1168 results about "Leak detection" patented technology

A system for AI-supported leak detection and localization in water distribution infrastructures, consisting of: a large number of distributed sensor nodes mounted along a water pipe, each sensor node comprising the following: a pressure sensor configured to measure local hydraulic pressure fluctuations within the pipeline with a resolution of at least 0.01 bar; a flow sensor configured to measure the volume flow within the pipeline with an accuracy of at least ±0.5% of the measured value; an acoustic sensor configured to detect vibration signatures caused by leaks in a frequency range between 50 Hz and 20 kHz; an embedded microcontroller with integrated analog-to-digital conversion circuitry for digitizing sensor outputs; a wireless communication module configured to transmit time-synchronized sensor data to a cloud-based processing platform; and a local energy subsystem with a rechargeable battery and an optional circuit for generating photovoltaic energy; the cloud-based processing platform includes an artificial intelligence engine that comprises the following: a data acquisition module configured to receive and decode the transmitted sensor data and to perform time alignment; a supervised learning module that is trained on historical data of flagged leaks and non-leaks to classify incoming sensor patterns; an unsupervised learning module configured to detect anomalies by modeling normal operating baselines of the pipeline; and a topology-aware localization module configured to determine leak coordinates using the topology of the pipeline network, modeling the propagation of hydraulic waves, and estimating the arrival time difference from multi-node acoustic detections; and wherein the system is configured to provide real-time leak alerts and georeferenced visualization via a remote monitoring interface.

The invention belongs to the technical field of leakage detection, and particularly relates to a leakage detection method based on multi-sensor time shift correction and deep learning. A plurality of sensors sensitive to leakage sound waves or vibration are arranged at different positions of a monitored pipeline or container and used for collecting leakage related signals; time alignment is carried out on the time sequence signals collected by the multiple sensors according to the position difference of the time sequence signals in the space; comparing and differentiating the plurality of sensor signals subjected to time alignment; modeling and updating environmental noise in real time, dynamically adjusting detection parameters according to noise level, continuously monitoring statistical characteristics of background signals, and establishing a noise model; a pre-trained deep learning model is used to carry out mode identification and classification determination on the signal after noise reduction processing; and the alarm and display module is used for triggering alarm and recording leakage information when suspected leakage is detected. According to the invention, weak leakage signals can be reliably detected under the condition of strong background noise.

The invention discloses a pipeline leakage detection device and method based on multi-modal data fusion, and relates to the technical field of pipeline safety monitoring. According to the device, modular sensor cabins (ultrasonic waves, LiDAR, millimeter wave radar and optics) are adopted for cooperative work, and high-precision leakage detection and positioning are achieved in combination with space-time registration, a cross-modal feature alignment network and a multi-level decision fusion mechanism. Through a dynamic weight distribution model and a multi-physical field joint inversion algorithm, the problems of low detection precision and poor anti-interference capability of a traditional single sensor are effectively solved. The method is suitable for detecting defects such as corrosion and leakage of metal and nonmetal pipelines, and has the advantages of high sensitivity, strong robustness and wide applicability.

The invention discloses an indoor gas leakage monitoring and intelligent joint control system, and belongs to the technical field of intelligent safety monitoring systems. Comprising a sensing layer, a network layer, a control layer and a platform layer, the sensing layer comprises a multi-sensor array and an edge computing node, the multi-sensor array is combined with a catalytic combustion sensor, an infrared sensor, a semiconductor sensor and a temperature and humidity sensor, and sensor data are fused through a weighting algorithm so as to reduce the false alarm rate; a temperature and humidity sensor dynamically adjusts a detection threshold value to adapt to different climate conditions, and an edge computing node is deployed with a lightweight machine learning model. Through multi-sensor data fusion and a weighting algorithm, the false alarm rate is reduced to be smaller than 0.1% per year, environment self-adaptive calibration and AI auxiliary decision making are performed, interference of alcohol, perfume and the like is effectively eliminated, different climate conditions are adapted, leakage judgment is completed at an equipment end through edge calculation, cloud delay is avoided, a multi-stage emergency strategy is provided, and intelligent safe ecology driven by the Internet of Things is constructed.

