Fire-fighting water system fault detection and early warning method and system based on fire-fighting internet of things

Abstract

The invention relates to the technical field of fire-fighting monitoring, in particular to a fire-fighting water system fault detection and early warning method and system based on the fire-fighting Internet of Things, and the method comprises the steps: 1, periodically collecting pressure data through pressure sensors disposed at a fire-fighting water pump outlet, a pipe network main pipe branch point and the most unfavorable tail end; 2, according to the current valve opening degree, the pump state and historical normal working condition data, theoretical pressure expected values and dynamic allowable deviation zones of all nodes are generated; 3, when actually measured pressure deviates from a theoretical pressure expected value and exceeds a dynamic allowable deviation band, marking abnormal nodes, reversely constructing a fault propagation tree along the topological model, and allocating weight factors for associated nodes according to fault types; and 4, triggering graded early warning based on the number of the abnormal nodes, the fault propagation path and the weight factor accumulated value. And the system can automatically take measures when a fault occurs through an equipment linkage function, so that the efficient operation of the fire fighting water system is guaranteed, and the efficiency and safety of fire emergency response are improved.

Description

Technical Field

[0001] The present invention relates to the field of fire monitoring technology, and in particular to a fire water system fault detection and early warning method and system based on a fire Internet of Things. Background Art

[0002] As a core component of building fire protection facilities, the reliability of the fire water system is directly related to the effectiveness of fire emergency response. Current mainstream monitoring systems deploy pressure sensors at key nodes in the pipe network and use a fixed threshold trigger alarm mechanism (such as triggering when the pressure falls below 0.5 MPa). However, in complex, branched pipe network topologies, this mechanism has the following inherent drawbacks: High misjudgment rate of a single parameter: Because faults such as pipeline leaks, pump failure, and sensor drift all manifest as pressure drops, the system is unable to distinguish the root cause. For example, in a large commercial complex project, sensor false alarms led to an average of 3.2 false activations of backup pumps per month, increasing annual maintenance costs by 150,000 yuan. The technical rationale for this problem lies in the fact that fire water systems are closed pressure vessels. Faults at any location are transmitted to the entire system through pressure waves, and traditional single-point threshold detection inevitably confuses fault types.

[0003] Lack of fault location capability: Existing technologies can only detect pressure anomalies but cannot trace the source of the fault. When a branch pipe leak causes a pressure fluctuation in the main pipe, maintenance personnel must manually check all branch valves, taking an average of over two hours. This is necessary because branched pipe networks have a unique fault propagation path (the pressure attenuation gradient upstream of the leak is significantly higher than downstream). However, existing methods fail to establish a topological correlation model, resulting in inefficient fault isolation.

[0004] Serious response lag: Fixed-threshold mechanisms require the pressure to drop to a critical value before triggering an alarm, making them incapable of detecting small, progressive leaks below 10%. Experimental data shows that a 2L / min leak in a DN100 pipe takes 6-8 hours to trigger an alarm, during which time the leak can reach up to 1 ton. This lag is likely due to the initial masking of small leaks by system noise, and traditional methods lack the ability to analyze dynamic deviations.

[0005] Therefore, there is an urgent need for a fire water system fault detection and early warning method and system based on the fire Internet of Things to solve the above problems. Summary of the Invention

[0006] Based on the above objectives, the present invention provides a fire water system fault detection and early warning method and system based on the fire protection Internet of Things, which is characterized by including: Step 1: Pressure data is periodically collected using pressure sensors deployed at the fire pump outlet, at the branch point of the main pipe network, and at the most unfavorable end. The total system flow is collected using an electromagnetic flowmeter at the pump group outlet manifold. Branch flow is collected using an ultrasonic flowmeter behind the zone control valve. The valve opening status and pump operating status are respectively obtained using a valve opening sensor and a pump group current sensor. Step 2: Establish a pressure transmission relationship based on the network topology node-edge relationship model, and generate the theoretical pressure expectation value and dynamic tolerance band of each node according to the current valve opening, pump status and historical normal operating condition data; Step 3: When the measured pressure deviates from the expected value of the theoretical pressure and exceeds the dynamic tolerance band, mark the abnormal node and reversely construct the fault propagation tree along the topology model, and assign weight factors to the associated nodes according to the fault type; Step 4: Trigger a hierarchical warning based on the number of abnormal nodes, fault propagation path, and accumulated weight factors: A single node abnormality and unspread weight triggers a Level I alarm; Adjacent nodes are abnormal and the fault pattern matches, triggering a Level II alarm; When a key node is abnormal and the accumulated weight value exceeds the adaptive threshold, level III equipment linkage is triggered.

[0007] Preferably, the method for determining the time interval for periodically collecting pressure data in step 1 includes: Obtain the elastic modulus of the pipe, fluid density, and maximum segment length of the pipe network; Calculate the propagation velocity of the water hammer wave in the pipe section, where the propagation velocity is proportional to the square root of the elastic modulus of the pipe and inversely proportional to the square root of the fluid density; Divide the maximum segment length of the pipe network by the water hammer wave propagation velocity to obtain the critical time for pressure wave transmission; Set the sampling interval to less than half of the critical time to avoid distortion of the pressure oscillation signal.

[0008] Preferably, the method for generating the dynamic tolerance band in step 2 includes: Extract pressure-flow data pairs in continuous time windows under historical normal operating conditions; Calculate the residual standard deviation of the measured pressure value and the theoretical value in each time window; Based on the moving average of the residual standard deviation, the boundary coefficient is dynamically adjusted according to the recent false alarm rate statistics: When the number of consecutive false alarms increases, the boundary coefficient is expanded at a fixed ratio; When an underreporting event occurs, the boundary coefficient is shrunk in proportion to the fault impact range.

