Radar level monitoring system based on intelligent sensors and data transmission

The radar level monitoring system, which uses intelligent sensors and data transmission, dynamically maintains false echo patterns and divides time slots in star-flash communication, enabling stable level measurement and accurate transmission under changing mud tank conditions and electromagnetic interference environments. This reduces false alarm and missed alarm rates and adapts to complex drilling site conditions.

CN123111412APending Publication Date: 2026-11-17东营职业学院 +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611208264.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-11
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing radar level gauges are prone to loss of lock under frequent changes in mud tank operating conditions and lack measurement confidence characterization. Abnormal data transmission delays are difficult to guarantee under strong electromagnetic interference environments, and the lack of multi-point data fusion in upper-level monitoring leads to a high false alarm and missed alarm rate.

Method used

A radar level monitoring system based on intelligent sensors and data transmission is adopted, including a tank-top intelligent radar monitoring terminal, a star-flash gateway node, and a field control host. The system calculates the true liquid level score by dynamically updating the false echo spectrum, signal-to-noise ratio, locking continuity, prediction consistency, and occlusion prior. Data transmission is carried out in the star-flash communication unit by dividing emergency time slots and normal time slots, and multi-tank fusion discrimination is performed on the host side.

Benefits of technology

It enables stable data measurement, deterministic low-latency transmission, and fusion early warning under complex operating conditions, reduces false alarm and missed alarm rates, ensures the reliability and accuracy of liquid level measurement, and adapts to frequent relocation and electromagnetic interference environments at drilling sites.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123111412A_ABST
    Figure CN123111412A_ABST
Patent Text Reader

Abstract

This invention relates to the field of level measurement technology and discloses a radar level monitoring system based on intelligent sensors and data transmission to solve the technical problems of easy loss of lock-in in level measurement and false alarms / missed alarms caused by the complex operating conditions of mud tanks. The system includes an intelligent radar monitoring terminal on the tank top, a star-flash gateway node, and a field control host. The terminal maintains a dynamic false echo spectrum according to operating conditions, performs forgetting-style updates between protected areas within a stable window, and scores candidate peaks based on overlap and occlusion priors to lock the liquid level and output confidence scores. The gateway prioritizes scheduling alarm data based on emergency time slot reservations during the scheduling cycle and terminal requests. The host constructs a fusion anomaly index based on the total liquid volume change deviation, average confidence loss, and the proportion of suspicious nodes to determine early warnings. This invention achieves stable measurement and reliable multi-tank linkage alarms under complex operating conditions, meeting safety monitoring requirements.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of level measurement technology, and in particular to a radar level monitoring system based on intelligent sensors and data transmission. Background Technology

[0002] In oil and gas drilling operations, drilling fluid flows back and forth along the drill pipe, wellbore, and surface circulation system. Various mud tanks on the well site, including circulation tanks directly involved in circulation, reserve tanks storing spare mud, and mixing tanks used for weighting and mixing, collectively undertake the tasks of storing, preparing, and circulating drilling fluid. The fluid level in the tanks and the total fluid volume calculated from it are the most direct surface evidence for assessing the wellbore condition. When formation fluid intrudes into the wellbore, the outflow rate exceeds the inflow rate, causing an abnormal rise in the surface tank fluid level, resulting in an overflow. When drilling fluid leaks into the formation, the outflow rate decreases, causing an abnormal drop in the tank fluid level, resulting in well leakage. If these two types of anomalies are not detected and addressed early, they may escalate into serious accidents such as blowouts and stuck pipe. Therefore, continuous and accurate monitoring of the tank fluid level, and timely judgment of changes in the total fluid volume, has become a fundamental aspect of drilling safety monitoring.

[0003] Liquid level measurement methods have evolved from manual gauging, float-based, and hydrostatic methods to ultrasonic and radar methods. Frequency-modulated continuous wave radar, with its non-contact measurement, insensitivity to medium density, and tolerance to temperature and pressure fluctuations, has been gradually applied in mud tank monitoring in recent years. However, the operating conditions of mud tanks are extremely harsh. The tank contains fixed components such as agitators, heating coils, ladders, and welds. The medium surface is accompanied by foam, splashes, and high-temperature steam, and the antenna aperture is easily adhered to by mud. These factors superimpose a large number of false echoes on the radar range spectrum. Existing radar level gauges generally undergo a one-time empty tank learning during the commissioning phase to generate a static false echo suppression curve and use it in conjunction with a fixed threshold. However, changes in agitator speed, feed nozzle impact, foam layer evolution, and increased material buildup can cause the position and intensity of false echoes to continuously drift. The static suppression curve quickly becomes mismatched after changes in operating conditions, and the instrument may lock onto the echoes from the agitator blades or internal tank components, resulting in fluctuating or even continuously erroneous liquid level readings. Usually, manual readjustment is required to restore the readings. At the same time, such instruments only output the liquid level value itself, without carrying any representation of the reliability of the value. The host system has no way of distinguishing whether a reading comes from the actual liquid level or from interference echoes.

[0004] There are also constraints at the data transmission and comprehensive judgment levels. Well sites have numerous tanks, are geographically dispersed, and are frequently relocated with drilling rig relocations. Wired cabling using current loops or fieldbuses involves a large amount of work and is difficult to migrate and reuse. Common wireless telemetry solutions are mostly based on contention-based access, making it difficult to guarantee transmission delays in environments with strong electromagnetic interference from well site frequency converters and high-power motors. Furthermore, data from each measuring point is polled and uploaded in a round-robin fashion, meaning that alarm data may be prioritized after regular data during abnormal moments of drastic liquid level changes and increased channel load. Upper-level monitoring typically uses instantaneous exceedances of single tank liquid levels as alarm criteria, directly amplifying occasional echo interference from a single measuring point into a process alarm. However, the total liquid volume change across multiple tanks, which truly reflects overflow and well leakage trends, lacks a fusion judgment method that combines measurement confidence levels, resulting in a long-term coexistence of false alarms and missed alarms.

[0005] In view of this, how to enable radar level measurement to reliably identify the true liquid level and provide confidence characterization in the mud tank environment where operating conditions change frequently, while ensuring timely and reliable transmission of abnormal data from multiple tanks in the well site environment and making reliable fusion alarm judgment based on it, is a technical problem that has not yet been effectively solved in this field. Summary of the Invention

[0006] The purpose of this invention is to provide a radar level monitoring system based on intelligent sensors and data transmission to solve the technical problems of existing radar level gauges being prone to lock-up under frequent changes in mud tank operating conditions and lacking measurement confidence characterization, having difficulty guaranteeing abnormal data transmission delay under strong electromagnetic interference environment, and having a high false alarm and missed alarm rate due to the lack of multi-point data fusion in upper-level monitoring.