The invention belongs to the technical field of fused salt heat storage tank leakage detection, and discloses a device for fused salt heat storage tank leakage detection and a working method thereof. According to the device for detecting the leakage of the fused salt heat storage tank, a rotary supporting assembly is rotatably mounted on a base, and an equipment mounting frame is further mounted on the base; a plurality of spray heads and a plurality of cameras are arranged on the equipment mounting rack; one end of the detection cylinder is provided with a detection pipe and an air inlet pipe, and the other end is provided with an air outlet; one end provided with the air outlet is provided with a distance sensor, and the other end provided with the air outlet is further provided with a piston plate through an elastic assembly; the control unit is used for controlling rotation of the rotary supporting assembly, controlling opening and closing of the spray head, obtaining data collected by the distance sensor and obtaining data collected by the camera. According to the technical scheme, bubble visual monitoring and pressure change monitoring are combined, and the sensitivity and accuracy of leakage detection are improved.

A system of evaluating a mobile device includes a sensing device that captures data while a force is being applied to a display of the mobile device while the mobile device is held stationary; and a processing device to process the data and determine the force applied and a displacement of the display while the force is being applied.

The invention provides a steel shell concrete interface void detection system and method, and relates to the technical field of civil engineering structure health detection, and the system comprises a mobile scanning unit, an excitation unit, an infrared imaging unit, a signal processing and acquisition unit and a comprehensive analysis terminal. The method comprises the following steps: controlling a mobile scanning unit to carry out step-by-step impact excitation and signal acquisition so as to obtain a frequency diagram, and synchronously sampling by using an infrared thermal imager so as to obtain a thermal analysis diagram; and then binarizing the two images respectively, fusing the two images by adopting a noise suppression algorithm to generate a comprehensive image, and judging that the interface is void according to the fused image. According to the method, the high sensitivity of the impact echo method to the layering defect and the accurate description capability of the infrared thermal imaging method to the defect shape are fused, so that nondestructive, efficient and high-precision detection of the steel shell concrete interface void is realized, and the misjudgment and leak detection risks of a single method are effectively reduced.

The invention discloses an AMC sensor array spatial feature recognition and leakage point detection method based on deep learning, and relates to the technical field of sensors, and the method comprises the steps: carrying out the spatial correlation analysis of a preprocessing result, and extracting a spatial feature signal reflecting the spatial distribution characteristics of gas; decomposing the denoised spatial features by using an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode components, and generating a reconstruction signal based on each intrinsic mode component; performing framing processing on the reconstructed signal by using a framing windowing technology, calculating an energy value and a zero-crossing rate of a corresponding frame based on a framing processing result, constructing a leakage sensing model in combination with the extracted spatial feature signal, and generating a leakage probability distribution sequence; and inputting the time sequence window sequence into a self-encoder, and performing gas leakage point detection in combination with a knowledge distillation mechanism and a comparative analysis technology. According to the invention, accurate extraction and de-noising processing of the spatial feature signals are realized, the signal quality is improved, and the detection precision is further improved.

The invention provides a machining method for improving the yield of a 2K negative pressure aluminum plate-fin heat exchanger. The machining method comprises the steps that a core body is manufactured; manufacturing an end socket; the core body is connected with the end socket to manufacture the heat exchanger; normal-temperature detection is conducted on the heat exchanger; performing low-temperature detection on the heat exchanger; the heat exchanger is subjected to cold shock treatment, and the background leakage rate and the vacuum degree of each heat exchange channel in the heat exchanger and the background leakage rate and the vacuum degree of the sealing device in the cooling process are obtained; and under the condition that the temperature is reduced to the preset temperature threshold value, low-temperature helium mass spectrum leakage detection is conducted on the heat exchanger, after low-temperature helium mass spectrum leakage detection is conducted, secondary vacuumizing is conducted, then third-time vacuumizing is conducted, the circulation is conducted till ith vacuumizing is conducted, and the total leakage rate is calculated, and i is an integer larger than 3. The heat exchanger with the micro leakage rate, high mechanical strength, compact size and high cleanliness can be obtained, and the machining efficiency and the yield are improved.

The invention relates to the technical field of oil pipe leakage detection, in particular to an underground oil pipe leakage detection system, which comprises an underground permanent monitoring unit, which comprises at least one optical fiber sensor, and the optical fiber sensor clings to the outer wall of a production oil pipe or is integrated in an oil pipe coupling and extends along the depth of a whole shaft; the ground signal processing and analyzing unit is in communication connection with the underground permanent monitoring unit; wherein the optical fiber sensor is used as a distributed temperature sensor and a distributed sound wave vibration sensor at the same time, real-time on-line monitoring of oil pipe leakage is achieved, early warning can be given out at the initial stage of leakage, the accuracy of leakage recognition is improved through collaborative analysis of bimodal sensing data, the false alarm situation is effectively reduced, and the reliability of oil pipe leakage detection is improved. A distributed measurement mode provides meter-scale spatial resolution, the depth position of a leakage point is accurately positioned, normal production operation does not need to be interrupted during system operation, and the operation cost of an oil field is reduced.

