Intelligent disposal method and system for workshop gas leakage partition positioning

Abstract

The present application relates to the field of workshop industrial gas leakage monitoring and disposal, especially to a workshop gas leakage partition positioning intelligent disposal method and system, which is based on multi-source data fusion and multi-dimensional risk quantification to match differentiated basic monitoring decisions, and corrects the basic inspection frequency, can adjust the inspection frequency in real time according to the speed of risk deterioration, greatly improves the real-time self-adaptive ability of the inspection strategy, and also constructs a composite sensing data frame and a benchmark database for comparison, realizes integrated output of leakage identification, leakage positioning, risk labeling and disposal suggestions, facilitates differentiated disposal by operation and maintenance personnel, and adds a pre-inspection analysis mechanism to mark attention objects of deteriorating equipment, improve the inspection level, realize intelligent, closed-loop and precise disposal of the whole domain of workshop gas leakage risk, and greatly improve the safety control level of the workshop gas.

Description

Technical Field

[0001] This invention relates to the field of monitoring and handling industrial gas leaks in workshops, and particularly to a method and system for intelligent handling of gas leaks in workshops by zoning and location. Background Technology

[0002] Pressure vessels, gas pipelines, and other equipment in industrial workshops operate continuously for extended periods. They are highly susceptible to gas leaks due to a combination of factors, including operating stress, media corrosion, component aging, and fluctuations in operating conditions. Gas leaks are characterized by their concealment, suddenness, rapid spread, and significant hazards. If they cannot be identified, located, and addressed promptly and accurately, they can easily lead to safety accidents such as explosions, poisoning, and environmental pollution, seriously threatening the safety of workshop production and the safety of personnel and equipment.

[0003] Currently, the monitoring and control system for gas leaks in industrial workshops is not yet perfect. The core problem lies in the significant technical shortcomings in the two major aspects of inspection and control, and leak identification, location and disposal, which seriously restricts the real-time and accuracy of gas safety control in workshops.

[0004] On the one hand, workshops generally adopt a static fixed inspection frequency control mode. Once the inspection frequency is set, it remains unchanged for a long time, which cannot adapt to the dynamic changes in equipment risk characteristics and is very easy to miss early leakage risks.

[0005] On the other hand, the accuracy and systematicness of gas leak identification, location and disposal technologies are seriously insufficient. Traditional monitoring mostly relies on a single gas concentration sensor to complete the detection. The monitoring dimension is single and it is easily affected by workshop environmental noise, dust, temperature and humidity fluctuations and equipment operating conditions. The false alarm and missed alarm rates of sensors remain high.

[0006] At the same time, single detection methods cannot accurately distinguish between single gas leaks and multi-gas mixed leaks, making it difficult to accurately locate the leak source, determine the leak risk level, and achieve precise positioning, intelligent classification, and efficient closed-loop handling.

[0007] To address the core pain points of the aforementioned technologies, such as rigid inspection strategies, low leak identification accuracy, and delayed intervention, there is an urgent need to develop a comprehensive, dynamic, and high-precision intelligent handling method for workshop gas leaks, enabling integrated operations that achieve precise risk quantification, intelligent zoned management and control, dynamic inspection adaptation, accurate leak location, and early detection and handling. Summary of the Invention

[0008] The purpose of this invention is to provide a method and system for intelligent handling of gas leaks in workshops by zoning and locating them, in order to solve the aforementioned technical defects. This is achieved through multi-source data fusion, multi-dimensional risk quantification, dynamic inspection and adjustment, multi-modal accurate identification, and hierarchical pre-inspection intervention. This addresses the problems of rigidity in traditional technical inspection strategies, low accuracy in leak identification, and lack of pre-emptive prevention, thereby realizing intelligent, closed-loop, and precise handling of gas leak risks across the entire workshop and greatly improving the level of gas safety management in workshops.

[0009] The objective of this invention can be achieved through the following technical solution: a workshop gas leak zoned location intelligent handling method, comprising the following steps:

[0010] Step 1: Collect three types of data and four key fields from the monitoring facilities in the target workshop, perform calculations on the remaining basic data, and filter the target monitoring risk indicators for each monitoring facility based on the calculation results.

[0011] Step 2: Based on the target monitoring risk indicators of each monitoring facility, the fatigue cumulative damage coefficient C and the estimated historical leakage kernel density of each monitoring facility are processed and calculated. The calculated comprehensive risk score S is judged, and the workshop risk level map and preset monitoring decisions are output.

[0012] Step 3: Based on the workshop risk level map, calculate the rate of change and acceleration of the rate of change of the comprehensive risk score S of each monitoring facility, and output the dynamic inspection frequency f1 by combining the rounding method.

[0013] Step 4: Based on the dynamic inspection frequency f1 of each area, collect target gas concentration data, spectral fingerprint data and ultrasonic signals of specific frequency bands to locate and identify leaks, and output alarm and handling data packets;

[0014] Step 5: Based on the alarm handling data packet, perform a pre-detection analysis on the rate of change and acceleration of the rate of change of the orange area, and output a pre-detection command or a regular command.

