A sewer robot toxic gas threshold closed-loop control method and system

Abstract

The application discloses a kind of sewer robot toxic gas threshold closed-loop control method and system, it is related to limited space gas safety monitoring technical field, for solving the control algorithm based on fixed air exchange trigger threshold and direct back-off release rule when air exchange is driven by patrol inspection, it is easy to misjudge the short-time back-off of local concentration after air exchange as risk release, thereby causing the closed-loop oscillation problem of fan repeated start-stop and early warning repeated switching, by collecting target gas original concentration value, environmental temperature value, sampling time and current position, determine main control gas and air exchange trigger time;After air exchange, the background concentration result is extracted, the dynamic release threshold is generated, the back-off amplitude result and the recheck deviation result are calculated, the release credible model is established, the release probability threshold and the number of continuous satisfaction are made according to the release credible probability, release probability threshold and the number of continuous satisfaction, and the model coefficient, release proportion coefficient and release probability threshold are updated online in combination with the label result in confirmation window.

Description

Technical Field

[0001] This invention relates to the field of confined space gas safety monitoring technology, and more specifically, to a closed-loop control method and system for toxic gas threshold of a drainage pipeline robot. Background Technology

[0002] In scenarios where drainage pipeline robots inspect and activate fans for ventilation, existing control methods typically collect target gas concentrations at fixed sampling intervals and initiate ventilation when the concentration reaches a preset ventilation trigger threshold. During ventilation, the system then determines whether to deactivate ventilation and switch the warning status based on the concentration drop at the current location. Combining the settings for the ventilation trigger threshold, ventilation trigger time, current location concentration drop, and release determination process in this application, it can be seen that existing technologies mainly revolve around a control approach based on fixed threshold triggering and direct deactivation upon drop.

[0003] The existing technology has the following shortcomings:

[0004] The gas distribution within the confined space of drainage pipes is easily affected by the action of fans, robot movement, disturbances at branch inlets, and local backflow. A short-term decrease in concentration at the current location after ventilation does not necessarily indicate that the risk has been truly eliminated. Combining the background concentration results, dynamic release threshold, neighboring location verification, release reliability model, consecutive satisfaction count, post-release confirmation window, and online update mechanism introduced in this application, it can be concluded that existing control algorithms based on fixed thresholds and direct drop-off elimination rules are prone to misjudging a short-term drop in local concentration after ventilation as a risk elimination, leading to repeated fan start-stops and warning switching, forming a closed-loop oscillation.

[0005] To address the above problems, this invention proposes a solution. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a closed-loop control method and system for toxic gas threshold control of a drainage pipeline robot, in order to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] A closed-loop control method for toxic gas threshold of a drainage pipeline robot, comprising the following steps;

[0009] Step S1: Collect the original concentration value of the target gas, the ambient temperature value, the sampling time and the current location according to a fixed sampling period. Perform temperature compensation and smoothing on the original concentration value, calculate the normalized risk result, compare the normalized risk results of each target gas, and determine the main control gas, the ventilation trigger threshold and the ventilation trigger time.

[0010] Step S2: Capture a reference window around the current ventilation event, calculate the background concentration result, generate a dynamic release threshold by combining the safety margin and release ratio coefficient, and calculate the drop amplitude result and the verification deviation result;

[0011] Step S3: When the current concentration level is not higher than the dynamic release threshold, calculate the threshold margin result, establish a release confidence model based on the threshold margin result, the fall amplitude result and the verification deviation result, output the release confidence probability, and make a release judgment by combining the release probability threshold and the number of consecutive satisfactions.

[0012] Step S4: After the ventilation is turned off, track the current concentration level in the confirmation window, generate a label result, and update the release confidence model coefficient, release ratio coefficient and release probability threshold online based on the label result.

[0013] In a preferred embodiment, step S1 includes the following:

[0014] Hydrogen sulfide, methane, ammonia, and carbon monoxide are detected according to a fixed sampling cycle, and the corresponding sampling time and current location are recorded.

[0015] The current location includes the entry section, narrow straight pipe section, and branch inlet section that the robot passes through as it moves along the preset inspection route within the limited space of the drainage pipe.

[0016] The original concentration value is temperature-compensated based on the temperature compensation coefficient corresponding to the target gas and the reference temperature value to form a compensated concentration value.

[0017] The compensation concentration value is smoothed using the median method to form the current concentration level;

[0018] The normalized risk result is calculated based on the ratio of the current concentration level of each target gas to its corresponding ventilation trigger threshold.

[0019] At each sampling moment, the target gas with the highest normalized risk result is selected as the current master gas;

[0020] If the current concentration level of the main control gas reaches or exceeds its ventilation trigger threshold, the ventilation state is entered, and this moment is recorded as the ventilation trigger moment. Before the end of this ventilation event, release assessment and parameter updates are performed around the current main control gas.

[0021] In a preferred embodiment, step S2 includes the following:

[0022] After the ventilation trigger moment is formed, a reference window containing a preset number of sampling points is extracted from before the ventilation trigger moment. The average value of the current concentration level in the reference window is calculated to form the background concentration result.