A coolant leak detection and mitigation system for a rechargeable energy storage system having multiple battery cells arranged in individual battery modules includes a cooling system having a main coolant loop and multiple parallel coolant branches. Each coolant branch adjusts the temperature of one battery module using a portion of coolant from the main coolant loop. The cooling system also has flow-valve(s) for regulating and distributing the coolant from the main coolant loop across the coolant branches. An electronic controller is configured to monitor the cooling system for coolant loss and, in response to an indication of coolant loss, assess each coolant branch for a coolant leak. The controller is also configured to identify a coolant branch having a coolant leak and shut off, via the flow-valve(s), coolant flow into the coolant branch having the coolant leak.

The invention discloses a frequency domain ground penetrating radar pipe leakage detection method and system, and relates to the technical field of pipeline leakage detection, and the method comprises the steps: employing a frequency domain ground penetrating radar to carry out the whole pipe segment detection of a pipeline, and obtaining an original electromagnetic wave signal; preprocessing the original electromagnetic wave signal to obtain an echo signal; extracting time-frequency domain features of the echo signals, and screening the time-frequency domain features; identifying the screened time-frequency domain features by using a leakage identification model to obtain suspected leakage pipe sections; starting a microwave impedance measuring device at the position of the suspected leakage pipe section to obtain a reflection signal; and determining the axial position and the radial position of the leakage point according to the reflection signal. According to the method, the frequency domain ground penetrating radar is adopted for primary screening of the whole pipe section, after the suspected leakage pipe section is recognized, the microwave impedance technology is utilized for fine recognition, the wire for impedance measurement is only arranged in the high-risk pipe section, construction and use convenience is improved, and meanwhile the precision of the detection result is improved.

Systems and methods are disclosed for a lid insert and cover for use with a utility box having any type of instrument, e.g., a leak detection or geo-location instrument, with an antenna for transmitting any type of signals related to the instrument, e.g., a leak detection, geo-location or electrical current or load sensing signal. The insert may include a top plate configured to be mounted in an upper recess of the cover and a bottom plate configured to be mounted in a lower recess of the cover. The bottom plate may define a plurality of openings for fasteners, e.g., screws, bolts or other mechanical mounting mechanisms and / or zip-ties, that are configured for securing the instrument and / or its antenna within the box. When properly mounted therein, the antenna of such instruments can effectively transmit or receive signals into and out of the utility box.

A method for detecting a leak in a cooling system comprising a plurality of cooling units, a control device and a cooling system are provided. The method comprises adjusting operation of a first cooling unit of the plurality of cooling units. The method further comprises, while the first cooling unit is operating according to the adjusted operation, determining at least one parameter of the first cooling unit and operating at least one other cooling unit of the plurality of cooling units to compensate for the adjusted operation of the first cooling unit. The method further comprises comparing the at least one parameter to a reference value.

The utility model is suitable for the field of air filter leak detection, and provides an air filter scanning leak detection device, which is provided with a driving assembly, a mounting shaft, a round rod and a gear, and the driving assembly can drive a plurality of filters mounted in a frame to move to the lower part of a support frame in succession in the process of driving a rotating disc to rotate intermittently, so that the air filter scanning leak detection can be realized. Through cooperation with scanning leak detection modules arranged on the supporting frame, continuous detection of a plurality of filters can be achieved, and in the process that the filters rotate from one scanning leak detection module to the position below the other scanning leak detection module, the gears are meshed with the gear teeth on the fence, the frame and the filters can be driven to be turned over together, and therefore the filter detection efficiency is improved. In the process that the filter rotates by one circle, double-face detection of the filter can be achieved, and compared with the prior art that a device can only conduct single-face detection on the filter, the comprehensiveness and accuracy of scanning leakage detection of the filter are improved.