[0015] Preferably, the three types of data include operating stress data, corrosion detection data, and maintenance record data; the four key fields include the three-dimensional spatial coordinates at the time of the leak, the time of the leak, the duration of the leak and the estimated amount of leakage, the equipment number and pipe section number to which the leak point belongs;

[0016] Preferably, the analysis process for the target monitoring risk indicators is as follows:

[0017] For each monitoring facility, a corresponding failure mode is matched, and the remaining basic data is calculated, including fatigue remaining life, corrosion remaining life, and aging remaining life; the failure modes include fatigue failure mode, corrosion thinning failure mode, and seal aging failure mode; the minimum value among the remaining basic data is selected as the target monitoring risk indicator of the monitoring facility.

[0018] Preferably, the analysis process of the workshop risk level map is as follows:

[0019] The cumulative fatigue damage degree and historical leakage intensity of each monitoring facility are obtained. The fatigue cumulative damage coefficient C is calculated based on the cumulative fatigue damage degree. The target monitoring risk index of the monitoring facility is transformed into the aging risk index L with a value range of [0, 1]. The historical leakage intensity is normalized by the Min-Max normalization method to obtain the estimated value of historical leakage kernel density.

[0020] The target workshop plane is discretized into a 1m×1m grid. For each monitoring facility, all grid cells corresponding to the spatial coverage of the monitoring facility are obtained. The maximum value of the historical leakage kernel density estimate after normalization of the grid cells is extracted, and the comprehensive risk score S of each monitoring facility is calculated.

[0021] The preset comprehensive risk score thresholds [Smin, Smax] are retrieved to perform discrimination processing on the comprehensive risk score S, and high-risk, medium-risk and low-risk levels are output. Based on the discrimination processing results, the facility risk level is projected and filled into the 1m×1m grid it covers to generate a workshop risk level map. The workshop risk level map includes red area: high-risk level, orange area: medium-risk level and green area: low-risk level. Grids without monitoring facilities are marked as blank areas.

[0022] Based on the preset monitoring decisions corresponding to each color area in the workshop risk level map.

[0023] Preferably, the analysis process for the dynamic inspection frequency f1 is as follows:

[0024] Calculate the rate of change and acceleration of the comprehensive risk score S for each monitoring facility. Based on the workshop risk level map, obtain the basic inspection frequency f0 for each area. Retrieve the preset speed sensitivity coefficient α and preset acceleration sensitivity coefficient β. Calculate the dynamic adjustment coefficient k1 based on 1 + α × rate of change + β × acceleration of change. Calculate the dynamic inspection frequency f1 based on the dynamic adjustment coefficient k1 × basic inspection frequency f0. Limit the dynamic inspection frequency f1 and round the limited dynamic inspection frequency f1 to the available level. Overlay and display the current dynamic inspection frequency f1 of each monitoring facility on the workshop risk level map.

[0025] Preferably, the target gas concentration data includes ammonia and CO2. Feature peaks are extracted from the collected spectral fingerprint data to construct a spectral fingerprint feature vector. The ultrasonic signal in a specific frequency band is denoised and time-frequency analysis is performed using short-time Fourier transform (STFT) to extract acoustic fingerprint features and construct an acoustic fingerprint vector.

[0026] The target gas concentration data, spectral fingerprint feature vector, and acoustic fingerprint vector of the same time and the same monitoring facility are fused into a composite sensing data frame.

[0027] Preferably, the analysis process of the alarm handling data packet is as follows:

[0028] The composite sensing data frame is compared and analyzed with a preset benchmark database. If an early warning command is triggered, the gas corresponding to the maximum value of the parameter concentration is set as an abnormal gas.

[0029] Calculate the similarity score between the spectral fingerprint feature vector and the standard spectral fingerprint feature vector, set the gas corresponding to the maximum similarity score as the identification gas, compare the abnormal gas with the identification gas, and output whether it is a single leak or a multi-gas leak.

[0030] When an early warning command is triggered, the cosine similarity between the acoustic fingerprint vector and the standard acoustic fingerprint vector is obtained and processed to output a high-risk label / low-risk label.

[0031] By combining high-risk / low-risk labels, warning instructions, and single / multiple gas leaks, an alarm response data package is obtained.

[0032] Preferably, the pre-inspection analysis process is as follows: perform pre-inspection analysis on the rate of change and acceleration of the rate of change of the orange area, and output a pre-inspection work order or a regular instruction. The pre-inspection work order includes the facility number, coordinates, and rate of change.

[0033] The monitoring facility is marked as a concern and indicated by a yellow dashed circle on the workshop risk level map.