[0023] A dynamic release threshold is generated based on the background concentration result and the ventilation trigger threshold. The dynamic release threshold is the larger of the background constraint formed by the background concentration result plus the safety margin and the global constraint formed by the ventilation trigger threshold multiplied by the release ratio coefficient.

[0024] The magnitude of the drop is the difference between the current concentration level at the moment the ventilation is triggered and the current concentration level at the current moment, divided by the ventilation trigger threshold.

[0025] When the current concentration level drops below the dynamic release threshold for the first time, the robot moves forward a preset verification distance along the current inspection direction and collects the verification concentration results of the neighboring locations. The verification deviation result is the difference between the verification concentration result and the current concentration level at the current location divided by the ventilation trigger threshold.

[0026] The preset verification distance is set according to the pipe section type.

[0027] In a preferred embodiment, step S3 includes the following:

[0028] When the current concentration level is not higher than the dynamic release threshold, the threshold margin result is calculated. The threshold margin result is the difference between the dynamic release threshold and the current concentration level divided by the dynamic release threshold.

[0029] The release confidence model takes the threshold margin result, the fallback magnitude result, and the verification deviation result as inputs, and outputs the release confidence probability in the form of logistic regression. The increase of the threshold margin result and the fallback magnitude result increases the release confidence probability, while the increase of the verification deviation result decreases the release confidence probability.

[0030] The release decision to lift ventilation is made only when the current concentration level continuously reaches the preset number of consecutive satisfactions, the number of sampling points is not higher than the dynamic release threshold, and the corresponding release confidence probability is not lower than the release probability threshold.

[0031] Before deployment, the release trust model is trained based on historical ventilation events. During training, whether the main control gas in the confirmation window has reached the ventilation trigger threshold again is used as the basis for labeling positive and negative samples.

[0032] In a preferred embodiment, step S4 includes the following:

[0033] After the ventilation is turned off, the current concentration level of the main control gas continues to be tracked within a preset confirmation window;

[0034] If the current concentration level of the main control gas does not reach the ventilation trigger threshold again within the confirmation window, the label result will be recorded as a successful release.

[0035] If the ventilation trigger threshold is reached again within the confirmation window, the tag result will be recorded as release failure.

[0036] Based on the labeling results, the model coefficients of the released reliable model are updated online using a single-step gradient method. When the release fails, the model coefficients are corrected in a more conservative direction, and when the release is successful, the model coefficients are strengthened.

[0037] The release ratio coefficient is corrected online. When the release fails, the release ratio coefficient is reduced. When the number of consecutive successes reaches the consecutive success threshold, the release ratio coefficient is increased. The update of the release ratio coefficient is constrained by the minimum boundary and the maximum boundary.

[0038] The release probability threshold is adjusted online. When the release fails, the release probability threshold is increased. When the number of consecutive successes reaches the consecutive successes threshold, the release probability threshold is decreased. The update of the release probability threshold is constrained by the minimum boundary and the maximum boundary.

[0039] Online updates maintain parameter sets separately for each pipe segment scenario.

[0040] A closed-loop control system for toxic gas threshold of a drainage pipeline robot includes: a gas detection trigger module, a dynamic threshold fall-off module, a release reliable release module, and a parameter update closed-loop module, with signal connections between the modules;

[0041] Gas detection trigger module: Collects the original concentration value of the target gas, ambient temperature value, sampling time and current location according to a fixed sampling period, performs temperature compensation and smoothing on the original concentration value, calculates the normalized risk result, compares the normalized risk results of each target gas, and determines the main control gas, ventilation trigger threshold and ventilation trigger time.

[0042] Dynamic threshold fall-off module: It captures a reference window around the current ventilation event, calculates the background concentration result, generates a dynamic release threshold by combining the safety margin and release ratio coefficient, and calculates the fall-off magnitude result and the verification deviation result.

[0043] Release Confidentiality Module: When the current concentration level is not higher than the dynamic release threshold, calculate the threshold margin result, establish a release confidence model based on the threshold margin result, the fall amplitude result and the verification deviation result, output the release confidence probability, and make a release judgment by combining the release probability threshold and the number of consecutive satisfactions;

[0044] Parameter update closed-loop module: After ventilation is turned off, the current concentration level is tracked in the confirmation window, a label result is generated, and the release confidence model coefficient, release ratio coefficient and release probability threshold are updated online based on the label result.

[0045] The technical effects and advantages of the toxic gas threshold closed-loop control method for drainage pipeline robots of the present invention are as follows:

[0046] This invention integrates target gas concentration acquisition, main control gas determination, ventilation triggering, release judgment, and online parameter updates into a unified closed-loop control framework. During the ventilation release phase, it no longer relies solely on instantaneous drop results under a fixed threshold for judgment. Instead, it generates a dynamic release threshold by combining background concentration results, and makes a release judgment by combining drop amplitude results, verification deviation results, and release confidence probability. Simultaneously, after release, it updates the release confidence model coefficient, release ratio coefficient, and release probability threshold online through the label results in the confirmation window. This reduces the likelihood of misjudging short-term local concentration drops as risk clearance, lowers the frequency of repeated fan start-ups and shutdowns and repeated warning switching, and gradually aligns control parameters in different pipe segment scenarios with the corresponding gas evolution characteristics. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the closed-loop control method for toxic gas threshold of a drainage pipeline robot according to the present invention.