The invention discloses a heat exchange station leak detection positioning system based on a pressure fluctuation rule, and particularly relates to the field of monitoring positioning, and the system comprises a dynamic site selection module which is used for determining the position of a detection point through multi-pipeline parameter cooperative matching; the working condition self-adaptive module is used for dynamically adjusting a sampling strategy according to the operation load of the heat exchange station; the leakage screening module is used for establishing a leakage judgment threshold value in combination with historical data and real-time working conditions; the propagation velocity correction module is used for analyzing the propagation velocity of the pressure waves by integrating medium and pipeline characteristics; the double-constraint positioning module is used for realizing leakage point positioning by combining time difference and characteristic value double constraints; the closed-loop verification and correction module is used for verifying a positioning result through theoretical verification and actual measurement to realize digital and intelligent management; through the three-dimensional coupling feature extraction and multi-working-condition threshold value dynamic correction method, the limitation of single parameter judgment is avoided, the leakage judgment accuracy is improved, and a scientific basis is provided for operation and maintenance decision making.

The invention relates to the technical field of micro-leakage detection, in particular to an online positive and negative pressure micro-leakage detection system. According to the technical scheme, the system comprises a foaming reagent module, a positive pressure and negative pressure detection equipment module, a data acquisition unit, an online operation module, a micro-leakage positioning module and a visual interaction module. According to the invention, the advantages of multiple modules are integrated, the foaming reagent has optimized components, is clean and convenient, is suitable for wide-temperature and high-calcium magnesium ion environments, and does not need surface pretreatment; the positive and negative pressure detection module realizes safe and flexible detection and accurate impurity separation through multi-stage adjustable pressurization, anti-explosion design, an oil-free vortex pump and a condenser; the data acquisition and online operation module records data in real time and ensures continuous and reliable detection through efficient switching of A / B sides; the micro-leakage positioning module utilizes a sound wave and infrared multi-source data fusion algorithm and an intelligent optimization algorithm to realize accurate positioning of leakage points, and the whole device has the advantages of high efficiency and accuracy in detection, safety and reliability, high environmental adaptability and convenience in operation.

The invention relates to the technical field of fusion reactors, and discloses an assembly tool and an assembly method of a divertor, a box body is positioned and mounted on the assembly tool through a first positioning surface, a second positioning surface and a third positioning surface; through the first reference surface and the second reference surface, a reference coordinate system is set during assembly and detection of the target plate, and the target plate is detected and adjusted by taking the reference coordinate system as a reference, so that the profile tolerance of the target plate conforms to an error allowable range; therefore, it is guaranteed that the divertor obtained through assembling of the assembling tool has enough assembling precision; by the adoption of the assembling tool, the box body is installed on the assembling tool, then the target plate and the box body are preliminarily positioned, a reference coordinate system is established in cooperation with a laser tracker, the actual contour of the target surface of the target plate is obtained, the target plate is correspondingly adjusted, the assembling precision is improved, and normal work of a divertor is effectively guaranteed; and the normal cooling work of the cooling channel can be ensured through the leak detection test.

A leak detection system and method are provided. The system includes: a heater configured to receive a refrigerant from a refrigeration circuit, to heat the refrigerant and to thereby output a first constituent component of the refrigerant in a vapor state and output a second constituent component of the zeotropic refrigerant in a liquid state. Relative mass flows of the first constituent component and the second constituent component may be determined directly or may be inferred from a change in pressure of the refrigerant prior to and after heating or from a liquid level change of the second constituent component in a tank. One or more of the relative mass flows, the change in pressure, and the liquid level change may be compared to a predetermined value, and a refrigeration leak alert may be output based on the comparison.

The invention provides noise-based tap water pipe network leak detection equipment, and relates to the technical field of water pipe network detection. The noise-based tap water pipe network leak detection equipment comprises a shell, an upper flange, a lower flange, an edge calculation module, a communication module and a mass block. According to the invention, the environmental noise monitoring unit and the high-frequency signal acquisition unit are cooperatively deployed, and the wavelet packet decomposition algorithm is operated in real time by adopting edge calculation, so that the problem that a traditional single-point sensor cannot distinguish pipeline vibration, water flow turbulence and leakage signal spectrum characteristic difference is effectively solved, and the leakage characteristic identification accuracy is relatively improved by more than 20%; through the synergistic effect of the laser micro-woven annular groove, the liquid metal sealing structure and the composite elastic sealing structure, the problem of signal distortion caused by sealing failure of an existing leak detection instrument in a humid and corrosive underground environment is solved, and it is ensured that equipment can maintain stable high-frequency vibration signal acquisition capacity for a long time under severe working conditions.