[0007] To achieve the above objectives, the present invention provides a radar level monitoring system based on intelligent sensors and data transmission. The system includes several intelligent radar monitoring terminals installed on the top of the monitored tank, a satellite gateway node wirelessly connected to the terminals, and a field control host connected to the satellite gateway node. Each intelligent radar monitoring terminal includes: a radar measurement unit for periodically transmitting frequency-modulated continuous wave signals and receiving echoes, generating a range spectrum containing several candidate peaks; and an auxiliary sensing unit for sensing and acquiring the current operating condition tag of the monitored tank. The echo analysis unit is used to independently maintain the corresponding dynamic false echo spectrum according to the operating condition label, and when the stable window condition is met, to perform a forgetting update on the dynamic false echo spectrum to which the amplitude within the current frame distance spectrum is located within the update upper bound range. The update upper bound is determined based on the smaller distance between the current locked liquid surface peak position and the candidate peak position with the highest signal-to-noise ratio in the current frame, and a protection interval is reserved on both sides of the corresponding peak position where no update is performed. The unit calculates the true liquid surface score of the candidate peak based on the signal-to-noise ratio, locking continuity, prediction consistency, false echo overlap with the dynamic false echo spectrum, and the occlusion prior determined according to the fixed component position. The unit locks the liquid surface according to the highest score and the score difference, and outputs the liquid surface measurement result including the confidence level calculated based on the highest score. The edge decision unit is used to classify the operating status of the terminal into normal, suspicious, or alarm states according to the liquid surface measurement result, and to classify the data to be sent into at least a state group including the confidence level and an alarm group generated in the alarm state. The StarFlash communication unit is used to send the alarm group within a reserved emergency time slot by sending a request with an alarm indication to the StarFlash gateway node. The StarFlash gateway node includes a group scheduling unit, which is used to broadcast a time slot table containing an emergency request window and emergency sub-time slots at fixed intervals. Upon receiving the request, it assigns the emergency sub-time slot to the corresponding alarm terminal to send the alarm group data. Under normal circumstances, it allocates the remaining time slots to other data groups. The field control host includes a multi-tank fusion discrimination unit, which is used to uniformly convert the liquid level of each monitored tank into volume, calculate the fusion anomaly index based on the estimation deviation of the total liquid volume change rate of multiple tanks, the average confidence loss normalized according to the number of monitored tanks, and the proportion of terminal nodes in a suspicious state. When the fusion anomaly index exceeds a preset threshold, an alarm linkage is triggered. The above system can form an effective closed loop of stable data measurement, deterministic low-latency transmission, and fusion early warning under complex operating conditions.

[0008] As a preferred embodiment of the present invention, the specific rules for updating the dynamic pseudo-echo spectrum by the echo analysis unit when the stable window condition is met are as follows: The smaller distance between the currently locked liquid surface peak position and the candidate peak position with the highest signal-to-noise ratio in the current frame is taken as the upper limit of the update, and a protection interval is formed by extending a preset number of distance units to both sides of the corresponding peak position; the distance units within the upper limit of the update but not belonging to the protection interval are updated by weighting the amplitude of the current frame distance spectrum with the amplitude of the original spectrum according to an exponential forgetting factor; the distance units within the protection interval and on the side farther from the upper limit of the update maintain the original dynamic pseudo-echo spectrum amplitude unchanged. This prevents the constant liquid surface from being mislearned and cuts off the self-reinforcing path of erroneous interference.

[0009] As a preferred embodiment of the present invention, when the echo analysis unit calculates the true liquid level score, it calculates the continuity score by the difference between the candidate peak position and the locked liquid level position at the previous moment; it extracts the liquid level distance change rate estimate by the reliable locking frame within the sliding time window to predict the current short-term liquid level distance, and calculates the prediction consistency score based on the difference between the candidate peak position and the short-term liquid level distance; it calculates the false echo overlap penalty by using the amplitude of the preset neighborhood of the distance unit where the candidate peak is located in the corresponding dynamic false echo spectrum; it calculates the occlusion penalty based on whether the candidate peak falls into the pre-marked distance range of the fixed components inside the tank and the transition zones on both sides, and when the short-term liquid level distance falls into a certain fixed component distance range or transition zone, the corresponding occlusion penalty for the candidate peak whose deviation from the short-term liquid level distance is less than a preset scale parameter is temporarily exempted to prevent the liquid level from being mistakenly suppressed when it crosses the depth of the component.

[0010] As a preferred embodiment of the present invention, the liquid level locking logic of the echo analysis unit includes: removing candidate peaks with scores not higher than zero from the current frame, selecting the winning peak with the highest score and not lower than the score confidence threshold from the remaining candidate peaks; when the winning peak deviates from the current locking position, and the difference between the winning peak score and the second highest score is not less than the difference threshold, entering the switching confirmation process, and performing locking position switching after maintaining dominance for multiple consecutive frames; when the difference between the winning peak score and the second highest score is less than the difference threshold, keeping the previous confidence locking position unchanged, marking the operating status of the corresponding terminal as suspicious, and triggering a request to send diagnostic group data; when the number of consecutive suspicious states reaches a preset number of frames, clearing the sliding time window and entering the recapture mode, re-selecting the peak value to establish locking based on the capture score that does not include the continuity item and the prediction consistency item within the full range, and calculating the liquid level change rate during the lockout period after re-locking and sending it to the alarm judgment, thereby significantly reducing lock jumps and false locks.

[0011] As a preferred embodiment of the present invention, the auxiliary sensing unit includes a triaxial vibration sensor; the stable window condition requires the following criteria to be met simultaneously: the terminal continuously obtains a reliable liquid level lock within a preset number of sampling periods, and the liquid level change rate calculated based on the reliable lock sequence is lower than a preset change rate threshold; the vibration index obtained based on the triaxial vibration sensor falls within the normal vibration range corresponding to the current working condition label; wherein, the upper limit of the normal vibration range corresponding to different working condition labels is set in stages according to being above the typical operating vibration level of the working condition and below the impact transient level; there is no liquid replenishment or drainage action; wherein, for the nozzle impact working condition accompanied by feeding, its separate update condition is that the feeding state is constant, a reliable lock is obtained every frame, and the liquid level change rate matches the feeding conversion rate.

[0012] As a preferred embodiment of the present invention, the edge decision unit divides the data to be transmitted into an alarm group, a status group, and a diagnostic group. The alarm group includes a timestamp, liquid level value, confidence level, alarm type code, and a liquid level change rate summary. The alarm group is generated only when the terminal liquid level exceeds the limit, the liquid level change rate exceeds the standard, or the number of consecutive periods in a suspicious state reaches a preset upper limit. The status group includes node status, liquid level value, confidence level, temperature, vibration index, and link quality, and is sent with the normal sampling cycle. The diagnostic group includes a candidate peak list, local echo feature vectors, false echo spectrum increments, and echo summaries compressed by differential coding. It is generated and uploaded when the terminal is in a suspicious or alarm state for traceability and evidence collection. This significantly reduces air interface occupancy while completely preserving the evidence chain at abnormal moments.

[0013] As a preferred technical solution of the present invention, the group scheduling unit of the StarSpark gateway node implements an emergency time slot scheduling mechanism that combines reservation and borrowing: Each scheduling cycle includes a beacon, an emergency request window, an emergency indication micro-time slot, a normal time slot, and the emergency sub-time slot. The emergency request window is composed of request micro-time slots that are fixedly mapped and allocated to each terminal. When there are no alarm group messages in the entire network, the emergency sub-time slot is temporarily conditionally authorized and borrowed to the other messages with the highest priority in the system, and it is agreed that the authorization will only take effect when the current emergency request window is silent. When any terminal generates an alarm group message and sends a request in its corresponding request micro-time slot, the StarSpark gateway node declares that the emergency sub-time slot belongs to the alarm terminal in the immediately following emergency indication micro-time slot. The terminal holding the conditional authorization immediately abandons the borrowing and sending, thereby ensuring that the alarm group message obtains a deterministic sending delay and preventing alarm queuing backlog.

[0014] As a preferred technical solution of the present invention, the remaining time slots not polled and arranged by the normal time slots are immediately incorporated into the elastic time slot pool on the network side; for the diagnostic group messages and low-priority retransmission messages after link recovery, the packet scheduling unit determines the transmission order in the elastic time slot pool based on the comprehensive priority sorting; the comprehensive priority is obtained by normalizing and weighting the event level corresponding to the data group to which the message belongs, the waiting time of the message in the transmission buffer, the critical weight of the monitored tank to which the message belongs, the retransmission request flag, and the message length.