[0048] Figure 2 This is a schematic diagram of a closed-loop control system module for a toxic gas threshold of a drainage pipeline robot according to the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Example: Please refer to Figures 1-2 As shown, this invention discloses a closed-loop control method for toxic gas threshold in a drainage pipeline robot, comprising the following steps:

[0051] Step S1: Collect the original concentration value of the target gas, the ambient temperature value, the sampling time and the current location according to a fixed sampling period. Perform temperature compensation and smoothing on the original concentration value, calculate the normalized risk result, compare the normalized risk results of each target gas, and determine the main control gas, the ventilation trigger threshold and the ventilation trigger time.

[0052] Step S2: Capture a reference window around the current ventilation event, calculate the background concentration result, generate a dynamic release threshold by combining the safety margin and release ratio coefficient, and calculate the drop amplitude result and the verification deviation result;

[0053] Step S3: When the current concentration level is not higher than the dynamic release threshold, calculate the threshold margin result, establish a release confidence model based on the threshold margin result, the fall amplitude result and the verification deviation result, output the release confidence probability, and make a release judgment by combining the release probability threshold and the number of consecutive satisfactions.

[0054] Step S4: After the ventilation is turned off, track the current concentration level in the confirmation window, generate a label result, and update the release confidence model coefficient, release ratio coefficient and release probability threshold online based on the label result.

[0055] In step S1, the original concentration value of the target gas, the ambient temperature value, the sampling time, and the current location are collected according to a fixed sampling period. Temperature compensation and smoothing are performed on the original concentration value. The normalized risk result is calculated, and the normalized risk results of each target gas are compared to determine the main control gas, the ventilation trigger threshold, and the ventilation trigger time. Specific details include:

[0056] Hydrogen sulfide, methane, ammonia, and carbon monoxide are detected according to a fixed sampling cycle, and the corresponding sampling time and current location are recorded.

[0057] The corresponding locations include the built-in entry section, the narrow straight pipe section, and the branch inlet section where the robot moves along the preset inspection route within the limited space of the drainage pipe.

[0058] The detection is performed in a shorter period, preferably 2 seconds as a sampling period, and the original concentration value of the target gas is collected in each sampling period. and ambient temperature value Different gas detection elements exhibit response shifts when ambient temperature changes. Temperature compensation is applied to the original concentration value to generate a compensated concentration value. The formula for the compensation concentration value can be expressed as: ;in, Target gas At any moment The compensation concentration value, Target gas At any moment The original concentration value, For a moment Ambient temperature value, Target gas Temperature compensation coefficient, This is a reference temperature value;

[0059] It should be noted that the sampling period of 2 seconds is based on the evolution characteristics of gas concentration within the confined space of the drainage pipe: the gas in the pipe is affected by the ventilation of the fan and local backflow, and the time constant of the concentration change is about 5-10 seconds. The 2-second sampling period can realize real-time tracking of concentration changes, and will not cause data redundancy or increase the computational load of the robot due to excessive sampling frequency. At the same time, the 2-second sampling period matches the response speed of the gas detection element of the existing drainage pipe robot. The response time of the detection element is ≤1 second, which can ensure the validity of the sampling data.

[0060] The compensated concentration value is further smoothed to form the current concentration level. To suppress localized jitter, instantaneous spikes, and single-point anomalies caused by robot movement, the three-point median method is preferred. The formula for the current concentration level can be expressed as: ;in, Target gas At any moment The current concentration level, This is for median operations;

[0061] After obtaining stable current concentration levels, warning levels are determined for each target gas based on these levels, and ventilation triggering is established. To ensure comparability among various gases within the same control framework, the ventilation triggering threshold for each target gas is first determined. Calculate normalized risk results The normalized risk result can be expressed by the following formula: ;in, Target gas At any moment Normalized risk results Target gas At any moment The current concentration level, Target gas The ventilation trigger threshold;

[0062] At each sampling moment, the target gas with the highest normalized risk result is selected as the current master gas. If the current concentration level of the master gas reaches or exceeds its ventilation trigger threshold, ventilation is initiated, and this moment is recorded as the ventilation trigger moment. Before the end of this ventilation event, release assessments and parameter updates will be conducted focusing on the current main control gas.

[0063] For example, if the current concentration levels of hydrogen sulfide, ammonia, carbon monoxide, and methane are all detected to be 0.65, 25, 80, and 25% of the lower explosive limit, respectively, and the ventilation trigger thresholds for hydrogen sulfide, ammonia, carbon monoxide, and methane are set to 0.51, 30, 200, and 30% of the lower explosive limit, respectively, then the normalized risk results for the four gases are approximately 1.27, 0.83, 0.40, and 0.83. In this case, hydrogen sulfide is identified as the current controlling gas, and it enters the ventilation state because it exceeds the ventilation trigger threshold.