The utility model discloses an SF6 gas insulated metal-enclosed switchgear intelligent leak detection device, which comprises a host and a plurality of detection parts, the detection parts are in contact with the host through gas pipes, each detection part comprises a gas filter, a gas mold cover and a sealing band, the gas filter is connected to a hand valve switch on the gas mold cover through a gas pipe, and the sealing band is connected to the gas mold cover through a gas pipe. The end part of the gas mold cover is connected with a sealing strap, the sealing strap is bound on the SF6 gas insulated metal-enclosed switchgear, and the gas filter is connected to a host through a gas pipe and is switched on and off in real time by using a hand valve. The SF6 gas-insulated metal-enclosed switchgear leak detection device has the advantages that the detection points are wrapped by the gas mold cover with excellent sealing performance, the detection part is conveniently fixed on the SF6 gas-insulated metal-enclosed switchgear by the sealing bridle, a plurality of leakage points can be detected at the same time, each detection point can be independently controlled, and the leak detection efficiency is high; the gas mold cover can be repeatedly used, the labor intensity is small, installation and operation are convenient, and the maintenance complexity is also reduced.

The invention belongs to the technical field of structural component testing, and discloses a leakage detection tool for GIS equipment, which comprises a leakage detection work frame assembly, a leakage detection sealing assembly is arranged at the top of the leakage detection work frame assembly in an elastic jacking and penetrating sliding manner, and a leakage detection tool table assembly is arranged at the bottom of the leakage detection work frame assembly in a lifting manner. The two ends of the leak detection tool table assembly are respectively used for limiting and fixing a cover body and a shaft body of the GIS equipment assembly during leak detection through a group of leak detection limiting assemblies, and the bottom of the leak detection tool table assembly is provided with a leak detection reset assembly which is used for controlling the GIS equipment assembly to carry out closing leak detection and opening leak detection. Comprising cooperation of a reset tooth arm and a driving gear and cooperation of a shifting arm, a traction arm and a leak detection limiting assembly, free switching of a GIS equipment assembly among a closing leak detection mode, an opening leak detection mode and a dynamic leak detection mode is achieved, and leak detection processing of the device is facilitated.

The disclosure relates to a leak detection system for a pneumatic braking system of a vehicle. The pneumatic braking system has a control unit, a pneumatic circuit, and an air tank configured to supply pneumatic pressure to the pneumatic circuit. The pneumatic circuit has a plurality of valves configured to be individually controlled by the control unit. The link detection system includes program code configured, when executed by a processor, to individually control the valves of the pneumatic circuit so as to isolate a sub-circuit of the pneumatic circuit. The leak detection module is configured to monitor a pressure in the pneumatic circuit for leaks, wherein the leak detection module detects a leak in the sub-circuit if the pressure decreases while the sub-circuit is isolated.

The invention relates to an air tightness detection device and method based on resonance frequency shift of a resonant cavity, belongs to the technical field of air tightness detection, and aims to solve the problems of high equipment cost, complex operation, low efficiency and difficulty in realizing multi-parameter quantitative detection in the traditional air tightness detection technology. High-purity helium with rated pressure is filled into a to-be-detected piece through a gas compressor, the to-be-detected piece is placed in a volume-adjustable detection container after pressure maintaining, and leaked helium is pumped by a vacuum pump to flow into a resonant cavity; resonance frequency deviation caused by helium concentration changes is monitored, temperature and humidity calibration data are combined, three air tightness parameters including the leakage rate (Pa.m / s), the leakage speed (mL / min) and the leakage pressure (Pa) are calculated in real time through a microcontroller, and the air tightness grade of a to-be-detected piece is output according to the air tightness detection standard. According to the device, the volume of the detection container is adaptively adjusted through the sliding rail, and the detection efficiency and adaptability can be remarkably improved by combining frequency mixing excitation and a rapid frequency calibration technology; according to the method, resonance frequency shift-helium concentration-airtightness parameters are converted, and high-precision, multi-parameter and integrated airtightness detection is realized. The invention has the advantages of fast response speed, strong anti-interference performance, low cost and the like, is suitable for fast automatic leak detection of industrial products such as sealing devices, pressure containers and the like, and provides a new scheme for trace gas detection and gas tightness grading.