[0009] Preferably, the weight factor allocation rules in step 3 include: The pump node weight factor is assigned based on the correlation between the pressure drop amplitude and the current change: If the pressure drop is accompanied by current exceeding the limit, the first weight level is assigned; If the pressure drop is accompanied by current fluctuation, the second weight level is assigned; The pipeline node weight factor is assigned according to the pressure gradient change rate: When the upstream pressure gradient drops sharply, the value is assigned according to the ratio of the gradient change rate to the historical maximum value; The valve node weight factor is assigned based on the delay time between the opening command and the flow response: When the delay time exceeds the preset threshold multiple, the weight is increased linearly according to the delay time.

[0010] Preferably, the verification method for fault pattern matching in step 3 includes: a: Build a fault feature mapping rule base, where: The criteria for determining a pipeline leak are: a sudden drop in pressure at the abnormal node, an upward trend in the flow rate of its associated downstream branch, and the pump group current is within the normal fluctuation range; The criteria for determining pump group efficiency degradation are: the abnormal node pressure shows a continuous slow drop, the total system flow remains unchanged, and the pump group current exceeds the rated operating range; The conditions for determining valve sticking are: the pressure at the abnormal node fluctuates irregularly, the flow rate of its associated branch does not respond to the valve opening command, and the pump group current does not show abnormalities; Collect the real-time parameter change direction of the abnormal node and its directly topologically related nodes. When the parameter combination matches the judgment conditions in any fault feature mapping rule base, perform similarity calculation: b: Extract the derivative sign of the pressure change trend as the first matching factor; c: Extract the Boolean value of the traffic response direction as the second matching factor; d: Extract the current over-limit status flag as the third matching factor; When the joint confidence of the three matching factors exceeds the dynamic similarity threshold, the fault type is confirmed.

[0011] Preferably, the method for dynamically adjusting the adaptive threshold in step 4 includes: The initialization threshold is based on the average weight of historical failures of key nodes; When a level III false alarm occurs, obtain the fault propagation path depth; When the path depth is greater than the set level, the threshold is increased in proportion to the depth value; When the path depth is less than the set level, the threshold is increased by a fixed step value; When a level III missed alarm occurs, the weight compensation value is calculated based on the actual impact range of the fault, and the threshold is adjusted down in inverse proportion to the compensation value.

[0012] Preferably, the calculation process of the water hammer wave propagation velocity includes: Obtain the slope of the stress-strain curve through the pipe elastic modulus test experiment; Obtain real-time medium density through a fluid density detection device; Input the ratio of pipe wall thickness to pipe diameter into the elastic deformation correction model and output the wave velocity correction coefficient; The actual propagation speed is calculated based on the reference speed of the sound wave in the fluid.

[0013] Preferably, the method for determining the similarity threshold includes: Calculate the cosine similarity between the parameter combination in the historical fault samples and the standard pattern; The 75th percentile of the similarity distribution is used as the initial threshold; Each time a fault sample is added, the threshold is adjusted according to the degree of deviation between the sample and the standard pattern: If the sample is correctly classified and the similarity is higher than the current threshold, the threshold remains unchanged; If a sample is misclassified, lower or raise the threshold according to the type of misclassification.

[0014] Preferably, the method for constructing the fault propagation tree in step 3 includes: Taking the abnormal node as the root node, expand the child nodes layer by layer in the reverse direction of the water flow along the topology model; Calculate the influence weight of each side: Obtain the ratio of the pipe section pressure drop change to the total system pressure drop change; Get the ratio of the pipe segment length to the total length of the traceability path; The product of the two proportions is used as the edge weight; Select the path with the largest edge weight as the main fault propagation chain; The execution of Level III equipment linkage in step 4 includes: When starting the standby pump: Obtain the water hammer sensitivity coefficient of the pipe network between the faulty pump and the backup pump; Calculate the maximum allowable flow rate change rate based on the sensitivity coefficient; Control the standby pump to increase the pressure gradually according to the allowable change rate; When adjusting the valve: Reversely calculate the required flow rate based on the target value of pressure recovery at the fault node; Calculate the opening adjustment amount according to the valve flow characteristic curve; Adjust in segments and verify the pressure recovery progress in real time.

[0015] Correspondingly, an embodiment of the present invention also provides a fire water system fault detection and early warning system based on the fire Internet of Things, including a memory configured to store instructions, a processor configured to call the instructions from the memory and to implement the fire water system fault detection and early warning method based on the fire Internet of Things as described in any embodiment of the present invention when executing the instructions.

[0016] Beneficial effects of the present invention: 1. The multi-parameter collaborative detection method proposed in this invention not only relies on pressure data but also incorporates multi-dimensional data such as flow rate, valve opening, and pump current. Combined with dynamic tolerance band analysis, when pressure anomalies are detected, the system can further accurately distinguish the fault type by matching historical data with system status, significantly reducing the false positive rate. By precisely constructing weighting factors and fault propagation paths, false alarms caused by a single parameter are avoided, ensuring that the system can accurately determine the fault type in complex fire water pipe networks.

[0017] 2. This invention builds a node-edge relationship model based on the pipeline network topology to monitor changes in pressure, flow, and pump group status in real time. It can reverse-engineer a fault propagation tree based on abnormal nodes, accurately trace the source of the fault, and mark the abnormal node. This method greatly improves the accuracy and efficiency of fault location, avoids the time-consuming manual troubleshooting required by existing methods, significantly reduces the workload of operations and maintenance personnel, and improves the efficiency of fault isolation and resolution.