[0015] As a preferred technical solution of the present invention, the calculation and linkage logic of the multi-tank fusion discrimination unit for the fusion anomaly index is as follows: Within the calculation period, the effective reported volume of each monitored tank is fitted using a sliding window least squares method to obtain the estimated value of the current total liquid volume change rate of the multi-tank system; the reference change rate of the total liquid volume calculated from the operating condition reference information is obtained, and the absolute difference between the two is normalized by the reference flow rate and limited to obtain the first term; the average confidence loss of all monitored tanks is obtained to obtain the second term; the proportion of terminal nodes currently in a suspicious state to the total number of monitored nodes in the system is statistically analyzed to obtain the third term; the first, second, and third terms are weighted and summed to output the fusion anomaly index; for terminal nodes in a suspicious state, the volume is extrapolated using the most recent reliable change rate before entering the suspicious state and included in the total volume; when the extrapolation continues for more than a preset time, its self-fitted sequence and multi-tank summation are removed; When any terminal directly reports a single-point limit exceedance or rate of change alarm, and the monitored tank is a circulating tank involved in drilling fluid circulation or multiple tanks triggering concurrent alarms, a secondary alarm is directly triggered and the corresponding diagnostic data is retrieved. This mechanism can limit occasional interference to the diagnostic level at the source, effectively reducing the system's false alarm and missed alarm rates.

[0016] As a preferred embodiment of the present invention, the field control host further includes a historical model management unit, used to classify and archive the dynamic false echo spectrum versions, threshold parameters, and event samples of each terminal according to tank number, operating condition label, and environmental segment. When a terminal's dynamic false echo spectrum changes at a fixed period or when the cumulative change exceeds a preset range, the diagnostic team sends a full snapshot of the current dynamic false echo spectrum in segments. During the observation period after the new version snapshot is registered, if the average confidence level of the corresponding monitored tank decreases by more than a preset range compared to the previous version period, the historical model management unit sends the dynamic false echo spectrum of the previous version to the corresponding terminal through the StarSpark gateway node for full coverage restoration and temporarily tightens the spectrum update parameters of that terminal. This mechanism facilitates the connection between the terminal's local real-time adaptation and the long-term parameter consistency of the entire system.

[0017] Compared with existing technologies, this invention maintains dynamic false echo maps for different operating conditions, strictly restricts map updates to within a stable window, and shields distance cells beyond the actual reflecting surface neighborhood and suspected liquid surface echoes from being written into the map. Candidate peak scores are jointly constructed using signal-to-noise ratio, locking continuity, prediction consistency, false echo overlap, and occlusion prior. When the liquid surface crosses the depth of internal tank components, the occlusion penalty is temporarily exempted based on the predicted position. The locking of the actual liquid surface no longer relies on the static suppression curve generated once during commissioning, nor is it affected by long-term constant liquid level or... It can lock short-term deviations and avoid mislearning them as false echoes. It can remain stable under changing operating conditions such as stirring start-up and shutdown, speed switching, foam evolution, and viscosity and weighting. When there is ambiguity in the scoring, it retains the previous reliable result and requests diagnostic data. After the lock is lost, it can re-lock through independent capture scoring and calculate the liquid level change rate during the lock loss period. The workload of reading the liquid level output jump and the readjustment after the operating conditions change is reduced. Rapid liquid surface movement is not missed due to tracking lock loss. Moreover, each measurement is accompanied by confidence output. The upper-level discrimination thus obtains the basis for measuring the quality of single-point data.

[0018] This invention divides the data to be sent into alarm groups, status groups, and diagnostic groups on the terminal side. Alarm group messages are sent through reserved emergency time slots divided into several sub-time slots. The remaining messages are sent in the order arranged by the network side according to event level, waiting time, tank criticality, retransmission requirements, and message length. Emergency sub-time slots can be conditionally borrowed for other messages. When an alarm occurs, it is immediately reclaimed through emergency request micro-time slots corresponding to each terminal. Under normal circumstances, the air interface only carries a small amount of characteristic data, allowing battery-powered terminals to operate at low power for a long time. When an anomaly occurs, alarm messages are sent in queue. Under the condition that the message is successfully transmitted once, the delay from the generation to the transmission of a single alarm does not exceed the sum of one scheduling cycle and a fixed offset. The completion delay of multi-point concurrent alarms has a definite upper bound that can be calculated with the number of concurrent alarms, and is not delayed by regular data occupying the channel. The original evidence required for interpretation is sent on demand, avoiding continuous occupation of air interface resources.

[0019] This invention converts the liquid level of each tank into volume on the host side. The fusion anomaly index is jointly constructed by the deviation of the total liquid volume change rate normalized to the reference flow rate, the confidence loss normalized by the number of tanks, and the proportion of suspicious nodes. It is supplemented by the direct-pass criterion for single tank over-limit alarm. The establishment of the alarm is jointly determined by the liquid level value, confidence, total liquid volume, and operating condition reference. Occasional echo interference from a single measuring point is limited to the diagnostic level and does not directly trigger process alarms. The overflow and well leakage trends reflected by the total liquid volume of multiple tanks can be identified by level before a single tank exceeds the limit. The false alarms and false alarms are reduced simultaneously.

[0020] In addition, once the terminal is powered on and registered, it can obtain the tank number, reference height and conversion table number and other configurations for monitoring, which is suitable for the deployment method of frequent relocation at the drilling site; the host side archives, reviews and writes back the false echo spectrum version, threshold parameters and event samples, so that the real-time adaptation of the terminal side is connected with the long-term parameter consistency of the whole system, and the discrimination parameters can be continuously corrected with the running data. Attached Figure Description

[0021] Figure 1 This is a system architecture diagram of a preferred embodiment of the present invention.

[0022] The components are as follows: 100, Intelligent Radar Monitoring Terminal on Tank Top; 101, Radar Measurement Unit; 102, Echo Analysis Unit; 103, Auxiliary Sensing Unit; 104, Star Flash Communication Unit; 105, Edge Decision Unit; 200, Star Flash Gateway Node; 201, Network Clock and Access Management Unit; 202, Group Scheduling Unit; 203, Reassembly Confirmation Unit; 204, Upper-Level Interface Unit; 300, Field Control Host; 301, Volume Conversion Unit; 302, Multi-Tank Fusion Discrimination Unit; 303, Alarm Linkage Unit; 304, Historical Model Management Unit. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the present invention. Without conflict, the technical features in the following embodiments can be combined with each other.

[0024] See Figure 1This embodiment provides a radar level monitoring system based on intelligent sensors and data transmission, including several tank-top intelligent radar monitoring terminals 100 installed on the top of each monitored tank, a star-flash gateway node 200 deployed in the well site control room or explosion-proof cabinet, and a field control host 300 connected to the star-flash gateway node 200. The monitored tanks are various mud tanks in the drilling site, classified by purpose into circulation tanks that directly participate in drilling fluid circulation, reserve tanks that store spare mud, and mixing tanks that are used for weighting and preparation operations; this classification will be used uniformly thereafter. In a specific deployment example, a well site has a total of eight monitored tanks, including four circulation tanks, two reserve tanks, and two mixing tanks, with one tank-top intelligent radar monitoring terminal 100 installed on the top of each tank. Each tank-top intelligent radar monitoring terminal 100 serves as a terminal node in the StarScan network, connecting to the StarScan gateway node 200 via a short-range wireless link. The StarScan gateway node 200 acts as a management node in the StarScan network, responsible for network-wide time synchronization, access management, and time slot scheduling of grouped data. It then transmits the reconstructed data to the field control host 300 via a wired upper-level link. The field control host 300 performs volume conversion, multi-tank fusion discrimination, hierarchical alarm linkage, and historical model management. The measurement and confidence level determination of a single tank are completed at the terminal side, the transmission order of multiple sets of data is arranged by the network side according to event value, and the comprehensive discrimination of multi-tank operating conditions is completed at the host side. These three levels together constitute a complete monitoring closed loop from echo acquisition to alarm archiving.