[0034] The beneficial effects of this invention are as follows:

[0035] (1) This invention is based on three types of on-site measured information and extracted four-dimensional leakage key fields, combined with failure modes, to screen target monitoring risk indicators, realize objective quantitative screening of facility risks under multiple failure modes, solve the problem of difficulty in objectively locking high-risk objects of concern for a large number of workshop equipment, and introduce the concept of multi-indicator weighted fusion to generate a visualized workshop risk level map, divide high, medium and low risk areas into red, orange and green, and match differentiated basic monitoring decisions. At the same time, a dynamic adjustment coefficient is constructed to correct the basic inspection frequency, and the inspection frequency can be adjusted in real time according to the speed of risk deterioration, thereby improving the real-time adaptive capability of the inspection strategy.

[0036] (2) The present invention also constructs a composite sensing data frame and compares it with a benchmark database, which can distinguish between single gas leaks and multi-gas mixed leaks. It uses the cosine similarity of sound waves to mark the risk level, generates a complete alarm handling data package and links the sound and light flashing alarm. Multimodal information is mutually verified, reducing the probability of false alarms from a single sensor, and realizing the integrated output of leak identification, leak location, risk labeling and handling suggestions, which facilitates differentiated handling by operation and maintenance personnel.

[0037] (3) The present invention also adds a pre-inspection analysis mechanism for medium-risk orange areas, marks the objects of concern for equipment that are deteriorating rapidly, and upgrades the inspection level, so as to realize the intelligent handling of gas leaks in the workshop from risk assessment to dynamic inspection to leak identification to early inspection. Attached Figure Description

[0038] The invention will now be further described with reference to the accompanying drawings;

[0039] Figure 1 This is a schematic diagram of the method of the present invention;

[0040] Figure 2 This is a flowchart of the system of the present invention;

[0041] Figure 3 This is a schematic diagram illustrating the target monitoring risk indicator analysis of the present invention;

[0042] Figure 4 This is a schematic diagram of the preset monitoring decision matching analysis of the present invention. Detailed Implementation

[0043] 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.

[0044] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments;

[0045] Example 1: Please refer to Figures 1 to 4 As shown, the present invention is a workshop gas leak zone location intelligent handling method, which includes the following steps:

[0046] Step 1: Collect three types of data and four key fields from the monitoring facilities in the target workshop, perform calculations on the remaining basic data, and filter the target monitoring risk indicators for each monitoring facility based on the calculation results.

[0047] Step 2: Based on the target monitoring risk indicators of each monitoring facility, the fatigue cumulative damage coefficient C and the estimated historical leakage kernel density of each monitoring facility are processed and calculated. The calculated comprehensive risk score S is judged, and the workshop risk level map and preset monitoring decisions are output.

[0048] Step 3: Based on the workshop risk level map, calculate the rate of change and acceleration of the rate of change of the comprehensive risk score S of each monitoring facility, and output the dynamic inspection frequency f1 by combining the rounding method.

[0049] Step 4: Based on the dynamic inspection frequency f1 of each area, collect target gas concentration data, spectral fingerprint data and ultrasonic signals of specific frequency bands to locate and identify leaks, and output alarm and handling data packets;

[0050] Step 5: Based on the alarm handling data packet, perform a pre-detection analysis on the rate of change and acceleration of the rate of change of the orange area, and output a pre-detection command or a regular command.

[0051] Furthermore, step 1 establishes a unified and comparable "quantitative indicator" for each monitoring facility through multi-source data fusion and remaining lifespan quantification, thereby solving the initial screening problem of "how to objectively select the most concerning high-risk targets from a large number of facilities," specifically including:

[0052] Collect three types of data from monitoring facilities such as pressure vessels and pipeline sections in the target workshop, and retrieve all real leakage alarm records and maintenance work orders from the workshop over the past 3-5 years, extracting four key fields from each record;

[0053] The three types of data from the monitoring facilities include:

[0054] Operating stress data: cumulative operating time, operating pressure fluctuation amplitude, etc.;

[0055] Corrosion detection data: wall thickness reduction rate obtained by ultrasonic thickness gauge, corrosion rate measured by online corrosion probe, etc.

[0056] Maintenance record data: repair and replacement dates, number of welding repairs, fastener replacement records, etc.;

[0057] The four key fields include the three-dimensional spatial coordinates at the time of the leak, the time of the leak, the duration of the leak and the estimated amount of leakage, the equipment number and pipe section number to which the leak point belongs;

[0058] Based on the three types of data and four key fields of the monitoring facilities, the corresponding failure mode is matched for each monitoring facility, and the remaining basic data is calculated.

[0059] The failure modes include fatigue failure mode, corrosion thinning failure mode and seal aging failure mode.

[0060] The remaining basic data includes fatigue life remaining, corrosion life remaining, and aging life remaining;

[0061] Furthermore, the remaining fatigue life = (preset allowable fatigue cycle count - accumulated stress cycle count) / preset allowable fatigue cycle count × preset design life;

[0062] Remaining corrosion life = (Measured wall thickness - Preset minimum allowable wall thickness) / Historical average wall thickness reduction rate;

[0063] Remaining service life due to aging = Preset design service life - Time already in service;

[0064] The minimum value among the remaining basic data is selected as the target monitoring risk indicator for the monitoring facility.