[0018] 3. By dynamically adjusting the tolerance band and analyzing the fault propagation tree, this invention can sensitively respond to subtle changes within the system and promptly detect small leaks or faults. In the early stages of a minor leak, the system optimizes the early warning mechanism through dynamic adjustment of the residual standard deviation, comparison with historical data, and false alarm rate statistics, significantly improving response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 is a flow chart of the steps of the method of the present invention; Figure 2 Flowchart of the steps for assigning the weight factors in step 3 of the method of the present invention; Figure 3 The figure is a flow chart of the steps of the process for calculating the water hammer wave propagation velocity according to the method of the present invention. DETAILED DESCRIPTION

[0021] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. It is also noted that, to provide a more detailed description, the following embodiments are best and preferred embodiments, and those skilled in the art may employ alternative methods for implementing certain known technologies. Furthermore, the accompanying drawings are intended only to provide a more detailed description of the embodiments and are not intended to limit the present invention.

[0022] See Figure 1-Figure 3 The embodiment of the present invention provides a fire water system fault detection and early warning method based on the fire Internet of Things. In step 1, first, the system deploys pressure sensors at the fire water pump outlet, the branch point of the pipeline network main pipe, and the most unfavorable end position to periodically collect pressure data in the pipeline network. At the same time, the total flow of the entire system is collected by the electromagnetic flowmeter installed on the pump group outlet main pipe, and the flow information of each branch is obtained by the ultrasonic flowmeter installed at the rear end of the partition control valve. The valve opening sensor and the pump group current sensor are further used to obtain the valve opening and pump operating status. This data collection process provides comprehensive real-time information, including pressure, flow, valve status, pump operating status, etc., which helps to comprehensively monitor the operating status of the fire water system.

[0023] In step 2, based on data collection, the system establishes the pressure transmission relationship of the pipeline network based on the node-edge relationship model of the pipeline network topology. By analyzing the current valve opening and pump status, combined with historical normal operating data, it generates a theoretical pressure expectation value for each node and sets a dynamic tolerance band based on the system characteristics. The core of this step is to accurately simulate the pressure variation pattern using the topological structure of the pipeline network and dynamically adjust the theoretical pressure expectation value based on the real-time status of the system and historical data, allowing the system to evaluate the pressure deviation of each node in real time.

[0024] In step 3, if the measured pressure deviates from the theoretically expected value by more than the dynamic tolerance band, the system marks the node as an abnormal node and reverse-engineers the fault propagation tree based on the pipeline network topology model. Based on the specific fault type (e.g., pipeline leak, pump failure, etc.), the system assigns weight factors to the associated nodes to further analyze the fault's propagation path and impact range. This approach not only detects pressure anomalies but also rapidly identifies the fault source by analyzing the pressure wave propagation path, enhancing fault location capabilities.

[0025] In step 4, based on fault detection and propagation path construction, the system triggers a hierarchical warning based on the number of abnormal nodes, the fault propagation path, and the accumulated weight factors of each node. Specifically, if a single node experiences an abnormality and the fault does not spread, the system will trigger a Level I alarm; if an adjacent node experiences an abnormality and the fault pattern matches, the system will trigger a Level II alarm; if a key node experiences an abnormality and the accumulated weight value exceeds the adaptive threshold, the system will trigger a Level III alarm and initiate a device linkage response, automatically making appropriate equipment adjustments or activating backup measures. Through this multi-level warning mechanism, the system can take timely response measures based on the severity and spread of the fault, preventing the expansion of potential disasters.

[0026] Through multi-dimensional data collection and analysis, the system addresses existing system issues such as high single-parameter misjudgment rates, difficulty locating faults, and delayed responses. By accurately modeling the pipe network topology and pressure conduction relationships, the system can accurately detect minor leaks and incipient faults, quickly locate the source of the fault, and provide a graded early warning mechanism based on weight factors, effectively reducing false alarms and missed alarms. Furthermore, through device linkage, the system automatically takes action when a fault occurs, ensuring the efficient and reliable operation of the fire water system and significantly improving the effectiveness and safety of fire emergency response.

[0027] In one possible implementation, the system first needs to obtain basic physical parameters of the pipe network, including the elastic modulus of the pipe material, the density of the fluid, and the maximum segment length of the pipe network. The elastic modulus describes the pipe material's ability to resist deformation, the fluid density affects the fluid's inertia, and the maximum segment length of the pipe network determines the distance the pressure wave can propagate.

[0028] After determining the pipe elastic modulus and fluid density, the system uses the formula for water hammer wave propagation velocity. According to the formula, the propagation velocity of water hammer waves in a pipe section is directly proportional to the square root of the pipe elastic modulus and inversely proportional to the square root of the fluid density. This relationship enables the system to accurately calculate the propagation velocity of water hammer waves in the pipe network, providing a theoretical basis for setting subsequent sampling intervals.

[0029] Based on the propagation speed of water hammer waves, the system divides the maximum segment length of the pipe network by the propagation speed to determine the critical propagation time of the pressure wave. This critical time is the time required for the pressure wave to propagate from one end point of the pipe network to the other. This time is a key factor affecting the sampling frequency. If the sampling interval is too long, changes in the pressure wave may not be captured in a timely manner, resulting in data distortion or missed samples.

[0030] To avoid distortion of pressure oscillation signals, the system sets the sampling interval to less than half the critical time. This setting ensures a high enough sampling frequency to accurately record changes in pressure waves, thus avoiding signal distortion caused by excessively long sampling intervals. This allows the system to capture subtle pressure fluctuations, ensuring accurate and timely fault detection.

[0031] By precisely calculating the propagation velocity of water hammer waves and the transit time of pressure waves, and setting appropriate sampling intervals, we effectively avoid signal distortion or data loss caused by inappropriate sampling frequencies. This design ensures that the system can capture dynamic changes in pressure waves promptly and accurately during pressure data collection, providing more precise data support for fault detection. Furthermore, a reasonable sampling interval increases the system's sensitivity to subtle anomalies, enhancing its ability to detect faults early, and contributing to the overall safety and reliability of the fire water system.