[0064] It should be noted that the temperature compensation coefficient The temperature compensation coefficient can be set based on the calibration results of the standard gas before deployment. For hydrogen sulfide, it can be 0.003; for methane, 0.002; for ammonia, 0.004; and for carbon monoxide, 0.0025. These values ​​are only preferred examples and can be adjusted under different pipeline networks and sensing element conditions. Another example is the ventilation trigger threshold. This can be configured according to on-site safety procedures. Preferably, hydrogen sulfide can be set to 0.51, ammonia to 30, carbon monoxide to 200, and methane to 30% of the lower explosive limit. By completing temperature compensation, smoothing, main control gas selection, and ventilation triggering in the same step, the current concentration level, main control gas, ventilation triggering threshold, and ventilation triggering time can be output to subsequent steps with a unified result caliber.

[0065] By the end of this step, the basic results upon which the entire closed-loop control process depends have been formed, including the current concentration level, the main control gas result, the ventilation trigger threshold, and the ventilation trigger time. The scattered measurement, compensation, smoothing, comparison, and triggering actions during the inspection process are unified into an operating benchmark that can be directly called upon later, enabling subsequent steps to continue to generate dynamic release thresholds, determine release reliability, and update online parameters based on the same set of results.

[0066] In step S2, a reference window is captured around the current ventilation event, the background concentration result is calculated, and a dynamic release threshold is generated by combining the safety margin and the release ratio coefficient. The result of the drop amplitude and the verification deviation are also calculated. Specific details include:

[0067] After a ventilation state has been established and the ventilation trigger time is recorded, a dynamic release threshold is generated around this ventilation event. At the same time, a small number of fallback behavior results are extracted, which are sufficient to support subsequent probability judgments. The results are used to determine whether the current concentration level has fallen back to an assessable range and whether the fallback looks like a real risk reduction.

[0068] At the moment of air exchange triggering Once formed, a short reference window is captured from before the air exchange trigger time to generate the background concentration results. Preferably, 10 sampling points can be used as a reference window, corresponding to approximately 20 seconds of sampling history. The background concentration result formula can be expressed as: ;in, The background concentration results of the control gas in this ventilation event are shown. The number of sampling points within the reference window. For a moment The current concentration level;

[0069] After obtaining the background concentration results, a dynamic release threshold is generated based on the background concentration results and the ventilation trigger threshold. Using a single-value threshold reduces the number of parameters and improves deployment clarity, the dynamic release threshold formula can be expressed as: ;in, The main control gas at time The dynamic release threshold, For background concentration results, The safety margin corresponding to the control gas. The release ratio coefficient corresponding to the main control gas. The ventilation trigger threshold corresponding to the main control gas;

[0070] The above formula reflects two constraints: the background constraint that the release assessment cannot be lower than the background concentration plus the safety margin, and the global constraint that the release assessment cannot be higher than the ventilation trigger threshold and then proportionally decreases. The larger of the two values ​​is taken.

[0071] For example, for hydrogen sulfide, if the average current concentration level of the 10 sampling points prior to a certain ventilation trigger is 0.12, the safety margin is... Set to 0.15 to release the scaling factor. Set to 0.6, ventilation trigger threshold If set to 0.51, the dynamic release threshold is... The result was 0.306. At this point, the release confidence assessment will only proceed if the current concentration level of hydrogen sulfide falls to 0.306 or below.

[0072] After establishing the dynamic release threshold, two pullback behavior results are extracted to determine the authenticity of the pullback: the pullback magnitude result. and verification deviation results ;

[0073] Result of the pullback This formula is used to depict how much the control gas has decreased from the moment the ventilation was triggered to the present moment. The magnitude of the decrease can be expressed by the following formula: ;in, The main control gas at time The magnitude of the decline, The current concentration level at the moment the air exchange is triggered. The current concentration level at the current moment. The air exchange trigger threshold for the main control gas.

[0074] Verification of deviation results This tool utilizes the robot's spatial mobility to verify the spatial consistency of the current concentration level's decline. Specifically, when the current concentration level first drops below the dynamic release threshold, the robot is controlled to move forward a short distance along the current inspection direction and collect the verification concentration results from nearby locations. The verification deviation result is calculated and can be expressed by the following formula: ;in, The main control gas at time The results of the review deviation, This is a verification of the concentration results at nearby locations. This represents the current concentration level at the current location. The air exchange trigger threshold for the main control gas;

[0075] If the verification deviation is small, it means that the current concentration levels at the current location and the adjacent locations are similar, and this decline is more likely to be a genuine spatial decline; if the verification deviation is large, it means that although the concentration at the current location has decreased, the concentration at the adjacent locations has not decreased synchronously, and this decline is more likely to be a local short-term dilution caused by the action of the fan.

[0076] It should be noted that the verification distance of adjacent locations can be set according to the pipe segment type. For example, in a narrow straight pipe segment, the verification distance can be 0.5 meters; in a branch merging section, the verification distance can be 0.3 meters. The former is suitable for long-distance unidirectional gas migration scenarios, while the latter is suitable for confluence areas with more significant local disturbances.

[0077] For example, safety margin and release ratio coefficient Alternatively, settings can be made for each target gas. The safety margin for hydrogen sulfide can be 0.15, and the release ratio coefficient can be 0.6; the safety margin for ammonia can be 5, and the release ratio coefficient can be 0.55; the safety margin for carbon monoxide can be 20, and the release ratio coefficient can be 0.6.