[0025] The intelligent radar monitoring terminal 100 on the tank top includes a radar measurement unit 101, an echo analysis unit 102, an auxiliary sensing unit 103, a star-flash communication unit 104, and an edge decision-making unit 105. Structurally, the intelligent radar monitoring terminal 100 is encapsulated in a shell that meets the explosion-proof requirements of the well site. It can be designed as either explosion-proof or intrinsically safe depending on the hazardous area of ​​the site. The shell protection rating is no lower than IP67, and the operating temperature range of the electronic components covers -40 degrees Celsius to 85 degrees Celsius. The intelligent radar monitoring terminal 100 is installed through a pre-reserved flange opening on the tank top. The flange specification is, for example, DN80 or larger. The antenna aperture is vertically downward, aligned with the liquid surface, and the horizontal distance between the antenna and the tank wall is maintained at no less than 300 mm to reduce the chance of multiple reflections from the wall surface entering the antenna main lobe. During the initial survey phase, the installation point avoids the area directly above the agitator blades and the direct impact area of ​​the feed nozzle. To address the issues of mud splashing and adhesion, a polytetrafluoroethylene (PTFE) anti-adhesion cover can be added to the outside of the antenna. This material has low surface energy, is not prone to mud adhesion, and its attenuation of millimeter-wave signals is negligible. Power can be supplied by lithium battery packs, such as 19 Ah lithium thionyl chloride batteries, or by using a readily available 24V DC power supply. When battery powered, the intelligent radar monitoring terminal 100 on the tank top enters sleep mode during non-transmit / receive periods, relying on the listening cycle mechanism of the star-flash communication unit 104, to ensure long-term maintenance-free operation.

[0026] The radar measurement unit 101 employs an 80GHz band frequency-modulated continuous wave radar chip, transceiver antenna, and analog-to-digital sampling circuit. During operation, it periodically transmits continuous wave signals using a linear frequency modulation (LFM) method. The echo signal is mixed with the local oscillator signal to obtain the beat frequency signal. Let the sweep slope of the transmitted signal be... The actual distance from the antenna aperture to the reflector is The shooting frequency is All three conditions are met.

[0027] in It is the speed of light. In one specific embodiment, the sweep bandwidth is... Using 4 GHz as the reference frequency and a single sweep duration of 1 millisecond, a frame consists of 16 consecutive sweeps. The beat frequency signal is sampled analog-to-digitally at a rate of several megahertz. After adding a Hanning window to suppress spectral sidelobes, a 2048-point Fast Fourier Transform is performed on the sampled sequence to obtain the range spectrum. The spectral amplitudes of each sweep within a frame are incoherently accumulated to improve the signal-to-noise ratio. The frame mean referred to below is the range spectrum obtained after this incoherent accumulation and its corresponding complete measurement. The range resolution is determined by the sweep bandwidth and satisfies the following conditions.

[0028] With a bandwidth of 4 GHz, the distance is approximately 3.75 cm. Further parabolic interpolation of the spectral peaks enables sub-range cell localization. Combined with multi-frame smoothing, the measurement deviation of the static liquid level can be reduced to the millimeter level. The tank top mounting reference height is recorded as... The corresponding distance is calculated from the position of the spectral peak. Afterwards, the liquid level height is

[0029] The measurement range is configured according to the tank height, for example, from 0.3 meters to 20 meters. Several range elements at the front of the range spectrum corresponding to the antenna near-field and feed leakage are permanently shielded and do not participate in subsequent processing. The antenna uses a lens design, with the beam angle controlled between 3° and 8°. The narrow beam helps reduce the chance of tank walls and internal components entering the antenna main lobe. Under normal conditions, the sampling period is 1 to 5 seconds, with each sampling period completing one frame of measurement. When the intelligent radar monitoring terminal 100 on the tank top is in a suspicious or alarm state, the sampling period is temporarily shortened, for example, to 1 second, to increase the density of observations. The temperature data provided by the auxiliary sensing unit 103 is used to correct the sweep slope and the drift of the ranging reference with temperature according to a pre-calibrated curve. A high-frequency modulated continuous wave system is chosen instead of an ultrasonic method because foam, dust, high-temperature steam, and temperature and pressure fluctuations are common inside mud tanks. Non-contact millimeter-wave measurement is more robust in such environments, and the continuous wave system is less prone to loss of lock on the liquid surface under agitation.

[0030] The echo analysis unit 102 performs constant false alarm rate (CFAR) detection on the range spectrum of each frame, and includes spectral peaks with local signal-to-noise ratios exceeding the detection threshold into the candidate peak set. Each candidate peak records its distance. With local signal-to-noise ratio The peak amplitude, peak width, and other peak shape parameters are also noted. These peak shape parameters are not included in the scoring described later; they are only submitted with the diagnostic team's report for the host side to review and verify. Unlike the conventional practice of directly taking the strongest peak as the liquid level, echo analysis unit 102 maintains dynamic pseudo-echo spectra divided according to operating conditions. ,in For operating condition status labels, For distance element indexing. Operating conditions must be at least divided into stop-mixing and stop-stirring states. Low-speed stirring High-speed mixing and nozzle impact The four levels each correspond to an independently maintained false echo pattern. The reason for stratifying by operating condition is that the reflection characteristics of internal tank components such as agitator blades, heating coils, and ladders will drift with changes in agitation speed, liquid surface disturbance, adhesion degree, and the steam space inside the tank. Adhesion at the antenna aperture will also form a gradually changing false echo family at the near distance, and the temperature gradient of the steam space will also cause a slight shift in the apparent distance. These effects are all closely related to the operating condition. A single static false echo template is often no longer effective after the operating condition is switched. If a fixed threshold suppression is used, the real liquid surface may be incorrectly suppressed.

[0031] Each false echo pattern is updated only within a stable window. The stable window is contingent upon the intelligent radar monitoring terminal 100 on the tank top continuously acquiring a reliable lock; its conditions include, for example, the most recent... A reliable liquid level lock is achieved in each frame within each sampling period, and the rate of change of liquid level calculated from this reliable lock sequence is lower than a threshold. The vibration index falls within the normal vibration range corresponding to the current operating condition label; there is no fluid replenishment or drainage operation. For example, take 20. For example, a value of 0.5 mm / s. Vibration criteria are set in tiers according to operating conditions: a label is assigned to each operating condition. The preset normal vibration range has an upper limit. The upper limit of the stop-stirring setting is close to the static background, while the upper limit of the high-speed stirring setting accommodates steady-state vibration at that speed. This condition thus refines the function from eliminating all vibrations to eliminating transient vibrations exceeding the normal level of the current operating condition, without blocking the updating of the high-vibration operating condition settings. When the process status input described later is connected, the replenishment / drainage action is directly determined by the pump and valve operation signal; otherwise, it is overridden by the liquid level change rate criterion. Nozzle impact condition. Feeding is always present, and its update conditions are separately defined as follows: the feeding state is constant, a reliable lock is obtained every frame, and the rate of change of liquid level matches the rate calculated from the feed rate. This is considered a controlled update window, and the protection interval rolls with each frame's lock position. During suspicious states, alarm states, and the recapture mode described later, no spectrum update is performed, and echoes from periods of unreliable locking are therefore not written into the spectrum. When a stable window is established, the spectrum corresponding to the current operating condition is updated exponentially using the following formula.

[0032] in This represents the range spectral amplitude distribution of the current frame after incoherent accumulation. The forgetting factor has a value between 0.85 and 0.98. The update is not applied to all range cells, but only to the segment from the antenna aperture to the update upper bound: the update upper bound is the smaller of the distance between the currently locked liquid surface peak position and the candidate peak position with the highest SNR in the current frame, with a guard zone extended towards the antenna side. The guard zone is centered on the corresponding peak position, with several range cells on each side; for example, four range cells on each side during operation, equivalent to approximately 15 cm at a range resolution of 3.75 cm. Range cells within the guard zone and those beyond the update upper bound retain their original base map values. Including the candidate peak with the strongest SNR in the current frame in the upper bound constraint ensures that even if the locked position deviates, suspected liquid surface echoes significantly stronger than environmental clutter will not be written into the spectrum, thus cutting off the self-reinforcing path of mislearning. The fundamental reason for setting up the shielding is that the establishment of a stable window precisely means that the liquid level is constant for a long period of time. If the entire frame of echoes is written into the base map indiscriminately, the real liquid level peaks will accumulate frame by frame with each update and be slowly learned as false echoes. Echoes that change with the liquid level, such as multiple reflections after penetrating the liquid surface, will also be incorrectly recorded. The shielding and stable window measures work together. The former prevents the constant liquid level from being mistakenly learned, and the latter avoids writing the real liquid level that moves rapidly during the replenishment and discharge process into the base map, thus ensuring the long-term effectiveness of the map. In terms of implementation, each map resides in memory as an amplitude array corresponding to a distance unit. The total storage overhead for the four levels is within tens of kilobytes, and it is fixed to the non-volatile memory of the intelligent radar monitoring terminal 100 on the tank top on an hourly cycle. The map can be restored after a power failure and restart. The liquid level lock state is not saved with the map and is re-established after a restart through the recapture mode described later.