[0032] In one possible implementation, pressure-flow data pairs are first extracted from the fire water system within continuous time windows under historical normal operating conditions. Each time window contains measured and theoretical pressure values ​​over a period of time, reflecting the operating status of the water system under normal operating conditions. Using this data, the system can establish pressure and flow characteristics under normal conditions, providing a benchmark for subsequent anomaly detection.

[0033] For each time window, the system calculates the residual between the measured and theoretical pressure values ​​and then determines the standard deviation of the residual. The residual standard deviation reflects the degree of fluctuation between the actual system operating pressure and the theoretical expectation and is an important indicator of system stability and error margin. If the system experiences minimal fluctuations under normal operating conditions, the residual standard deviation is small; conversely, the standard deviation is large.

[0034] To improve data stability, the system applies a moving average to the calculated residual standard deviation. This smooths data fluctuations, reduces the impact of short-term mutations, and makes test results more reliable. The moving average serves as the benchmark for the dynamic tolerance band, providing a reference for subsequent error adjustments.

[0035] Under normal operating conditions, the system dynamically adjusts the margin coefficient based on recent false alarm rates. Specifically, when the system detects an increase in the number of consecutive false alarms, it expands the margin coefficient of the tolerance band by a fixed ratio to avoid frequent false alarms caused by overly stringent tolerance settings. This adjustment effectively reduces the false alarm rate and unnecessary alarms. Conversely, when the system experiences a missed alarm, the margin coefficient is reduced based on the proportion of the fault's impact range, thereby reducing missed alarms and improving the system's sensitivity to faults and detection accuracy.

[0036] This dynamic adjustment mechanism effectively adapts to the varying states that arise during system operation by flexibly adjusting the boundary coefficients based on historical data and real-time false alarms and missed alarms, avoiding the false alarm and missed alarm issues that can result from fixed deviation band settings. By expanding the deviation band when false alarms increase and narrowing it when missed alarms occur, the system maintains a high level of early warning accuracy and reduces the frequency of false and missed alarms. Furthermore, this method can be adjusted in real time based on changes in the actual operating environment of the fire water system, making fault detection more accurate, improving system reliability and response efficiency, and enhancing the safety and security capabilities of the entire fire water system.

[0037] In one possible implementation, the weighting factors are assigned based on different node types and specific abnormal signal characteristics. This ensures that different types of faults can be accurately identified and that warnings are assigned with appropriate severity based on the severity of the anomaly, helping the system achieve more intelligent and accurate fault detection.

[0038] Specifically, the fault detection of the pump node relies on the correlation between pressure and current changes: If a sudden pressure drop in the pump system is accompanied by current exceeding the limit, this indicates a possible serious pump failure (such as idling, blockage, or mechanical failure). This failure typically reduces system efficiency and should be assigned a higher alarm level (i.e., the first weighted level). This allocation rule ensures a timely system response to critical failures, preventing serious incidents.

[0039] If the pressure drops slowly and is accompanied by current fluctuations, this usually means that the pump is operating in an unstable state or a minor fault state (such as unstable flow or foreign matter in the pump). This situation has a minor impact on the system, so the weighting factor is set to the second level, indicating that it requires attention but will not immediately trigger a serious alarm.

[0040] Through the above rules, the fault warning of the pump node can be accurately classified according to the degree of change of pressure and current, thereby realizing dynamic alarm for different fault severities.

[0041] Failures at pipeline nodes usually manifest as changes in pressure gradients, especially a sharp drop in upstream pressure: When the upstream pressure gradient drops sharply, the system determines a weighting factor based on the ratio of the gradient's rate of change to the historical maximum. Specifically, when the pressure gradient changes significantly, far exceeding the historical maximum, the system assigns a higher weighting factor. This is often a warning sign of a serious pipeline blockage or rupture, which, if not addressed promptly, could cause the pipeline system to malfunction.

[0042] This rule flexibly adjusts the alarm level by comparing historical data, effectively avoiding false alarms caused by excessive fluctuations in historical data.

[0043] If the delay between the valve opening command and the flow response exceeds a preset threshold multiple, the system linearly increases the weighting factor based on the delay duration. Delays often reflect issues with valve response efficiency. Longer delays may indicate valve malfunctions such as sticking, leakage, or control system instability, so the weighting is dynamically increased based on the severity of the delay. This rule helps the system more accurately detect valve control issues and take timely corrective measures.

[0044] This method provides an accurate, flexible and efficient solution for intelligent fault detection and early warning of fire water systems, effectively improving the safety and reliability of the system.

[0045] In one possible implementation, the system first defines clear judgment criteria for common fault types by establishing a fault feature mapping rule base. Each fault mode has a set of predefined parameter change characteristics, which are obtained through historical data and fault simulation. For example: If the pressure at the abnormal node drops suddenly, the flow rate in the associated downstream branch increases, and the pump current remains within the normal fluctuation range, it can be determined to be a pipeline leak. Pipeline leaks typically cause a sudden drop in pressure and a change in flow rate, but the change in pump current is minimal.

[0046] When the pressure at the abnormal node shows a continuous slow drop, while the total system flow remains constant, and the pump current exceeds the rated operating range, it can be determined that the pump efficiency has decreased. This type of fault indicates that the pump is operating but inefficiently, possibly due to mechanical damage or blockage.

[0047] If the pressure at the abnormal node fluctuates irregularly, the associated branch flow does not respond to valve opening commands, and the pump group current does not show abnormalities, it can be determined that the valve is stuck. A stuck valve will cause pressure fluctuations but will not cause current abnormalities.