[0065] Step 2, based on the "target monitoring risk indicators" selected in Step 1, introduces two dimensions: "historical leakage nuclear density estimate" and "cumulative damage level." Through weighted fusion and threshold grading, discrete facility risks are transformed into a visualized overall workshop risk level map. Based on this, differentiated basic monitoring strategies are matched, specifically including:

[0066] Obtain the cumulative fatigue damage degree of each monitoring facility. The cumulative fatigue damage degree is the sum of the ratios between the actual number of cycles under different stress amplitudes and the preset allowable number of fatigue cycles under the corresponding stress amplitude, i.e., ∑(ni / Ni), where ni represents the actual number of cycles under the i-th stress amplitude, Ni represents the preset allowable number of fatigue cycles under the i-th stress amplitude, and i is a natural number greater than zero.

[0067] The target workshop plane was discretized into a 1m×1m grid, and the historical leakage intensity was calculated for each grid point using two-dimensional Gaussian kernel density estimation. : ;

[0068] Where M is the total number of historical leakage events, (xg, yg) is the coordinate of the g-th leakage event, h is the bandwidth parameter, K is the standard Gaussian kernel function, and the normalized sample range is all confirmed historical leakage events in this workshop, excluding false alarms and invalid records.

[0069] The fatigue cumulative damage coefficient C is calculated based on the cumulative fatigue damage degree, where C = min(1, cumulative fatigue damage degree / preset cumulative fatigue damage degree threshold (e.g., 0.7)).

[0070] The target monitoring risk index of the monitoring facility is normalized to the [0-1] interval using Min-Max to obtain a dimensionless normalized value. The target monitoring risk index of the monitoring facility is then transformed into an aging risk index L, with a value range of [0,1] and a positive correlation with risk, where L = e （-a×目标监控风险指标） , where 'a' is the attenuation coefficient, which is a dimensionless coefficient and has a preset value range of 0.01-0.1 depending on the equipment type;

[0071] The historical leakage intensity was normalized using the Min-Max normalization method to obtain the historical leakage kernel density estimate;

[0072] For each monitoring facility, all grid cells corresponding to the spatial coverage area of ​​the monitoring facility are obtained, the maximum value of the historical leakage kernel density estimate after normalization of the grid cells is extracted, and the comprehensive risk score S of each monitoring facility is calculated. The comprehensive risk score S = aging risk index L × w1 + fatigue cumulative damage coefficient C × w2 + maximum historical leakage intensity × w3, where w1, w2 and w3 are preset weight coefficients, and w1, w2 and w3 are all greater than zero.

[0073] The preset comprehensive risk score threshold [Smin, Smax] is retrieved to judge the comprehensive risk score S. If the comprehensive risk score S > Smax, it is judged as high risk level; if the comprehensive risk score S ∈ [Smin, Smax], it is judged as medium risk level; if the comprehensive risk score S < Smin, it is judged as low risk level.

[0074] Based on the discrimination and classification results, the facility risk level is projected and filled into the 1m×1m grid it covers to generate a workshop risk level map. The workshop risk level map includes red area: high risk level, orange area: medium risk level and green area: low risk level. Grids without monitoring facilities are marked as blank areas and do not participate in risk control.

[0075] Based on the discriminant processing results, a workshop risk level map is obtained, which includes a red area (high risk), an orange area (medium risk), and a green area (low risk).

[0076] Based on the preset monitoring decisions corresponding to each color area in the workshop risk level map, for example, the red area has a basic inspection frequency of 10Hz, automatically triggers drone / robot inspections daily, and must arrange a manual ultrasonic flaw detection review once a week; the orange area has a basic inspection frequency of 5Hz, automatically triggers robot inspections every 3 days, and conducts a manual review once a month; and the green area has a basic inspection frequency of 1Hz, automatically triggers robot inspections weekly, and conducts a manual review once a quarter.

[0077] Method for determining the spatial coverage of monitoring facilities: For pressure vessels, the spatial coverage is defined as 1.5m outward from the outer contour of the equipment; for gas pipelines, the spatial coverage is defined as 1.2m outward from the centerline of the pipeline on both sides, mapped onto a 1m×1m grid on the workshop plane.

[0078] Multi-facility overlapping grid conflict trade-off rule: If the same grid is covered by multiple monitoring facilities, the grid risk level is taken as the highest risk level among the multiple facilities;

[0079] Blank area boundary: Grids that are not within the coverage area of ​​any monitoring facilities are defined as blank areas. Blank grids are not included in inspections or risk scoring calculations.

[0080] Risk level map and dynamic inspection frequency update time scale: The risk level map is fully recalculated quarterly; the comprehensive risk score S and dynamic inspection frequency f1 are updated in real time according to the sampling step size Δt; when a manually confirmed leak event occurs, the estimated value of the grid kernel density around the leak point is immediately updated locally and the local risk level map is refreshed without waiting for the quarterly cycle.