[0078] For example, if in a certain hydrogen sulfide ventilation event, the current concentration level at the time of ventilation triggering is 0.8, and the current concentration level at the current moment is 0.29, then the resulting decrease would be approximately... That is, approximately 2.96; if the concentration of hydrogen sulfide at a nearby location is measured to be 0.33 after the robot moves forward for verification, then the verification deviation result is approximately... The value is approximately 0.08. This set of results indicates that this pullback was not only significant in magnitude, but also that the difference between the current and nearby locations was small, making it more likely a genuine pullback.

[0079] This resulted in three key results required for subsequent release assessment: dynamic release threshold, drop magnitude result, and verification deviation result. The phenomenon of concentration decrease was broken down into two levels: whether it has fallen below the dynamic release threshold and whether this drop is spatially consistent. This provides a concise and targeted basis for calculating the release reliability probability in subsequent steps.

[0080] In step S3, when the current concentration level is not higher than the dynamic release threshold, the threshold margin result is calculated. Based on the threshold margin result, the fallback magnitude result, and the verification deviation result, a release confidence model is established, and the release confidence probability is output. A release determination is then made by combining the release probability threshold and the number of consecutive satisfactions. Specific details include:

[0081] After the dynamic release threshold, the drop amplitude results, and the verification deviation results have been established, a release credibility model is built, and the decision on when to lift the ventilation state is made based on this model. Ventilation is lifted only when the current concentration level is lower than the dynamic release threshold and the state below the threshold has a sufficiently high degree of credibility in terms of behavior; otherwise, ventilation is maintained.

[0082] Determine whether the current concentration level is no higher than the dynamic release threshold. This indicates that the current concentration level is still outside the release assessment range, so we will continue to maintain the ventilation state and will not enter the release confidence model.

[0083] Only when Then, continue calculating the threshold margin result. The threshold margin result can be expressed by the following formula: ;in, The main control gas at time The threshold margin results To dynamically release the threshold, This represents the current concentration level.

[0084] In this embodiment, the release confidence model uses three input results, including the threshold margin result. Result of the decline and the results of the review deviation The three results respectively reflect the degree to which the current concentration level is below the dynamic release threshold, the overall decline from the trigger time to the current time, and the consistency between the current location and the neighboring locations. Based on these three results, a release confidence probability is established. The release probability can be expressed by the following formula: ;in, The main control gas at time The release probability of credibility. Target gas The bias coefficient, These are the model coefficients for the threshold margin result. The model coefficients represent the magnitude of the pullback. To verify the model coefficients of the deviation results, This is the threshold margin result. The result is the magnitude of the decline. To verify the deviation results;

[0085] When the threshold margin and the pullback magnitude increase, the release reliability probability increases; when the verification bias increases, the release reliability probability decreases. The model directly reflects the positive and negative effects of the three factors on the release reliability.

[0086] It should be noted that the release reliability model needs to be trained before deployment. During training, a separate training sample set is established for each target gas. For each training sample, the moment when the current concentration level first drops below the dynamic release threshold is found in the historical ventilation events. The threshold margin, drop magnitude, and verification deviation results corresponding to that moment are calculated and used to form a three-dimensional input. Then, a confirmation window is set after that moment. If the main control gas does not reach the ventilation trigger threshold again within the confirmation window, the sample is marked as a positive sample and recorded as 1. If the ventilation trigger threshold is reached again within the confirmation window, the sample is marked as a negative sample and recorded as 0. Subsequently, the above logistic regression model is trained using the conventional gradient descent algorithm to obtain the bias coefficient and the model coefficients of each input result.

[0087] For example, for hydrogen sulfide, at least 300 ventilation events can be collected as training samples; the confirmation window is preferably set to 90 seconds; after training, a set of preferred model coefficients can be obtained, such as the bias coefficient. Set to -0.85, threshold margin coefficient Take 2.2, the pullback amplitude coefficient Take 1.35 and verify the deviation coefficient. Taking 1.6, the above values ​​indicate that in the hydrogen sulfide scenario, the greater the margin of the current concentration level relative to the dynamic release threshold, the more sufficient the overall decline after ventilation, and the more consistent the verification of adjacent locations, the higher the probability of release.

[0088] After obtaining the reliable probability of release, instead of immediately releasing the product upon fulfillment of the requirement, a release probability threshold is further introduced. and the number of consecutive satisfactions Only when the current concentration level is continuous Each sampling point is no higher than the dynamic release threshold, and this Ventilation is only lifted when the release confidence probability corresponding to each sampling point is not lower than the release probability threshold. This release determination can be expressed by the following formula: ;in, , The main control gas at time The release determination result, Target gas The release probability threshold, Target gas The number of consecutive satisfactions;

[0089] For example, if the hydrogen sulfide release probability threshold is set to 0.72 and the number of consecutive satisfactions is set to 3, it means that the fan will only stop when the current concentration level is not higher than the dynamic release threshold and the release confidence probability is not lower than 0.72 at least 3 consecutive sampling points. Since the sampling period is 2 seconds, satisfying the condition 3 times means satisfying the release condition for 6 consecutive seconds, avoiding misunderstandings caused by single-point jumps or instantaneous disturbances.