[0033] Based on this, the echo analysis unit 102 analyzes each candidate peak Calculate the true liquid level score

[0034] All components used in the scoring were first normalized to the range of 0 to 1. Signal-to-noise ratio (SNR) term. The amplitude is linearly normalized and limited according to a reference signal-to-noise ratio, for example, 30 dB; the continuity term characterizes the candidate peak's position relative to the previous locked liquid surface position. The degree of closeness, taking

[0035] Scale parameters The value is determined by the product of the maximum allowable replenishment rate and the sampling period, for example, by taking twice that product; the prediction consistency term is calculated as follows:

[0036] Calculate the scale parameters For example with Using the same value, predict the liquid surface distance in a short time. Depend on

[0037] Given. In the formula, The time interval between the current frame and the last frame that formed a reliable lock is equal to the sampling period when the liquid level is continuously and reliably locked. , The first one in the sliding window The sampling time of the frame and the distance of the liquid surface locked in that frame. This is the length of the sliding window, for example, 10. , These are the average values ​​of the sampling time and the locking distance for each frame within the window. This is an estimated value of the rate of change of liquid level distance, which is equal in magnitude but opposite in sign to the rate of change of liquid level. Throughout this paper, when the rate of change of liquid level is used as a criterion, its absolute value is compared. When the number of reliable locked frames within the sliding window is less than two frames... Take zero, Take immediately False echo coincidence term according to

[0038] Calculation and normalization of the spectrum ,in The equivalent threshold amplitude for constant false alarm detection is the distance cell in the spectrum where the amplitude reaches the detection threshold, which is then penalized. The distance cell index where the candidate peak is located, and the neighborhood half-width. For example, take 2; occlusion penalty. The distance range is given by the fixed components such as the agitator, coil, and internal welds of the tank, which are manually marked during the installation survey phase. When a candidate peak falls within the marked range, it is assigned a value of 1; when it falls into the transition zone on either side of the range, it linearly decays to zero based on the distance from the range boundary. The width of the transition zone is, for example, two distance units; otherwise, it is assigned a value of 0. When predicting the liquid surface distance in a short time... When the liquid falls into a marked area or its transition zone, it is determined that the liquid surface is crossing the depth of the component. The deviation is less than The candidate peak is temporarily exempted from the occlusion penalty in this interval. Locking during the crossing period is dominated by the continuity term and the prediction consistency term, ensuring that the actual liquid level is not mistakenly suppressed when passing the depth of the agitator and coil. Weighting coefficients to The initial value is set to 0.2, and the theoretical upper limit of the score is... Initial value The score is 0.6, which is the confidence threshold for the rating. For example, a threshold value of 0.35 is used for the difference. For example, a value of 0.08 is used; after commissioning, the weights and thresholds are adjusted based on on-site calibration data or historical event samples confirmed manually, using methods such as logistic regression. During the adjustment, the values ​​are updated simultaneously according to the above formula. and keep Less than Constraints, such as limitations No more than Eighty percent. Due to the existence of the two penalties, the scores of individual candidate peaks may be negative under extreme interference. Candidate peaks with scores not higher than zero are directly judged as non-liquid surface echoes and removed from the candidate set of the current frame. That is, non-positive scores are truncated to zero. The removed candidate peaks will no longer participate in the subsequent dominance comparison and liquid surface locking. Their peak shape parameters are still kept in the diagnostic group message for future reference.

[0039] After scoring is completed, the liquid level locking is performed according to the following rules: the highest score among the remaining candidate peaks in the current frame that is not lower than the minimum score is recorded. The winner is the peak When the winning peak is consecutive to the current locked position, that is... When the position does not exceed the consecutive position threshold, the threshold is, for example, taken as... The lock is updated directly using the winning peak; when the winning peak deviates from the lock position and the difference between its score and the second highest score is not less than [missing information], [the lock is updated]. At that time, the switching confirmation process begins, only if the peak is continuous. Switching is only performed when the frame remains dominant. For example, if we take 3, when switching execution, we calculate the liquid level change rate based on the difference between the old and new locked positions and the confirmation duration, and send it to the alarm judgment to avoid single-frame noise or occasional interference causing locking jumps; if the original locking peak is in the candidate set of the current frame and the score is not lower than 3 during the confirmation period, then the liquid level change rate is calculated and sent to the alarm judgment. If the original locked peak is not found, the output will still be treated as the liquid level peak; otherwise, it will be treated as if there is no reliable liquid level. If the score difference between the winning peak and the second highest peak is less than 1, the output will be processed normally. This indicates that the current frame's identification of the actual liquid surface is ambiguous. The echo analysis unit 102 does not switch its locking position, maintains the previous reliable result, marks the node status as suspicious, and requests diagnostic data. When the frame contains only a single candidate peak, it does not constitute ambiguity and does not fall into this branch. The suspicious state continues until the recapture start frame count is reached. hour, For example, if the value is 10, echo analysis unit 102 clears the sliding window and enters recapture mode: using capture scores without continuity and prediction constraints.

[0040] Peak selection is performed frame-by-frame across the entire measurement range, with weights... For example, take 0.6, and For example, each is set to 0.2, which is the capture threshold. For example, taking 0.15, continuously When the position difference between the selected capture peaks in a frame is less than the position continuity threshold, the frame is reconstructed and locked based on their positions. , The system re-accumulates and estimates from zero, then resumes the normal scoring process. Immediately after lock reconstruction, the liquid level change rate is calculated based on the difference between the old and new lock positions and the duration of the lock loss, and this is sent to the alarm for judgment. In cases where overflow, well leakage, or other situations cause the liquid surface movement speed to exceed the conventional tracking capability, the change rate alarm is triggered immediately after recapture, preventing prolonged delays due to track loss. Combined with encrypted observation under suspicious conditions, the time from lock loss to lock reconstruction can be controlled within tens of seconds. Upon initial power-on, power-off restart, or re-registration after relocation, there is no lock history. Similarly, an initial lock is established first through the recapture mode. Before lock establishment... and It is not involved in the calculation. This method of scoring first, maintaining when in doubt, and independently recapturing after loss of lock ensures that the system will not suddenly jump to the agitator echo or component echo when the stirring speed changes, the foam morphology changes, or the adhesive weights are added. It will also not miss rapid liquid surface movement, and at the same time significantly reduces the need to re-calibrate the empty tank after changes in operating conditions.