[0081] Example 2: Step 3. Based on the risk level and basic inspection frequency (f0) generated in Step 2, introduce the dimension of "risk change trend" (rate of change and acceleration) to achieve real-time dynamic adjustment of the inspection frequency. This allows monitoring resources to be tilted towards areas where "risk is accelerating," solving the problem that "static grading cannot cope with sudden risk changes." Specifically, this includes:

[0082] Obtain the comprehensive risk score S(t) of each monitoring facility at the current time t (t>0), and at the same time obtain the historical comprehensive risk score sequence {S(t-Δt), S(t-2Δt), ..., S(t-nΔt)} of each monitoring facility, where n is a natural number greater than zero and Δt is the sampling time step;

[0083] Calculate the rate of change and acceleration of the rate of change of the comprehensive risk score S for each monitoring facility;

[0084] Rate of change = [S(t) - S(t-Δt)] / Δt;

[0085] Rate of change acceleration = (Current rate of change - Previous rate of change) / Δt;

[0086] Based on the workshop risk level map, the basic inspection frequency f0 of each area is obtained;

[0087] Retrieve the preset velocity sensitivity coefficient α and preset acceleration sensitivity coefficient β. α defaults to 0.5-2.0, and β defaults to 0.1-0.5.

[0088] The rate of change and acceleration of the rate of change of the comprehensive risk score S are respectively subjected to maximum-min normalization to the dimensionless interval of [0-1].

[0089] The dynamic adjustment coefficient k1 is calculated based on the acceleration of 1 + α × rate of change + β × rate of change. The amplitude of k1 is limited, k1∈[k1min, k1max], where k1min and k1max are the preset minimum and maximum values ​​of the dynamic adjustment coefficient, such as k1∈[0.5, 3.0].

[0090] Limit: If k1 > k1max, then k1 takes the value k1max; if k1 < k1min, then k1 takes the value k1min; if k1 ∈ [k1min, k1max], then k1 takes the value k1.

[0091] The dynamic inspection frequency f1 is calculated based on the dynamic adjustment coefficient k1 × the basic inspection frequency f0. The dynamic inspection frequency f1 is then limited, f1∈[f1min, f1max], where f1min and f1max are the preset minimum and maximum values ​​of the dynamic inspection frequency, such as f1∈[0.5, 20].

[0092] Limit: If f1 > f1max, then f1 takes the value f1max; if f1 < f1min, then f1 takes the value f1min; if f1 ∈ [f1min, f1max], then f1 takes the value f1.

[0093] Round the limited dynamic inspection frequency f1 to the available range (e.g., 0.5, 1, 2, 5, 10, 20Hz).

[0094] The current dynamic inspection frequency f1 of each monitoring facility is overlaid on the workshop risk level map.

[0095] Step 4: Based on the dynamic inspection frequency f1 of each area, collect target gas concentration data, spectral fingerprint data, and ultrasonic signals of specific frequency bands to locate and identify leaks, and output an alarm handling data package, specifically including:

[0096] Based on the workshop risk level map, target gas concentration data for each area other than the non-blank area are obtained. The target gas concentration data includes ammonia, CO2, etc. At the same time, spectral fingerprint data of each area is obtained by scanning the characteristic absorption spectrum of the target gas in the near-infrared / mid-infrared band. In addition, ultrasonic signals of specific frequency bands generated by target gas leakage are captured by an ultrasonic microphone array (response frequency range 20kHz-100kHz).

[0097] Feature peaks are extracted from the collected spectral fingerprint data. The spectral fingerprint data can be processed using a polynomial fitting algorithm from traditional techniques to remove baseline drift caused by laser intensity fluctuations, thereby improving the accuracy of baseline representation. Feature peaks include the center wavelength of the absorption peak, the peak amplitude of the absorption peak, and the full width at half maximum (FWHM), and a spectral fingerprint feature vector is constructed.

[0098] The collected ultrasonic signals in a specific frequency band are denoised by using a digital bandpass filter (passband frequency 25kHz-80kHz) to filter out low-frequency noise in the workshop environment. The denoised ultrasonic signals are then subjected to time-frequency analysis using short-time Fourier transform (STFT) to extract acoustic features and construct an acoustic fingerprint vector.

[0099] Voiceprint features include:

[0100] Peak frequency (Hz): The frequency at which sound wave energy is concentrated;

[0101] Spectral centroid (Hz): The average frequency position of the centroid of the spectrum;

[0102] Spectral entropy (dimensionless): The degree of disorder in the distribution of spectral energy, used to distinguish between "continuous jet sound" and "sudden pulse sound";

[0103] Zero-crossing rate in the time domain (times / second): characterizes the oscillation rate of the sound wave waveform;

[0104] The target gas concentration data, spectral fingerprint feature vector, and acoustic fingerprint vector of the same time and the same monitoring facility are fused into a composite sensing data frame.

[0105] The composite sensing data frames are compared and analyzed with a preset benchmark database.

[0106] If the concentration of a parameter in the target gas concentration data is greater than the corresponding preset concentration alarm threshold, the concentration is judged to be out of control, an early warning command is immediately triggered, and the gas corresponding to the maximum value of the parameter concentration is set as an abnormal gas.