[0090] For example, if in a certain hydrogen sulfide ventilation event the current concentration level is 0.29 and the dynamic release threshold is 0.306, then the threshold margin result is approximately The value is approximately 0.502. If the concentration at the moment of ventilation triggering is 0.8, the drop will be approximately 2.96. If the concentration at a nearby location is 0.33, the verification deviation will be approximately 0.08. Substituting this three-dimensional input into the hydrogen sulfide model above, the release confidence probability is approximately 0.81. If the release confidence probability at the next three consecutive sampling points is still not lower than 0.72, the release determination result is one, and ventilation is stopped. Otherwise, as long as any point is below the threshold, ventilation continues.

[0091] It should be noted that the release probability threshold and the number of consecutive satisfactions can be configured separately for each target gas. For example, the release probability threshold for hydrogen sulfide can be 0.72 and the number of consecutive satisfactions can be 3; for ammonia, it can be 0.7 and 3; for carbon monoxide, it can be 0.68 and 3; and for methane, it can be 0.75 and 4. The reason for setting a higher release probability threshold and a larger number of consecutive satisfactions for methane is that methane is associated with flammability risk, and a more conservative release requirement is preferred.

[0092] This step makes a clear judgment on whether to lift the ventilation ban based on the dynamic release threshold, release confidence probability, release probability threshold, and number of consecutive satisfactions. It compresses three types of information—whether the current concentration is low enough, whether the decrease is sufficient, and whether the adjacent locations are consistent—into a unified release confidence probability. Through the continuous satisfaction mechanism, the instantaneous judgment is transformed into a stable judgment, thereby ensuring the continuity and consistency of the ventilation ban.

[0093] In step S4, after ventilation is stopped, the current concentration level is tracked within the confirmation window to generate a label result. Based on the label result, the release confidence model coefficient, release ratio coefficient, and release probability threshold are updated online. Specific details include:

[0094] After making a probabilistic judgment on the release of ventilation, the relevant parameters of the release credibility model and dynamic release threshold are updated online based on the actual results after release, so that the control strategy can gradually conform to the real gas evolution characteristics under different pipe segment scenarios in multiple rounds of inspection.

[0095] Specifically, after step S3 makes a release determination and cancels the ventilation, the ventilation event does not end immediately. Instead, it continues to track the current concentration level of the main control gas within a confirmation window. The confirmation window can extend for a fixed duration from the release time. In this embodiment, 90 seconds is preferably used as the confirmation window length.

[0096] If the current concentration level of the main control gas does not reach the ventilation trigger threshold again within the confirmation window, the release is considered successful.

[0097] If the ventilation trigger threshold is reached again within the confirmation window, the release is considered to have failed. For ease of subsequent updates, this result will be recorded as a tagged result. The labeling results can be represented by the following formula: ;in, The labeling results for this release event. This is the confirmation window after release. To confirm the current concentration level within the window, The air exchange trigger threshold for the main control gas;

[0098] After obtaining the labeling results, the coefficients of the released reliable model are first updated online. The online update adopts a single-step gradient method, so that the model can self-correct based on the true results after each release. The model coefficient update can be expressed by the following formula: ;in, For the updated number Each model coefficient For the previous version Each model coefficient Target gas Online learning rate For the label results, The release probability at the release time. For the model input result corresponding to the release time, when When the values ​​are 1, 2, and 3, they correspond to the threshold margin result, the fallback magnitude result, and the verification deviation result, respectively. When the value is 0, It is a constant of 1;

[0099] When the model gives a high probability of release at the release time, but the actual release fails, When the value is negative, the model coefficients will be corrected in a more conservative direction; when the model gives a high probability of release and the release is actually successful, the model coefficients will be strengthened. For example, the online learning rate... It can be preferably set to 0.02 to ensure smooth updates;

[0100] After updating the model coefficients, this step continues with the release ratio coefficient in the dynamic release threshold. and the release probability threshold in step S3 Online correction is performed, and the release ratio coefficient is used to control the degree to which the dynamic release threshold falls back relative to the ventilation trigger threshold;

[0101] If the release fails, it means that the current release ratio is too high and needs to be reduced.

[0102] If the release is successful continuously, it indicates that the current release ratio may be too conservative and can be appropriately increased.

[0103] The release ratio coefficient can be updated in the following ways: when hour, ;when And reach the consecutive success rate threshold hour, When neither of the above two conditions is met, ;in, This is the updated release ratio coefficient. This is the release ratio coefficient before the update. and These represent the minimum and maximum boundaries of the release ratio coefficient, respectively. For single update step size, This is a threshold for the number of consecutive successful attempts.

[0104] The release probability threshold is used to control at what level of reliable release probability is allowed to terminate ventilation.

[0105] If the release fails, it means that the current release probability threshold is too low and needs to be increased.

[0106] If the release is successful consecutively, it indicates that the current release probability threshold may be too conservative and can be appropriately lowered.