The invention relates to the technical field of leakage ultrasonic detection, and discloses SF6 gas leakage ultrasonic detection equipment for a GIS pipeline, which comprises a handheld mechanism, a leakage detection mechanism, a switching mechanism and a controller, and is characterized in that the leakage detection mechanism is arranged on the handheld mechanism and comprises a first sensor mounting position and a posture adjusting mechanism; the posture adjusting mechanism is configured to drive the sensor mounted on the first sensor mounting position to perform posture adjustment; the switching mechanism is arranged on the handheld mechanism and comprises a second sensor mounting position; the controller is arranged in the handheld mechanism and is electrically connected with the leak detection mechanism and the switching mechanism; the equipment is at least provided with the ultrasonic micro-leakage fine detection sensor and the ultrasonic global quick scanning sensor, the ultrasonic micro-leakage fine detection sensor and the ultrasonic global quick scanning sensor can be exchanged between the first sensor mounting position of the leakage detection mechanism and the second sensor mounting position of the switching mechanism through the switching mechanism, and the detection efficiency and accuracy are remarkably improved.

The loop type natural gas secondary calibration system comprises a loop unit, a flow and pressure control unit, a temperature control unit, a leak detection unit and a control unit, and natural gas circulates in the loop unit; the flow and pressure control unit, the temperature control unit and the leak detection unit are electrically connected with the control unit; the flow and pressure control unit is used for stabilizing the flow and pressure of the natural gas in the loop unit; the temperature control unit is used for stabilizing the temperature of natural gas in the loop unit; the leak detection unit is used for detecting whether two ends of the detected flowmeter leak or not; the loop unit is further provided with a first electric ball valve, a differential pressure transmitter, a second temperature transmitter and a sonic nozzle standard meter. The calibration method comprises the steps of calibration operation preparation, calibration operation and leak detection operation. The system can simplify the natural gas verification and adjustment process, reduces the dependence on a natural gas transmission station, has the leakage detection capability, and is worthy of popularization and application.

The invention provides a gas leakage detection method, device and system and computer equipment, and the method can preliminarily position a potential leakage gas region through detecting a moving target region in an infrared image, further recognize the type of the moving target region, eliminate an interference target region, and finally, determine the leakage gas region. And judging whether the suspicious gas region is a leakage gas region or not only based on the gray data of the suspicious target gas region. The data volume detected only based on the gray data of the suspicious gas region is far lower than the data volume detected based on the whole infrared image, so that the computing power requirement of target detection operation on a hardware platform can be reduced, and whether the region is the leaked gas region is judged only based on the gray data of the suspicious gas region, so that the detection efficiency is improved. And for the detection of the leaked gas area, interference targets are reduced, the real leaked gas area can be more accurately classified from the suspicious gas area, and the false detection rate is reduced.

A battery case leak testing device includes a housing having a testing port. A pump is in fluid communication with the testing port and is configured to generate a prescribed pressure at the testing port. A pressure sensor is in fluid communication with the testing port to measure fluid pressure at the testing port. A controller is in operative communication with the pump and the pressure sensor, and is configured to facilitate operation of the pump in a first mode and a second mode. In the first mode, the pump operates to generate a first prescribed pressure at the testing port. In the second mode, the pump operates to generate a second prescribed pressure at the testing port. The controller transitions the pump from the first mode to the second mode when the pressure sensor measures a prescribed pressure characteristic when the pump is operating in the first mode.

A method and system for detecting the integrity of a high-voltage (HV) battery enclosure. The system uses sensors located within the battery enclosure to generate pressure and temperature sensor data. The system analyzes these data to detect any lack of integrity in the enclosure, such as leaks or breaches, which could compromise the battery's performance and safety. Upon detecting a potential integrity issue, the system generates an alert.

The invention relates to the technical field of refrigeration system fault detection, and discloses a deep separable convolution-BiLSTM multi-head attention refrigerant leakage detection method. The method comprises the following steps: acquiring historical data of a unit and performing standardized preprocessing; constructing a time sequence sample by adopting a sliding window method; building a parallel hybrid model of deep separable convolution and BiLSTM (Bidirectional Long Short-Term Memory Network), and fusing local space features and global time sequence features through channel splicing; a multi-head attention mechanism is introduced to dynamically weight key time steps, and a dynamic category weight compensation strategy is adopted to optimize a loss function; and training the model by using an adaptive learning rate and an early stop strategy, and finally outputting a high-precision fault diagnosis result. According to the method, the accuracy rate of a test set reaches 98%, the AUC fraction reaches 0.99 and is improved by 3%-15% compared with a traditional mixed model, the detection efficiency and robustness of the refrigerant leakage fault of the refrigeration system are remarkably improved, and the method is suitable for accurate maintenance in a high-precision industrial scene.