[0107] The similarity score between the spectral fingerprint feature vector and the standard spectral fingerprint feature vector is calculated in parallel. The gas corresponding to the maximum similarity score is set as the identification gas. If the identification gas is the same as the abnormal gas, a single leak is determined. If the identification gas is different from the abnormal gas, multiple gas leaks are determined.

[0108] When an early warning command is triggered, the cosine similarity between the acoustic fingerprint vector and the standard acoustic fingerprint vector is obtained. If the cosine similarity is greater than the preset maximum acoustic similarity value, a high-risk label is assigned. If the cosine similarity is less than or equal to the preset maximum acoustic similarity value, a low-risk label is assigned. The high-risk label / low-risk label, the early warning command, and the single leak / multiple gas leak are combined to obtain an alarm handling data package.

[0109] The alarm handling data package includes a real-time workshop risk level map, high-risk / low-risk labels, early warning instructions, and single / multiple gas leak combinations, gas names, etc.

[0110] Output alarm handling data packets and simultaneously display the text "Abnormal" flashing in the corresponding area on the workshop risk level map, and issue an audible and visual alarm. This allows operation and management personnel to conduct targeted operation and maintenance management of the leak area based on high-risk / low-risk labels and single / multiple gas leaks. For example, the operation and maintenance order for alarm data packets with low-risk labels is lower than that for alarm data packets with high-risk labels, so as to make reasonable operation and maintenance arrangements. For single leaks / multiple gas leaks, single or combined treatments are carried out, such as single ventilation, or combined ventilation and sprinkler systems.

[0111] The preset benchmark database includes standard spectral fingerprint feature vectors, standard acoustic fingerprint vectors, preset concentration alarm thresholds, etc.

[0112] Step 5: Based on the alarm handling data packet, perform a pre-detection analysis on the rate of change and acceleration of the rate of change of the orange area, and output a pre-detection command or a regular command.

[0113] Preliminary analysis is performed on the rate of change and acceleration of the rate of change within the orange region:

[0114] If the rate of change is greater than the preset rate of change, or the rate of change acceleration is greater than the preset rate of change acceleration, a pre-inspection work order is generated, including facility number, coordinates, rate of change, rate of change acceleration, etc.

[0115] If the rate of change is less than or equal to the preset rate of change, and the acceleration of the rate of change is less than or equal to the preset rate of change acceleration, then a regular instruction is generated to continuously monitor the alarm handling data packet or the workshop risk level map.

[0116] Mark the monitoring facility as a "focused object" and mark it with a yellow dotted circle on the workshop risk level map. At the same time, trigger the pre-inspection command and immediately execute the preset early warning operation of the pre-inspection command, such as assigning robots or maintenance personnel to conduct ultrasonic flaw detection or wall thickness sampling in advance, and at the same time increase the dynamic inspection frequency f1 of the area by one level.

[0117] It should be noted that the same monitoring facility will not trigger pre-inspection analysis again within 72 hours after the pre-inspection is triggered.

[0118] Example 3: The present invention also proposes a workshop gas leak zoned location intelligent handling system, including:

[0119] The monitoring data processing module is used to collect three types of data and four key fields from each monitoring facility, calculate and process the remaining basic data, output the remaining basic data, and select the minimum value among the remaining basic data as the target monitoring risk indicator of the monitoring facility.

[0120] The workshop risk mapping module analyzes the output target monitoring risk indicators, combines the fatigue cumulative damage coefficient C of each monitoring facility with the estimated historical leakage kernel density, and outputs a comprehensive risk score S. The comprehensive risk score S is further divided and mapped to output a workshop risk level map. Pre-set monitoring decisions are obtained based on the workshop risk level map.

[0121] The dynamic monitoring and management module relies on the workshop risk level map to calculate the rate of change and acceleration of the rate of change of the comprehensive risk score S of each monitoring facility, and uses a weighted fusion method and rounding to calculate the dynamic inspection frequency f1.

[0122] The alarm location and handling module continuously collects target gas concentration data, spectral fingerprint data and specific frequency ultrasonic signals in each area based on the dynamic inspection frequency f1 of each area, and performs leak location identification, outputs alarm handling data packets, and simultaneously displays the text "abnormal" on the corresponding area in the workshop risk level map, and issues audible and visual alarms.

[0123] The pre-inspection management module performs pre-inspection analysis on the rate of change and acceleration of the rate of change of the orange area, determines whether to generate a pre-inspection work order, and outputs a pre-inspection work order or a regular instruction.

[0124] The visual feedback module is used to display information such as workshop risk level map, alarm handling data package, and pre-inspection work order;

[0125] In summary, based on three types of on-site measured information and extracted four-dimensional leakage key fields, combined with failure modes, target monitoring risk indicators are screened to achieve objective quantitative screening of facility risks under multiple failure modes. This solves the problem of objectively identifying high-risk objects of concern among a large number of workshop equipment. Furthermore, the concept of multi-indicator weighted fusion is introduced to generate a visualized workshop risk level map, which divides high, medium, and low risk areas into red, orange, and green, and matches differentiated basic monitoring decisions.