[0048] During system operation, the direction of change of parameters such as pressure, flow, and current of abnormal nodes and their directly topologically related nodes is collected in real time. This real-time data is used to match the judgment conditions in the fault feature mapping rule base and perform similarity calculation: The system extracts the sign of the derivative of the pressure trend as the first matching factor. The sign of the derivative reflects the direction of pressure change and is key to determining whether the pressure change matches the fault mode. For example, a sudden or gradual drop in pressure is an important indicator for determining certain faults, such as pipeline leaks or decreased pump efficiency.

[0049] The Boolean value for the flow response direction indicates whether the flow rate changed in the expected direction. For example, in the case of a pipe leak, the flow rate will typically increase, while in other fault types, the flow rate may remain the same or change in some other way.

[0050] The current overlimit status flag is used to identify whether the pump unit is in an abnormal current state. For example, when the pump unit efficiency decreases, the current may exceed the normal range. However, in the case of a pipeline leak or a stuck valve, the current fluctuation is smaller.

[0051] After the system extracts the three matching factors, it performs a similarity calculation. These matching factors are combined based on their respective weights to generate a composite confidence score. If the combined confidence of the three matching factors exceeds a dynamically set similarity threshold, the fault type is confirmed.

[0052] Through this method based on feature mapping and similarity calculation, the system can achieve more accurate fault diagnosis, ensure the safe and stable operation of the fire water system, and warn of potential faults at an early stage to reduce the risk of accidents.

[0053] In one possible implementation, thresholds are initialized based on the average historical fault weights of key nodes. This means that when the system begins operation, the average fault weight of each key node is calculated based on historical fault data. This historical average provides a basis for threshold setting, ensuring that the initial threshold is reasonable, thereby avoiding incorrect judgments of excessively high or low thresholds.

[0054] When a Level III false alarm is detected, the system determines the depth of the fault propagation path. This depth refers to the length of the path from the faulty node to other affected nodes in the system. Path depth is an important indicator of the fault's impact. Based on the depth of the fault propagation path, the system dynamically adjusts the threshold: When the path depth exceeds the set level, the threshold increases proportionally to the depth. A deeper path indicates a more widespread fault impact, requiring the system to increase sensitivity to identify more potential faults. Therefore, the threshold increases proportionally to the path depth.

[0055] When the path depth is less than the set level, the threshold is increased in fixed increments. For shorter fault propagation paths, the system adjusts the threshold in smaller increments to avoid overly increasing sensitivity and affecting system stability and accuracy.

[0056] When a Level III missed fault occurs, it means the system failed to detect or predict the fault in a timely manner. In this case, the system calculates a weighted compensation value based on the actual impact range of the fault. This weighted compensation value reflects the actual impact of the missed fault on the system. Generally, faults with a wider impact range should be assigned a higher compensation value. The system then adjusts the threshold inversely proportional to this compensation value. In other words, as the fault impact range increases, the threshold becomes lower, making the system more sensitive and enabling faster detection of similar faults.

[0057] By dynamically adjusting the threshold in real time, a more flexible and accurate fault detection and early warning method is provided, which can effectively deal with various complex fault scenarios and improve the overall reliability and security of the system.

[0058] In one possible implementation, a stress-strain curve is first obtained by conducting an elastic modulus test on the pipe. In this test, different external pressures (stresses) are applied to the pipe, and the deformation (strain) is measured. The slope of the stress-strain curve represents the material's elastic modulus—the degree to which the material deforms under pressure. The elastic modulus is one of the fundamental parameters for calculating the propagation velocity of water hammer waves, as it directly affects the inertia and elastic response of the water flow in the pipe.

[0059] The density of the medium (water or other liquid) in the pipeline is measured in real time using a fluid density detection device. Fluid density is one of the key factors affecting the propagation speed of water hammer waves. The higher the density, the greater the fluid's inertia, which in turn affects the propagation speed of water hammer waves. Therefore, obtaining accurate fluid density data is crucial.

[0060] The ratio of pipe wall thickness to pipe diameter is input into the elastic deformation correction model, which then applies corrections based on the pipe geometry and outputs a velocity correction factor. Pipe wall thickness and diameter influence the flow velocity and wave propagation within the pipe, which in turn affects the propagation speed of water hammer waves. The correction model adjusts the calculated propagation velocity based on these parameters to suit the specific conditions of different pipes.

[0061] The actual propagation velocity of water hammer waves in a pipe can be calculated by combining the baseline velocity of sound waves in a fluid with the known fluid density, pipe elastic modulus, and correction factor. The baseline velocity of sound waves in a fluid generally depends on the fluid's physical properties (such as density and compressibility), while the propagation velocity of water hammer waves is closely related to the elasticity and geometry of the pipe.

[0062] This calculation process is dynamic and continuously updated based on real-time data (such as fluid density and pipe stress), ensuring that the calculated water hammer wave propagation velocity is consistent with actual conditions. This means the system can adapt to real-time changes, ensuring detection accuracy.

[0063] This calculation method of water hammer wave propagation velocity can not only improve the accuracy of fault detection, but also provide the system with a more efficient early warning function, ensuring the stability and safety of the fire water system in complex operating environments.

[0064] In one possible implementation, the system first quantifies the similarity between historical fault samples and the standard pattern by calculating the cosine similarity between each parameter combination in the sample and the standard pattern. Cosine similarity measures similarity by calculating the cosine of the angle between two vectors, with values ​​closer to 1 indicating higher similarity. Parameter combinations typically include multiple metrics, such as pressure, flow, and temperature. By calculating cosine similarity on this data, it is possible to identify which fault patterns are similar to the standard pattern and which are significantly different.