[0041] The auxiliary sensing unit 103 includes at least a temperature sensor for temperature compensation and steam space trend identification, and a triaxial vibration sensor for operating condition identification; as an optional configuration, it may also include an tilt sensor, a power health monitoring circuit, a process status input interface, and a measurement input channel. The temperature measurement point preferably covers both the RF cavity and the environment; if only a single point is used, it is located in the RF cavity. The data is used, on the one hand, to compensate for the thermal drift of the radar measurement unit 101 according to a pre-calibrated curve, and on the other hand, to identify the changing trend of the steam space inside the tank. The vibration sensing unit takes the root mean square of the triaxial acceleration signal after bandpass filtering in the tens to hundreds of hertz frequency band to obtain the vibration index. The process status input interface is used to input the operating signal of the stirring motor or the output signal of the frequency converter, as well as the operating status signals of the replenishment pump, discharge pump, and corresponding valves; the measurement input channel is a 4 to 20 mA current loop input or an RS485 interface, used to connect to comparison channels such as hydrostatic level gauges. Operating condition label. The following methods are used to generate the stirring gear: When there is a process status input, the stirring gear is directly given by the signal from the stirring motor or frequency converter, and the nozzle impact gear is directly determined by the signal from the feed pump valve. When there is no external signal, the stirring gear is inferred from the vibration index according to the preset gear threshold. The nozzle impact gear is further identified by combining the spectral characteristics of the vibration signal, such as distinguishing the broadband impact component from the narrow-band periodic component of stirring by the proportion of high-frequency band energy or kurtosis. Any gear switch must be confirmed for at least 3 seconds to eliminate instantaneous vibration. When an angle sensor is selected, the angle measurement is used to detect changes in the attitude of the skid-mounted tank after relocation or foundation settlement and to prompt verification of the installation reference height. When not selected, this verification is completed manually after relocation; the power health monitoring circuit tracks the changing trends of battery voltage and internal resistance, and provides a status data warning before the capacity is nearly depleted. It can be understood that the auxiliary sensing unit 103 does not replace radar in measuring liquid level, but rather provides bypass information required for operating condition understanding to the echo analysis unit 102 and the edge decision unit 105, and is an indispensable component of the terminal-side software closed loop.

[0042] The StarScan communication unit 104 adopts the StarScan short-range wireless communication module, operating under the low-power access technology of the StarScan protocol for IoT sensing scenarios. It is responsible for completing node registration, network-wide time synchronization, data transmission and reception according to the time slot table, emergency request transmission, transmission of buffer summary reports, link status reports, and acknowledgment retransmission. The StarScan air interface is centrally scheduled by the management node, which is a centralized, non-contention-based access: the StarScan gateway node 200 broadcasts beacon frames according to the scheduling cycle. The beacon frames carry the synchronization reference and the time slot table for the current cycle. Each tank-top intelligent radar monitoring terminal 100 uses this information to calibrate its local clock and complete transmission and reception within the authorized time slot, achieving a synchronization accuracy at the microsecond level. Following the beacon frame, an emergency request window is set up. This window consists of request micro-time slots corresponding one-to-one with each intelligent radar monitoring terminal 100 on the tank top. The mapping relationship is fixed during network registration. Each request micro-time slot carries only a very short request sequence. The intelligent radar monitoring terminal 100 on the tank top only sends a request within its assigned request micro-time slot when it has an alarm group message to be sent on its own device. At other times, it remains silent. The fixed mapping ensures that there are no collisions in the request process, and the total window overhead accounts for a negligible proportion of the scheduling cycle. Immediately following the emergency request window is an emergency indication micro-time slot broadcast by the StarScan gateway node 200, which is used to announce the ownership of each emergency sub-time slot in this cycle. The aforementioned emergency request micro-time slots, emergency indication micro-time slots, and the emergency time slots that can be conditionally borrowed as described later are all time slot usage agreements defined by the StarScan gateway node 200 in the scheduling information based on the StarScan centralized time slot orchestration mechanism, and do not involve any changes to the physical layer frame structure. Because the intelligent radar monitoring terminals 100 on each tank top do not collide due to random contention, the air interface delay of a single transmission can be controlled to the order of hundreds of microseconds. The physical layer uses polar code encoding, which has good resistance to strong electromagnetic interference sources such as well site motors and frequency converters. Unacknowledged messages are retransmitted according to the selective retransmission mechanism. The StarScan communication unit 104 switches between sleep and wake-up according to the listening cycle issued by the StarScan gateway node 200. The radio frequency circuit is only powered on in the beacon window, the corresponding request micro-time slot, the emergency indication micro-time slot, and the authorized time slot, with a very low duty cycle. Once the intelligent radar monitoring terminal 100 on the tank top generates an alarm group message while in sleep mode, it immediately wakes up on its own, calculates and synchronizes with the next beacon frame based on the local clock, and sends a request in the following emergency request window. Therefore, the sleep mechanism does not affect the timely reporting of alarms. The above operating mode, combined with battery power supply, keeps the overall power consumption at a very low level.

[0043] The edge decision unit 105 classifies the operating status of the tank-top intelligent radar monitoring terminal 100 into three levels: normal, suspicious, and alarm, based on the liquid level value, confidence level, liquid level change rate, fluctuation within the adjacent time window, and current operating condition label. It also classifies the data to be transmitted into three categories: alarm group, status group, and diagnostic group. The alarm group includes the tank number, timestamp, message sequence number, liquid level value, confidence level, alarm type code, and a summary of the liquid level change rate over a recent period. The status group includes the tank number, timestamp, message sequence number, node status, liquid level value, confidence level, temperature, vibration index, battery or power supply status, and link quality. The diagnostic group includes a candidate peak list, local echo feature vectors, operating condition label, false echo spectrum increments, and compressed echo summaries from the most recent frames. The echo summary compression method involves extracting the amplitude of 8 distance units before and after each candidate peak, quantizing it to 8 bits, and performing differential encoding. The single-frame summary can be controlled to be within tens of bytes. The false echo spectrum increment only includes the distance unit number and its new value whose amplitude change exceeds a preset threshold since the last transmission. Each uplink message header has a pending alarm indication bit and a transmission buffer summary field. The former indicates that there are alarm group messages that have not yet been sent on the local machine, and the latter records the number of pending messages in each group in the transmission buffer, the earliest generation time, and the total length, which are used by the StarScan gateway node 200 to calculate sorting parameters when scheduling time slots. Separating the measurement results, operating status, and interpretation evidence into different data groups is significant because in the normal stage, only a minimal amount of air interface resources are needed to transmit the feature data required for the interpretation conclusion. When a suspicious event or alarm occurs, the field control host 300 can obtain the original evidence to interpret the conclusion as needed, and the integrity of the evidence chain does not come at the expense of normal air interface overhead.

[0044] Within each sampling period, the edge decision unit 105 executes the following decision process. After the echo analysis unit 102 completes candidate peak extraction, peak-by-peak scoring, and non-positive score truncation and removal, if the candidate set for this frame is empty or the highest score of the remaining candidate peak is lower than the scoring confidence threshold... If no reliable liquid surface is found in this frame, the node status is set to suspicious, the liquid level output retains the reliable value from the previous moment, and the confidence level is set to a low level, for example, 0.2. The status group and diagnostic group are sent in this cycle; if the highest score is not lower than... However, the difference between it and the second-highest score is less than The node status is also set to suspicious, the liquid level remains unchanged, and the confidence level is set to medium (e.g., 0.4). The status group and diagnostic group are then sent; if the highest score is not lower than... If there is no ambiguity as described above, then the locking and switching rules shall apply. If the winning peak and the locked position are consecutive, or if the switching confirmation is completed, the distance corresponding to the winning peak shall be used. Solution of liquid level Confidence level according to

[0045] The calculation and limit are between 0 and 1. When the original locked peak is still reliable during the switching confirmation period, the liquid level and confidence level are calculated based on the current frame position and score of the original locked peak. The node status remains normal and a diagnostic group is attached to retain evidence of the switching process. When the original locked peak is no longer reliable, it is output as a suspicious case with no reliable liquid level. After obtaining the effective liquid level, the edge decision unit 105 further judges whether the liquid level exceeds the high or low limit and whether the absolute value of the liquid level change rate exceeds the process allowable range. If either is true, the node status is set to alarm and a high limit, low limit, or change rate alarm type code is written. The liquid level change rate calculated after recapture is also sent to this judgment. In parallel with the above branches, the edge decision unit 105 maintains a continuous suspicious cycle counter. It is reset to zero when a reliable lock is obtained. When the count reaches the preset upper limit, such as 30 consecutive cycles, regardless of which branch the current frame falls into, the node status is set to alarm and a continuous low confidence alarm type code is written. The liquid level field of the alarm group is filled with the current holding value and marked with a low confidence level. The anomaly of continuous loss of lock is thus not shelved for a long time. When a node is in an alarm state, it sends an alarm group, a status group, and a diagnostic group for the current cycle. When it is in a suspicious state, it sends a status group and a diagnostic group. In a normal state, it only sends a status group. Limit judgment introduces hysteresis. Exiting an alarm state requires the corresponding alarm condition to be cleared and the system to remain normal for several consecutive frames, such as 10 consecutive frames, to avoid repeated alarms caused by fluctuations in the liquid level near the limit. At the end of each cycle, if the stability window condition is met, the false echo spectrum of the current operating condition is updated using a forgetting method according to the aforementioned masking rules. The entire process is automatically executed on the terminal side using a fixed software pipeline, without manual intervention.