[0126] Simultaneously, a dynamic adjustment coefficient is constructed to correct the basic inspection frequency, which can adjust the inspection frequency in real time according to the speed of risk deterioration, improving the real-time adaptive capability of the inspection strategy. Furthermore, a composite sensing data frame is constructed and compared with a benchmark database to distinguish between single gas leaks and multi-gas mixed leaks. Based on the cosine similarity of sound waves, the risk level is marked, a complete alarm handling data package is generated, and an audible and visual flashing alarm is triggered. Multimodal information is cross-verified to reduce the probability of false alarms from a single sensor. This achieves integrated output of leak identification, leak location, risk labeling, and handling suggestions, facilitating differentiated handling by maintenance personnel. At the same time, a pre-inspection analysis mechanism is added for medium-risk orange areas to mark equipment that is deteriorating rapidly as objects of concern and increase the inspection level, realizing intelligent handling of gas leaks in the workshop from risk assessment to dynamic inspection, leak identification, and pre-inspection.

[0127] The above-mentioned comprehensive risk score weights w1, w2, and w3 are: w1∈[0.2-0.6], w2∈[0.2-0.5], and w3∈[0.1-0.4]; satisfying w1+w2+w3=1. The determination method is as follows: the analytic hierarchy process (AHP) is used, and weights are assigned based on historical leakage accident samples from the workshop. For pressure vessels, w2 (fatigue weight) is emphasized, and for corroded pipelines, w3 (historical leakage weight) is emphasized.

[0128] Comprehensive risk scoring thresholds [Smin, Smax]: S is normalized to the 0-1 range; Smin=0.4, Smax=0.7; can fluctuate by ±0.1 according to the workshop safety level;

[0129] Velocity sensitivity coefficient α∈[0.5-2.0], acceleration sensitivity coefficient β∈[0.1-0.5]; it is preferentially obtained by regression fitting of historical failure samples in the workshop; when there are no historical samples, the intermediate values ​​of the interval α=1.0, β=0.3 are used directly;

[0130] The dynamic adjustment coefficient limit k1∈[0.5, 3.0]; the dynamic inspection frequency limit f1∈[0.5Hz, 20Hz];

[0131] Acoustic fingerprint cosine similarity threshold: 0.75; calibrated based on laboratory standard leak samples; values ​​greater than the threshold are considered high-risk jet leaks.

[0132] Pre-detection trigger thresholds: Rate of change threshold is set to 0.08 / h; Rate of change acceleration threshold is set to 0.04 / h. 2 The risk level can be adjusted up or down by 20% depending on the workshop's risk control level.

[0133] Preset benchmark database construction method: Standard spectral fingerprint vector: obtained by collecting spectra of standard gases such as ammonia and CO2 at different concentrations in the laboratory; Standard acoustic fingerprint vector: obtained by collecting ultrasonic signals and extracting acoustic features from a test bench with different orifice diameters of leakage nozzles to establish a sample library; The database allows adding and deleting samples according to the type of media in the workshop;

[0134] That is, each preset parameter is obtained through the analytic hierarchy process, fault sample regression fitting, or laboratory calibration;

[0135] The threshold is set for result comparison and analysis to determine whether it is good or bad. The value of the threshold is determined by a combination of large-scale model analysis of the sample data and human experience, and can also be adjusted appropriately based on seasonal or common-sense influencing factors. The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A workshop gas leak zoned location intelligent handling method, characterized in that, Includes the following steps: Step 1: Collect three types of data and four key fields from the monitoring facilities in the target workshop, perform calculations on the remaining basic data, and filter the target monitoring risk indicators for each monitoring facility based on the calculation results. Step 2: Based on the target monitoring risk indicators of each monitoring facility, the fatigue cumulative damage coefficient C and the estimated historical leakage kernel density of each monitoring facility are processed and calculated. The calculated comprehensive risk score S is judged, and the workshop risk level map and preset monitoring decisions are output. Step 3: Based on the workshop risk level map, calculate the rate of change and acceleration of the rate of change of the comprehensive risk score S of each monitoring facility, and output the dynamic inspection frequency f1 by combining the rounding method. Step 4: Based on the dynamic inspection frequency f1 of each area, collect target gas concentration data, spectral fingerprint data and ultrasonic signals of specific frequency bands to locate and identify leaks, and output alarm and handling data packets; Step 5: Based on the alarm handling data packet, perform a pre-detection analysis on the rate of change and acceleration of the rate of change of the orange area, and output a pre-detection command or a regular command.

2. The workshop gas leak zone location intelligent handling method according to claim 1, characterized in that, The three types of data include operational stress data, corrosion detection data, and maintenance record data; the four key fields include the three-dimensional spatial coordinates at the time of the leak, the time of the leak, the duration of the leak and the estimated amount of leakage, the equipment number and pipe section number to which the leak point belongs.