[0065] Based on the similarity distribution of historical fault samples, the system determines the 75th percentile as the initial similarity threshold. Percentiles reflect the distribution of similarity data, and the 75th percentile typically indicates a high similarity value, meaning that only highly similar samples are considered normal. This setting helps filter out fault samples that deviate significantly from the standard pattern, preventing incorrect patterns from being misclassified as normal faults.

[0066] As new fault samples are added, the system will dynamically adjust the threshold based on the degree of deviation of each new sample from the standard pattern. If the new sample has a high degree of similarity to the standard pattern and is correctly classified, the system will maintain the current similarity threshold to ensure system stability. However, if the new sample has a low degree of similarity to the standard pattern and is incorrectly classified, the threshold will be adjusted: If a sample is misclassified: Depending on the type of misclassification, the system will raise or lower the threshold accordingly. For example, if the misclassification is due to a false negative caused by an overly strict threshold, the system may lower the threshold; conversely, if it is due to a false positive caused by an overly loose threshold, the threshold may be raised.

[0067] Through this dynamic adjustment, the system continuously optimizes threshold settings based on real-time data, adapting to different fault types and operating environments. Each threshold adjustment makes the system more sensitive and accurate in responding to new fault samples, preventing false positives or missed negatives caused by improper threshold settings.

[0068] Through dynamic adjustment of the similarity threshold, the fault detection system can identify fault modes more intelligently and accurately, enhancing the robustness and adaptability of the system and helping to improve the overall operational safety of the fire water system.

[0069] In one possible implementation, when constructing a fault propagation tree, the abnormal node where the fault occurred is first identified. This node represents the critical location of the fault in the system. Then, based on the topology of the pipe network, the tree is expanded layer by layer in the direction opposite to the water flow to identify the pipes, valves, pumps, and other equipment connected to the faulty node, thus forming a fault propagation tree. Each node represents a piece of equipment or pipe section, while each child node represents a piece of equipment or area potentially affected by the fault.

[0070] For each pipeline edge connecting the nodes, calculate its influence weight. This process is achieved through two important parameters: The ratio of the pressure drop change of a pipe section to the total pressure drop change of the system: This item measures the proportion of the pressure drop change of a certain pipe section to the pressure drop change of the entire system, reflecting the impact of the pipe section failure on the system.

[0071] The ratio of the pipe section length to the total length of the traceability path: This parameter reflects the relative length of the pipe section in the entire traceability path during fault propagation. Long pipe sections usually have a greater influence on fault propagation.

[0072] The product of these two ratio values ​​is used as the weight of the edge. An edge with a larger weight indicates that the pipe section or equipment has a higher influence on fault propagation.

[0073] By comparing the edge weights of different paths, the path with the largest weight is selected, representing the most important fault propagation chain. This path can indicate the propagation process of the fault from the initial node to other areas and provide a key reference for subsequent fault diagnosis and response.

[0074] When starting the standby pump: Obtain the water hammer sensitivity coefficient of the pipeline network between the faulty pump and the backup pump: The water hammer sensitivity coefficient measures the response of the fluid in the pipeline to pressure fluctuations. Before starting the backup pump, first determine the water hammer sensitivity coefficient between the faulty pump and the backup pump. This coefficient will be used to calculate the maximum allowable flow rate change.

[0075] Calculate the maximum allowable flow rate change: Based on the water hammer sensitivity coefficient and pipeline characteristics, calculate the maximum allowable flow rate change when the backup pump starts. This value determines the startup speed of the backup pump, avoiding excessive water hammer caused by starting too quickly, thereby reducing pipeline pressure fluctuations.

[0076] Control the standby pump to increase the pressure gradually according to the allowable change rate: According to the calculated flow change rate, control the standby pump to increase the pressure gradually to avoid water hammer phenomenon causing impact on the pipeline.

[0077] When adjusting the valve Reverse-calculate the required flow rate based on the target pressure recovery value at the fault node: When a fault occurs, the system calculates the target pressure to restore the fault node and uses this target to calculate the required flow rate. This ensures that the predetermined pressure level can be restored when the valve is adjusted.

[0078] Calculate the opening adjustment based on the valve flow characteristic curve: Based on the valve flow characteristic curve, calculate the opening adjustment required to achieve the desired flow rate. The valve opening adjustment determines the change in flow rate, ensuring the system can be precisely regulated.

[0079] To avoid system instability caused by large adjustments, valve adjustments are typically made in small increments, with real-time pressure monitoring. The system verifies the progress of pressure recovery based on real-time pressure data to ensure that the pressure returns to the target value.

[0080] Based on the construction of the fault propagation tree and the implementation of Level III equipment linkage, the accuracy and efficiency of fire water system fault detection and early warning can be greatly improved, ensuring the rapid response and safe operation of the fire protection system in emergency situations.

[0081] Correspondingly, an embodiment of the present invention also provides a fire water system fault detection and early warning system based on the fire Internet of Things, including a memory configured to store instructions, a processor configured to call the instructions from the memory and to implement the fire water system fault detection and early warning method based on the fire Internet of Things as described in any embodiment of the present invention when executing the instructions.

[0082] The following is a detailed explanation using examples: The present invention relates to a fault diagnosis and control method for a fire water system, and in particular to a system fault handling method based on the construction of a fault propagation tree model and the coordinated control of Level III equipment. The specific application scenario is the fire water system of a large industrial park, which consists of multiple water pumps, pipes, valves, and firefighting equipment, and has a relatively complex topology. The implementation of the present invention aims to quickly locate the source of system faults through the construction of a fault propagation tree and edge weight analysis, and to avoid water hammer and ensure the efficient operation of the fire protection system through the coordinated control of Level III equipment.

[0083] The fire water system consists of the following equipment: Water pump: 3 units, model JYW-1000, flow rate 1000m 3 / h, rated power 75kW.

