration conversion value, and record the duration conversion value; The state weight, amplitude conversion value and duration conversion value are summed, and the summed result is used as the risk indicator value corresponding to the current pressure time series. The risk level code is compared with the valve closing trigger condition recorded in the control unit. When the risk level code meets the valve closing trigger condition, a valve closing control command is generated in the control unit and output to the valve actuator. The time when the risk level code meets the trigger condition in this round is recorded. When the risk level code does not meet the valve closing trigger condition, the current valve opening and closing state remains unchanged. The risk level code is associated with the corresponding polymorphic gas state and two-dimensional phase residual signal, and output as the risk assessment result of this round to the upper monitoring or alarm module. After completing the risk level calculation and valve closing command judgment, the two-dimensional phase characteristics, pressure phase hypergraph path, two-dimensional phase residual signal, multi-state gas state, current state of risk state machine, risk level code, and valve closing command execution results generated during this round of gas use are written into the historical event buffer in chronological order to form event record entries.

Images