[0107] The release probability threshold can be updated in the following ways: when hour, ;when And reach the consecutive success rate threshold hour, When neither of the above two conditions is met, ;in, This is the updated release probability threshold. The release probability threshold before the update. and These are the minimum and maximum boundaries of the release probability threshold, respectively. This is the step size for a single update;

[0108] For example, for hydrogen sulfide, the initial value of the release proportion coefficient can be set to 0.6, the minimum boundary can be set to 0.45, the maximum boundary can be set to 0.75, and the single update step size can be set to 0.02; the initial value of the release probability threshold can be set to 0.72, the minimum boundary can be set to 0.65, the maximum boundary can be set to 0.85, the single update step size can be set to 0.01, and the consecutive success number threshold can be set to 5. If, in a certain branch confluence entrance scenario, ventilation stops twice and then resumes within 30 seconds, the threshold for consecutive successes is set to 5. When the wind trigger threshold is triggered, the release ratio coefficient will gradually decrease from 0.6 to 0.58 and then to 0.56; the release probability threshold will gradually increase from 0.72 to 0.73 and then to 0.74. In this way, when entering the same scenario again, the dynamic release threshold will be lower and the release probability requirement will be higher, so a more conservative release strategy will be adopted. Conversely, if the release is successful for five consecutive times, the release ratio coefficient can be increased by 0.01 and the release probability threshold can be decreased by 0.005 to avoid being overly conservative.

[0109] It should be noted that the online update in step S4 can maintain parameter sets separately according to the pipe segment scenario. That is, the entry segment, the narrow straight pipe segment, and the branch merging entry segment can use the same calculation framework, but maintain their own model coefficients, release ratio coefficients, and release probability thresholds respectively. This can form differentiated release habits in different pipe segments. For example, due to the more obvious local backflow and confluence disturbances at the branch merging entry, its release ratio coefficient will usually automatically converge to a lower level after multiple rounds of updates, while its release probability threshold will converge to a higher level; while the parameters of the narrow straight pipe segment will gradually converge to relatively loose values ​​because the backflow process is more stable.

[0110] The actual results after one release are fed back into the judgment of the next release, forming a complete closed loop. The success or failure after release is explicitly converted into the basis for updating model coefficients and threshold parameters. After multiple rounds of inspection, it can gradually conform to the gas evolution characteristics of various pipeline scenarios, thereby continuously reducing the probability of misdiagnosis and repeated start-stop.

[0111] This invention discloses a closed-loop control system for toxic gas threshold of a drainage pipeline robot, comprising: a gas detection trigger module, a dynamic threshold fall-off module, a release reliable release module, and a parameter update closed-loop module, with signal connections between the modules;

[0112] Gas detection trigger module: Collects the original concentration value of the target gas, ambient temperature value, sampling time and current location according to a fixed sampling period, performs temperature compensation and smoothing on the original concentration value, calculates the normalized risk result, compares the normalized risk results of each target gas, and determines the main control gas, ventilation trigger threshold and ventilation trigger time.

[0113] Dynamic threshold fall-off module: It captures a reference window around the current ventilation event, calculates the background concentration result, generates a dynamic release threshold by combining the safety margin and release ratio coefficient, and calculates the fall-off magnitude result and the verification deviation result.

[0114] Release Confidentiality Module: When the current concentration level is not higher than the dynamic release threshold, calculate the threshold margin result, establish a release confidence model based on the threshold margin result, the fall amplitude result and the verification deviation result, output the release confidence probability, and make a release judgment by combining the release probability threshold and the number of consecutive satisfactions;

[0115] Parameter update closed-loop module: After ventilation is turned off, the current concentration level is tracked in the confirmation window, a label result is generated, and the release confidence model coefficient, release ratio coefficient and release probability threshold are updated online based on the label result.

[0116] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0117] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0118] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and inventive constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0119] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0120] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0121] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A closed-loop control method for toxic gas threshold in a drainage pipeline robot, characterized in that, Includes steps; Step S1: Collect the original concentration value of the target gas, the ambient temperature value, the sampling time and the current location according to a fixed sampling period. Perform temperature compensation and smoothing on the original concentration value, calculate the normalized risk result, compare the normalized risk results of each target gas, and determine the main control gas, the ventilation trigger threshold and the ventilation trigger time. Step S2: Capture a reference window around the current ventilation event, calculate the background concentration result, generate a dynamic release threshold by combining the safety margin and release ratio coefficient, and calculate the drop amplitude result and the verification deviation result; Step S3: When the current concentration level is not higher than the dynamic release threshold, calculate the threshold margin result, establish a release confidence model based on the threshold margin result, the fall amplitude result and the verification deviation result, output the release confidence probability, and make a release judgment by combining the release probability threshold and the number of consecutive satisfactions. Step S4: After the ventilation is turned off, track the current concentration level in the confirmation window, generate a label result, and update the release confidence model coefficient, release ratio coefficient and release probability threshold online based on the label result.

2. The closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 1, characterized in that, Hydrogen sulfide, methane, ammonia, and carbon monoxide are detected according to a fixed sampling cycle, and the corresponding sampling time and current location are recorded. The current location includes the entry section, narrow straight pipe section, and branch inlet section that the robot passes through as it moves along the preset inspection route within the limited space of the drainage pipe. The original concentration value is temperature-compensated based on the temperature compensation coefficient corresponding to the target gas and the reference temperature value to form a compensated concentration value. The compensation concentration value is smoothed using the median method to form the current concentration level.