[0046] like Figure 1 As shown, the StarScan gateway node 200 is installed in the well site control room or in an explosion-proof cabinet as a centralized management and control node for the entire wireless domain. It includes a network clock and access management unit 201, a packet scheduling unit 202, a reassembly confirmation unit 203, and a host interface unit 204.

[0047] The network clock and access management unit 201 is responsible for network-wide time synchronization, node registration, address allocation, requesting micro-timeslot mapping allocation, issuing monitoring cycles, and statistical analysis of link quality. Link quality statistics cover received signal strength and packet error rate, which can be used to instruct the corresponding tank-top intelligent radar monitoring terminal 100 to adjust its transmission power or modulation / coding level. Considering the frequent relocation of drilling site equipment and frequent adjustments to tank configurations, the network clock and access management unit 201 issues static configuration information, including tank number and installation reference height, to each tank-top intelligent radar monitoring terminal 100 during initial network setup. The table includes: tank volume conversion table number, default sampling period, allowable range of high and low limits and liquid level change rate, occlusion prior labeling interval, scoring weights and decision thresholds, and the corresponding emergency request micro-slot mapping. The decision thresholds cover... , , Vibration ranges under various operating conditions , , With forgetting factor Meanwhile, the key weights of each tank are registered on the StarScan gateway node 200 side for scheduling purposes. When a smart radar monitoring terminal 100 on the top of a tank is re-powered after being moved with the tank, it only needs to be re-registered and receive a new configuration to start monitoring, without needing to reconfigure the entire network.

[0048] The group scheduling unit 202 employs different scheduling mechanisms for alarm group messages and other messages: alarm group messages do not participate in priority sorting and are always sent at the highest level via the emergency time slot mechanism described later; the priority of status group, diagnostic group, and low-priority retransmission messages after link recovery is calculated and sorted according to the following formula.

[0049] All terms in the formula are normalized: Event Level The value is assigned according to the message group: 1 for the status group, 0.5 for the diagnostic group, and 0.2 for low-priority retransmission messages. This is the ratio of the message's waiting time in the transmission buffer to the allowed waiting limit, and it is capped at 1. The waiting limit is, for example, 2 seconds. The critical weight of the tank to which the message belongs is given by static configuration between 0 and 1, with the value of the circulating tank participating in the cycle being higher than that of the storage tank and the mixing tank. Indicates whether the message carries a retransmission request, and is either 0 or 1; This is the ratio of the message length to the maximum transmission unit length of the air interface. Weighting coefficient. to All can be configured, and a set of feasible initial values ​​can be provided. Take 0.5, , , Take 0.15 from each. The value is set to 0.05. The required waiting time and message length are calculated from the transmit buffer digest reported by the intelligent radar monitoring terminal 100 on each tank top along with the uplink message header. During the period when no new digest is received, the time is extrapolated based on the most recent digest of that node and its message generation cycle. This priority sorting applies to the reallocation of surplus normal time slots, the transmission order within the elastic time slot pool, and the selection of emergency time slot borrowing objects. The basic transmission guarantee obtained by the state group message through normal time slot polling is independent of this sorting and is not affected by the sorting result.

[0050] The scheduling is executed cyclically with a fixed period, which in one embodiment ranges from 100 milliseconds to 500 milliseconds. The time slots in each scheduling period consist of: beacons, emergency request windows, emergency indications, and management overhead, which together account for about 5%; normal time slots, with a planned upper limit of about 60%, are actually arranged by the group scheduling unit 202 according to the sampling period and cache digest of each node, and allocated to status group messages in a polling manner supplemented by waiting time compensation. Unarranged normal time slots are immediately merged into the elastic time slot pool. The so-called elastic time slot pool capacity fluctuates with the remaining air interface resources, which comes from this recycling mechanism; emergency time slots account for 10% to 20%, which are divided into several emergency sub-time slots. Each emergency sub-time slot can carry one alarm group message, and the number of emergency sub-time slots is, for example, 4; the remaining time slots constitute the elastic time slot pool, which is used by diagnostic groups and low-priority retransmission messages. When there are no alarm group messages in the system, the emergency sub-time slot can be temporarily borrowed to the message with the highest ranking in the elastic time slot pool. The borrowing is granted by conditional authorization: the group scheduling unit 202 marks the emergency sub-time slot of this period as available for borrowing in the beacon and specifies the message to be borrowed. At the same time, it is agreed that the authorization will only take effect when the emergency request window of this period is silent. Once a new alarm group message is generated by the intelligent radar monitoring terminal 100 on the tank top, it sends a request within the corresponding request micro-slot of the most recent emergency request window. The group scheduling unit 202 directly identifies the alarm node based on the fixed mapping between the request micro-slot and the node, and then announces the final assignment of each emergency sub-slot in the current cycle in the emergency indication micro-slot. Nodes holding borrowing authorization relinquish their borrowing and sending upon receiving the indication. When multiple requests occur in the same window, if the number of concurrent requests does not exceed the number of emergency sub-slots, all requests are sent within the emergency sub-slots of the current scheduling cycle. Requests exceeding this number are assigned according to the tank's criticality weight and the order of request registration. Alarm nodes that are not assigned are carried over to the next scheduling cycle and are always placed at the head of the queue. The group scheduling unit 202 also temporarily compresses the normal time slots and the elastic time slot pool to expand the emergency sub-slots for the next cycle. The group scheduling unit 202 also continuously guarantees emergency sub-slots for nodes in the alarm state in the time slot table of subsequent scheduling cycles, and allocates dual-redundant time slots for them when necessary to further reduce the probability of packet loss. With the help of this mechanism that combines reservation, conditional borrowing and request assignment, the air interface bandwidth is not idle and wasted under normal circumstances. Under the premise that the message is successfully transmitted once, the delay from the generation of a single node alarm to the completion of the transmission does not exceed the sum of one scheduling cycle and a fixed processing offset. When the scheduling cycle is 500 milliseconds, this upper limit is within one second. When the number of concurrent alarms does not exceed the number of emergency sub-time slots, they share this upper limit. When they exceed it, the number of cycles is rounded up, which is still a calculable and definite value. Taking the aforementioned eight-tank deployment and four emergency sub-time slots as an example, even if all tanks alarm at the same time, the latest alarm group message will be sent within two scheduling cycles. Alarm group messages with transmission errors are resent through the selective retransmission mechanism, always occupying the emergency sub-time slot of the most recent cycle at the highest level.This fundamentally avoids the situation in traditional contention-based access networks where abnormal packets queue due to channel congestion. Taking the aforementioned eight-tank deployment as an example, the state group packet length is tens of bytes, and the effective payload of the entire network per second is less than one thousand bytes under normal conditions. The air interface occupancy rate is very low, thus the elastic time slot pool and emergency time slot have ample allocation margin.