[0084] Valves: 10, model DVC-50, flow control range 50~200m 3 / h, the maximum opening allowed is 90°.

[0085] Pipeline: 10 km of steel pipe, 200 mm diameter, maximum flow rate 1200 m 3 / h.

[0086] Sensor: Pressure sensors are installed at the pipeline inlet, outlet and key nodes, with a sampling frequency of 1 Hz and a measurement accuracy of ±0.5%.

[0087] The parameters of these devices are monitored in real time and used as input for fault diagnosis and control.

[0088] At a certain moment in the embodiment of the present invention, a pipe rupture event occurs in the fire water system. Based on the system topology, a fault propagation tree can be constructed. The fault propagation tree construction algorithm is as follows: System topology: Each node in the diagram represents a device (such as a pump, valve, pipe, etc.), and each edge represents the connection between devices.

[0089] Fault node: Assume that the rupture occurs in pipeline 3 (numbered C3).

[0090] Using the breadth-first search (BFS) algorithm, we start from node C3 where the fault occurs and gradually expand outward until all affected devices are marked as fault-related.

[0091] The weight of each edge Calculated according to the following formula: ; in, : Change in pressure drop in the pipe section (unit: Pa). : Change in total system pressure drop (unit: Pa). : Length of pipe section (unit: m). : The total length of the fault path (unit: m).

[0092] Assume that the total system pressure drop change is 300 Pa, the pressure drop change of pipe 3 is 50 Pa, the length of pipe 3 is 100 m, and the total length of the fault path is 500 m. Then the weight of this edge is: ; Based on the above calculations, the weights of all relevant edges are calculated. The system selects the propagation chain with the largest weight as the main fault path. Assume that the calculated main fault propagation chain is: Main transmission chain: .

[0093] The total weight of this propagation chain is 0.1, indicating that this path has the greatest impact on the system and needs to be restored first.

[0094] In order to avoid water hammer and ensure the system is restored to operation, it is necessary to adjust the system through the linkage control of Class III equipment (such as standby pumps and valves). The specific process is as follows: Conditions for starting the standby pump: When the pipeline pressure is lower than the set threshold (for example, the set value is 1.0 MPa), the system starts the standby pump.

[0095] Flow calculation formula when the standby pump starts: ; in, : Current pipeline pressure. : Minimum pressure. : Maximum pressure. : Maximum flow rate.

[0096] In the embodiment of the present invention, the current pressure is 0.8 MPa, the minimum pressure is 0.5 MPa, the maximum pressure is 1.2 MPa, and the maximum flow rate is 1000 m 3 / h, the flow rate of the standby pump is: ; When starting the standby pump, the system needs to close valve V5 to reduce the sharp change of pipeline pressure. The flow adjustment of the valve is carried out according to the following formula: ; Assume that the standby pump flow rate is 428.57m 3 / h, the current pressure is 0.8MPa, the maximum pressure is 1.2MPa, and the regulating flow of the valve is: ; This formula ensures that the valve adjusts the flow rate according to actual needs and avoids water hammer.

[0097] Through this linked control, the system can simultaneously adjust the valve opening when the pump is started, maintaining stable pipeline pressure and preventing water hammer. Compared to traditional manual adjustment methods, the automated control system of this invention offers faster response speeds, higher control accuracy, and the ability to dynamically adjust in real time, reducing errors caused by manual operation.

[0098] Before implementation, using traditional manual valve adjustment methods, system recovery time was 30 minutes and the water hammer rate was 20%. After implementing this invention, the automated control system was able to recover within 10 minutes and completely avoid water hammer. Comparative experiments showed that the automated control system significantly improved fault recovery speed and system stability compared to traditional methods.

[0099] This example describes in detail how to utilize a fault propagation tree model and Level III device linkage control methods for fault diagnosis and control. By applying specific algorithms, formulas, and parameter values, this approach ensures rapid fault location and efficient system recovery when a fault occurs. Compared to traditional methods, this approach significantly improves system response speed and stability, demonstrating outstanding practical application results.

[0100] The present invention encompasses any alternatives, modifications, equivalents, and solutions that fall within the spirit and scope of the present invention. To provide a thorough understanding of the present invention, specific details are described in detail below in connection with the preferred embodiments of the present invention, but those skilled in the art will be able to fully understand the present invention without these detailed descriptions. Furthermore, to avoid unnecessary confusion regarding the essence of the present invention, well-known methods, processes, procedures, components, and circuits have not been described in detail.

[0101] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A fire water system fault detection and early warning method based on the fire Internet of Things is characterized by: include: Step 1: Pressure data is periodically collected using pressure sensors deployed at the fire pump outlet, at the branch point of the main pipe network, and at the most unfavorable end. The total system flow is collected using an electromagnetic flowmeter at the pump group outlet manifold. Branch flow is collected using an ultrasonic flowmeter behind the zone control valve. The valve opening status and pump operating status are respectively obtained using a valve opening sensor and a pump group current sensor. Step 2: Establish a pressure transmission relationship based on the network topology node-edge relationship model, and generate the theoretical pressure expectation value and dynamic tolerance band of each node according to the current valve opening, pump status and historical normal operating condition data; Step 3: When the measured pressure deviates from the expected value of the theoretical pressure and exceeds the dynamic tolerance band, mark the abnormal node and reversely construct the fault propagation tree along the topology model, and assign weight factors to the associated nodes according to the fault type; Step 4: Trigger a hierarchical warning based on the number of abnormal nodes, fault propagation path, and accumulated weight factors: A single node abnormality and unspread weight triggers a Level I alarm; Adjacent nodes are abnormal and the fault pattern matches, triggering a Level II alarm; When a key node is abnormal and the accumulated weight value exceeds the adaptive threshold, level III equipment linkage is triggered.

2. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 1 is characterized in that: The method for determining the time interval for periodically collecting pressure data in step 1 includes: Obtain the elastic modulus of the pipe, fluid density, and maximum segment length of the pipe network; Calculate the propagation velocity of the water hammer wave in the pipe section, where the propagation velocity is proportional to the square root of the elastic modulus of the pipe and inversely proportional to the square root of the fluid density; Divide the maximum segment length of the pipe network by the water hammer wave propagation velocity to obtain the critical time for pressure wave transmission; Set the sampling interval to less than half of the critical time to avoid distortion of the pressure oscillation signal.

3. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 1 is characterized in that: The method for generating the dynamic tolerance band in step 2 includes: Extract pressure-flow data pairs in continuous time windows under historical normal operating conditions; Calculate the residual standard deviation of the measured pressure value and the theoretical value in each time window; Based on the moving average of the residual standard deviation, the boundary coefficient is dynamically adjusted according to the recent false alarm rate statistics: When the number of consecutive false alarms increases, the boundary coefficient is expanded at a fixed ratio; When an underreporting event occurs, the boundary coefficient is shrunk in proportion to the fault impact range.

4. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 1 is characterized in that: The weight factor allocation rules in step 3 include: The pump node weight factor is assigned based on the correlation between the pressure drop amplitude and the current change: If the pressure drop is accompanied by current exceeding the limit, the first weight level is assigned; If the pressure drop is accompanied by current fluctuation, the second weight level is assigned; The pipeline node weight factor is assigned according to the pressure gradient change rate: When the upstream pressure gradient drops sharply, the value is assigned according to the ratio of the gradient change rate to the historical maximum value; The valve node weight factor is assigned based on the delay time between the opening command and the flow response: When the delay time exceeds the preset threshold multiple, the weight is increased linearly according to the delay time.

5. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 1 is characterized in that: The verification method for fault pattern matching in step 3 includes: a: Build a fault feature mapping rule base, where: The criteria for determining a pipeline leak are: a sudden drop in pressure at the abnormal node, an upward trend in the flow rate of its associated downstream branch, and the pump group current is within the normal fluctuation range; The criteria for determining pump group efficiency degradation are: the abnormal node pressure shows a continuous slow drop, the total system flow remains unchanged, and the pump group current exceeds the rated operating range; The conditions for determining valve sticking are: the pressure at the abnormal node fluctuates irregularly, the flow rate of its associated branch does not respond to the valve opening command, and the pump group current does not show abnormalities; Collect the real-time parameter change direction of the abnormal node and its directly topologically related nodes. When the parameter combination matches the judgment conditions in any fault feature mapping rule base, perform similarity calculation: b: Extract the derivative sign of the pressure change trend as the first matching factor; c: Extract the Boolean value of the traffic response direction as the second matching factor; d: Extract the current over-limit status flag as the third matching factor; When the joint confidence of the three matching factors exceeds the dynamic similarity threshold, the fault type is confirmed.

6. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 1 is characterized in that: The method for dynamically adjusting the adaptive threshold in step 4 includes: The initialization threshold is based on the average weight of historical failures of key nodes; When a level III false alarm occurs, obtain the fault propagation path depth; When the path depth is greater than the set level, the threshold is increased in proportion to the depth value; When the path depth is less than the set level, the threshold is increased by a fixed step value; When a level III missed alarm occurs, the weight compensation value is calculated based on the actual impact range of the fault, and the threshold is adjusted down in inverse proportion to the compensation value.

7. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 2 is characterized in that: The calculation process of water hammer wave propagation velocity includes: Obtain the slope of the stress-strain curve through the pipe elastic modulus test experiment; Obtain real-time medium density through a fluid density detection device; Input the ratio of pipe wall thickness to pipe diameter into the elastic deformation correction model and output the wave velocity correction coefficient; The actual propagation speed is calculated based on the reference speed of the sound wave in the fluid.

8. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 5 is characterized in that: Methods for determining the similarity threshold include: Calculate the cosine similarity between the parameter combination in the historical fault samples and the standard pattern; The 75th percentile of the similarity distribution is used as the initial threshold; Each time a fault sample is added, the threshold is adjusted according to the degree of deviation between the sample and the standard pattern: If the sample is correctly classified and the similarity is higher than the current threshold, the threshold remains unchanged; If a sample is misclassified, lower or raise the threshold according to the type of misclassification.

9. The fire water system fault detection and early warning method based on the fire Internet of Things according to claim 1 is characterized in that: The method for constructing the fault propagation tree in step 3 includes: Taking the abnormal node as the root node, expand the child nodes layer by layer in the reverse direction of the water flow along the topology model; Calculate the influence weight of each side: Obtain the ratio of the pipe section pressure drop change to the total system pressure drop change; Get the ratio of the pipe segment length to the total length of the traceability path; The product of the two proportions is used as the edge weight; Select the path with the largest edge weight as the main fault propagation chain; The execution of Level III equipment linkage in step 4 includes: When starting the standby pump: Obtain the water hammer sensitivity coefficient of the pipe network between the faulty pump and the backup pump; Calculate the maximum allowable flow rate change rate based on the sensitivity coefficient; Control the standby pump to increase the pressure gradually according to the allowable change rate; When adjusting the valve: Reversely calculate the required flow rate based on the target value of pressure recovery at the fault node; Calculate the opening adjustment amount according to the valve flow characteristic curve; Adjust in segments and verify the pressure recovery progress in real time.

10. The fire water system fault detection and early warning system based on the fire Internet of Things is characterized by: It includes a memory configured to store instructions, a processor configured to call the instructions from the memory and to implement the fire water system fault detection and early warning method based on the fire Internet of Things as described in any one of claims 1-9 when executing the instructions.

Images