3. The closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 2, characterized in that, The normalized risk result is calculated based on the ratio of the current concentration level of each target gas to its corresponding ventilation trigger threshold. At each sampling moment, the target gas with the highest normalized risk result is selected as the current master gas; If the current concentration level of the main control gas reaches or exceeds its ventilation trigger threshold, the ventilation state is entered, and this moment is recorded as the ventilation trigger moment. Before the end of this ventilation event, release assessment and parameter updates are performed around the current main control gas.

4. The closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 1, characterized in that, After the ventilation trigger moment is formed, a reference window containing a preset number of sampling points is extracted from before the ventilation trigger moment. The average value of the current concentration level in the reference window is calculated to form the background concentration result. A dynamic release threshold is generated based on the background concentration result and the ventilation trigger threshold. The dynamic release threshold is the larger of the background constraint formed by the background concentration result plus the safety margin and the global constraint formed by the ventilation trigger threshold multiplied by the release ratio coefficient.

5. The closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 4, characterized in that, The magnitude of the drop is the difference between the current concentration level at the moment the ventilation is triggered and the current concentration level at the current moment, divided by the ventilation trigger threshold. When the current concentration level drops below the dynamic release threshold for the first time, the robot moves forward a preset verification distance along the current inspection direction and collects the verification concentration results of the neighboring locations. The verification deviation result is the difference between the verification concentration result and the current concentration level at the current location divided by the ventilation trigger threshold. The preset verification distance is set according to the pipe section type.

6. The closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 1, characterized in that, When the current concentration level is not higher than the dynamic release threshold, the threshold margin result is calculated. The threshold margin result is the difference between the dynamic release threshold and the current concentration level divided by the dynamic release threshold. The release confidence model takes the threshold margin result, fallback magnitude result, and verification deviation result as inputs and outputs the release confidence probability in the form of logistic regression. The increase of the threshold margin result and fallback magnitude result increases the release confidence probability, while the increase of the verification deviation result decreases the release confidence probability.

7. The closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 6, characterized in that, The release decision to lift ventilation is made only when the current concentration level continuously reaches the preset number of consecutive satisfactions, the number of sampling points is not higher than the dynamic release threshold, and the corresponding release confidence probability is not lower than the release probability threshold. Before deployment, the release trust model is trained based on historical ventilation events. During training, whether the main control gas in the confirmation window has reached the ventilation trigger threshold again is used as the basis for labeling positive and negative samples.

8. The closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 1, characterized in that, After the ventilation is turned off, the current concentration level of the main control gas continues to be tracked within a preset confirmation window; If the current concentration level of the main control gas does not reach the ventilation trigger threshold again within the confirmation window, the label result will be recorded as a successful release. If the ventilation trigger threshold is reached again within the confirmation window, the label result will be recorded as release failure.

9. A closed-loop control method for toxic gas threshold of a drainage pipeline robot according to claim 8, characterized in that, Based on the labeling results, the model coefficients of the released reliable model are updated online using a single-step gradient method. When the release fails, the model coefficients are corrected in a more conservative direction, and when the release is successful, the model coefficients are strengthened. The release ratio coefficient is corrected online. When the release fails, the release ratio coefficient is reduced. When the number of consecutive successes reaches the consecutive success threshold, the release ratio coefficient is increased. The update of the release ratio coefficient is constrained by the minimum boundary and the maximum boundary. The release probability threshold is adjusted online. When the release fails, the release probability threshold is increased. When the number of consecutive successes reaches the consecutive successes threshold, the release probability threshold is decreased. The update of the release probability threshold is constrained by the minimum boundary and the maximum boundary. Online updates maintain parameter sets separately for each pipe segment scenario.

10. A closed-loop control system for a toxic gas threshold of a drainage pipeline robot, used to implement the closed-loop control method for a toxic gas threshold of a drainage pipeline robot as described in any one of claims 1-9, characterized in that... ； Gas detection trigger module: Collects the original concentration value of the target gas, ambient temperature value, sampling time and current location according to a fixed sampling period, performs temperature compensation and smoothing on the original concentration value, calculates the normalized risk result, compares the normalized risk results of each target gas, and determines the main control gas, ventilation trigger threshold and ventilation trigger time. Dynamic threshold fall-off module: It captures a reference window around the current ventilation event, calculates the background concentration result, generates a dynamic release threshold by combining the safety margin and release ratio coefficient, and calculates the fall-off magnitude result and the verification deviation result. Release Confidentiality Module: When the current concentration level is not higher than the dynamic release threshold, calculate the threshold margin result, establish a release confidence model based on the threshold margin result, the fall amplitude result and the verification deviation result, output the release confidence probability, and make a release judgment by combining the release probability threshold and the number of consecutive satisfactions; Parameter update closed-loop module: After ventilation is turned off, the current concentration level is tracked in the confirmation window, a label result is generated, and the release confidence model coefficient, release ratio coefficient and release probability threshold are updated online based on the label result.

Images