3. The workshop gas leak zoning and intelligent handling method according to claim 2, characterized in that, The analysis process for the target monitoring risk indicators is as follows: For each monitoring facility, a corresponding failure mode is matched, and the remaining basic data is calculated, including fatigue remaining life, corrosion remaining life, and aging remaining life; the failure modes include fatigue failure mode, corrosion thinning failure mode, and seal aging failure mode. The minimum value among the remaining basic data is selected as the target monitoring risk indicator for the monitoring facility.

4. The workshop gas leak zone location intelligent handling method according to claim 1, characterized in that, The analysis process of the workshop risk level map is as follows: The cumulative fatigue damage degree and historical leakage intensity of each monitoring facility are obtained. The fatigue cumulative damage coefficient C is calculated based on the cumulative fatigue damage degree. The target monitoring risk index of the monitoring facility is transformed into the aging risk index L with a value range of [0, 1]. The historical leakage intensity is normalized by the Min-Max normalization method to obtain the estimated value of historical leakage kernel density. The target workshop plane is discretized into a 1m×1m grid. For each monitoring facility, all grid cells corresponding to the spatial coverage of the monitoring facility are obtained. The maximum value of the historical leakage kernel density estimate after normalization of the grid cells is extracted, and the comprehensive risk score S of each monitoring facility is calculated. The preset comprehensive risk score thresholds [Smin, Smax] are retrieved to perform discrimination processing on the comprehensive risk score S, and high-risk, medium-risk and low-risk levels are output. Based on the discrimination processing results, the facility risk level is projected and filled into the 1m×1m grid it covers to generate a workshop risk level map. The workshop risk level map includes red area: high-risk level, orange area: medium-risk level and green area: low-risk level. Grids without monitoring facilities are marked as blank areas. Based on the preset monitoring decisions corresponding to each color area in the workshop risk level map.

5. The workshop gas leak zoning and intelligent handling method according to claim 4, characterized in that, The analysis process of the dynamic inspection frequency f1 is as follows: Calculate the rate of change and acceleration of the rate of change of the comprehensive risk score S for each monitoring facility. Based on the workshop risk level map, obtain the basic inspection frequency f0 for each area. Retrieve the preset speed sensitivity coefficient α and preset acceleration sensitivity coefficient β. Perform maximum-minimum normalization on the rate of change and acceleration of the comprehensive risk score S to the dimensionless interval [0-1]. The dynamic adjustment coefficient k1 is calculated based on 1 + α × rate of change + β × rate of change acceleration. The dynamic inspection frequency f1 is calculated based on the dynamic adjustment coefficient k1 × the basic inspection frequency f0. The dynamic inspection frequency f1 is limited and rounded to the available level. The current dynamic inspection frequency f1 of each monitoring facility is overlaid and displayed on the workshop risk level map.

6. The workshop gas leak zone location intelligent handling method according to claim 1, characterized in that, The target gas concentration data includes ammonia and CO2. Feature peaks are extracted from the collected spectral fingerprint data to construct a spectral fingerprint feature vector. Noise is denoised from the ultrasonic signal in a specific frequency band and time-frequency analysis is performed using short-time Fourier transform to extract acoustic features and construct an acoustic fingerprint vector. The target gas concentration data, spectral fingerprint feature vector, and acoustic fingerprint vector of the same time and the same monitoring facility are fused into a composite sensing data frame.

7. The workshop gas leak zone location intelligent handling method according to claim 6, characterized in that, The analysis process of the alarm handling data packet is as follows: The composite sensing data frame is compared and analyzed with a preset benchmark database. If an early warning command is triggered, the gas corresponding to the maximum value of the parameter concentration is set as an abnormal gas. Calculate the similarity score between the spectral fingerprint feature vector and the standard spectral fingerprint feature vector, set the gas corresponding to the maximum similarity score as the identification gas, compare the abnormal gas with the identification gas, and output whether it is a single leak or a multi-gas leak. When an early warning command is triggered, the cosine similarity between the acoustic fingerprint vector and the standard acoustic fingerprint vector is obtained and processed to output a high-risk label / low-risk label. By combining high-risk / low-risk labels, warning instructions, and single / multiple gas leaks, an alarm response data package is obtained.

8. The workshop gas leak zone location intelligent handling method according to claim 1, characterized in that, The pre-detection analysis process is as follows: Pre-inspection analysis is performed on the rate of change and acceleration of the rate of change of the orange area, and a pre-inspection work order or routine instruction is output. The pre-inspection work order includes the facility number, coordinates, and rate of change. The monitored facility is marked as an object of concern and marked with a yellow dashed circle on the workshop risk level map.

9. A workshop gas leak zoning and intelligent handling system, wherein the system is applied to the workshop gas leak zoning and intelligent handling method as described in any one of claims 1-8, characterized in that, It includes a monitoring data processing module, a workshop risk mapping module, a dynamic monitoring management module, an alarm location and handling module, a pre-inspection management module, and a visual feedback module.

Images