[0051] The reassembly confirmation unit 203 performs deduplication, reassembly, timestamp correction, and packet missing labeling on the data from each tank-top intelligent radar monitoring terminal 100 on the StarScan gateway node 200 side. Deduplication is based on the message sequence number. Diagnostic group messages exceeding the single time slot carrying capacity are fragmented on the terminal side and reassembled sequentially here. The timestamps are uniformly converted to the network-wide synchronization time base. When an alarm group message is lost or the cyclic redundancy check fails, the reassembly confirmation unit 203 reports the need to retransmit the tag to the group scheduling unit 202, so that the message will receive higher priority in the next scheduling cycle. When it is found that a tank-top intelligent radar monitoring terminal 100 only sends status groups for several consecutive cycles and the confidence level is consistently lower than the preset level, for example, for 5 consecutive cycles, the reassembly confirmation unit 203 actively issues a diagnostic start command to the tank-top intelligent radar monitoring terminal 100, requiring it to send diagnostic group data in the next cycle. Accordingly, a bidirectional closed loop with network-side diagnostic triggering is formed between the tank-top intelligent radar monitoring terminal 100 and the StarScan gateway node 200, rather than a simple one-way upload relationship.

[0052] The upper-level interface unit 204 connects to the field control host 300 via Ethernet, serial port, or industrial bus. It outputs a unified data model using a register mapping method or publish-subscribe method commonly used in industrial settings. This model includes single-tank liquid level, single-tank confidence level, total volume of multiple tanks, anomaly level, and diagnostic evidence index. The diagnostic evidence index points to an archived list of candidate peaks and echo summaries, which the field control host 300 can retrieve as needed when reviewing alarms, avoiding the repeated transmission of original evidence in the regular data stream. As an optional implementation, the upper-level interface unit 204 can also report the summarized results to the remote monitoring center via the existing remote transmission link at the well site; this embodiment does not limit this approach.

[0053] The field control host 300 is implemented using a well site industrial control computer and monitoring software running on it, including a volume conversion unit 301, a multi-tank fusion discrimination unit 302, an alarm linkage unit 303, and a historical model management unit 304.

[0054] The volume conversion unit 301 converts the liquid level height according to the structural calibration table of each tank. Mapped to liquid volume For regular cubic or vertical cylindrical tanks, A linear function can be used; for tanks with sloping bottoms, compartments, or other irregular structures, Piecewise linear lookup table interpolation is used, with calibration points set along the liquid level direction, for example, one every 50 mm. In sections with drastic geometric changes, the density is increased to 20 mm. The calibration table can be calculated from the tank design drawings or obtained through water injection calibration. Converting the liquid level of each tank into volume is the basis for subsequent multi-tank fusion and judgment, because anomalies of concern to drilling technology, such as overflow, well leakage, and fluid loss, all macroscopically manifest as abnormal changes in the total fluid volume at the well site.

[0055] The multi-tank fusion discrimination unit 302 does not perform isolated limit judgments on the liquid level of a single tank, but instead constructs a joint discrimination index based on the total volume and confidence level. Assume the well site has... The monitored tanks have a total liquid volume of [number missing].

[0056] The discrimination is performed discretely according to a preset calculation cycle, for example, 5 seconds. The liquid level, confidence level, and node status of each tank are reported via status group messages according to their respective sampling cycles. The reconstructing confirmation unit 203 has uniformly converted their timestamps to the network-wide synchronization time base. The multi-tank fusion discrimination unit 302 takes the most recent valid reported value of each tank at each calculation cycle and performs hold-up interpolation to obtain equally spaced synchronization sequences for subsequent fitting. The sampling cycles of the intelligent radar monitoring terminals 100 on the top of each tank should be uniformly configured. If the configurations are inconsistent, the largest sampling cycle is used to verify the fitting window, ensuring that the window length is not less than ten times that cycle. Within each calculation cycle, the multi-tank fusion discrimination unit 302 performs a sliding window least squares fitting using the most recent total liquid volume sequence, for example, 60 seconds, to obtain an estimate of the total liquid volume change rate. The selection of the fitting window takes into account both the suppression of short-term fluctuations caused by stirring disturbances; reference volume The estimated rate of change is calculated based on the operating status of the replenishment pump, pump flush count, inlet and outlet flow meters, or manually entered operating condition markings. Obtained using the same fitting method, when the well site does not incorporate the aforementioned reference information. Zeroing is used. For tanks in a suspicious state, the reported liquid level is a held value and cannot reflect the true change. The multi-tank fusion discrimination unit 302 extrapolates the volume of the tank based on the most recent reliable rate of change from the moment it enters the suspicious state, and participates in the summation. The extrapolation stops after exceeding a preset time, for example, 5 calculation cycles. After that, the tank is removed from the fitted sequence and simultaneously removed within the window to maintain consistency. The abnormal status of the tank is then carried by the confidence term and the suspicious count term in the index. Based on this, a fusion anomaly index is defined.

[0057] In the formula, For a pre-configured normalized reference flow rate, such as the rated displacement of the well site drilling pump or the maximum allowable make-up flow rate, the first term is... After normalization and amplitude limiting, it becomes a dimensionless quantity between 0 and 1. When no reference information is connected, it degenerates into a normalized monitoring of the rate of change of the total liquid volume itself. The second term is the average confidence loss normalized by the number of tanks, and the third term is the proportion of suspicious nodes. The three terms are thus unified to the same quantity scale of 0 to 1, and the setting of the weight coefficient is no longer affected by volume, time unit and number of tanks. For the first The confidence level reported by the intelligent radar monitoring terminal on the top of each tank (100) This represents the number of 100 intelligent radar monitoring terminals on the tank tops currently in a suspicious state. Suspicious nodes are counted both in the second item (due to their relatively low confidence level) and in the third item (based on their count), a deliberate design choice: the mean confidence level is insensitive to the suspicion of individual nodes when the number of tanks is large, while the percentage item ensures that the index still has a perceptible response when a few nodes become suspicious, thus allowing the output information carried by the suspicious state to be expressed independently. Weighting coefficients. , , A feasible initial set of values ​​is 0.5, 0.25, and 0.25. Under normal operating conditions, the confidence level will not always be at its maximum; the second term has an inherent background. Therefore, the threshold is adjusted according to the operating baseline: after commissioning, statistical analysis is performed using a period of operating data confirmed to be free of any anomalies. The baseline mean and fluctuation range are calculated, with the first and second thresholds taken as the baseline mean plus a multiple of the fluctuation range, for example, three and six times the standard deviation, respectively. The baseline is recalculated as environmental segments change. A first-level warning is triggered when the index exceeds the first-level threshold for a preset number of calculation cycles, such as 3. A second-level alarm is triggered when the index exceeds the second-level threshold. A hysteresis is introduced into the threshold judgment, with the hysteresis magnitude set to, for example, 10% of the difference between the corresponding threshold and the baseline mean, to prevent the index from repeatedly crossing the threshold and causing alarm oscillation. In addition to index judgment, the multi-tank fusion judgment unit 302 also has a direct criterion for single-tank alarms: when any intelligent radar monitoring terminal 100 on the top of a tank reports a high limit, low limit, or rate of change alarm, regardless of... Regardless of the value, at least a Level 1 alarm must be triggered and the diagnostic data for that tank must be activated. If the alarmed tank is a circulating tank or two or more tanks are simultaneously alarming, a Level 2 alarm will be triggered directly. According to this judgment logic, if only a single tank shows a low confidence level while the total liquid volume is not abnormal, the field control host 300 will prioritize issuing a single-point suspicion and requesting diagnosis prompt rather than a process alarm. Conversely, when the total liquid volume changes of multiple tanks are significantly inconsistent with the reference operating conditions, and multiple intelligent radar monitoring terminals 100 on the top of the tanks are simultaneously in a suspicious or alarm state, the alarm level will escalate step by step. The occasional echo interference from a single measuring point is thus limited to the diagnostic level and will not be directly amplified into a global operating condition alarm. This is consistent with the actual needs of drilling sites, which require early detection of overflows and well leaks but cannot tolerate a large number of false alarms.

[0058] The alarm linkage unit 303 executes graded actions based on the output of the multi-tank fusion discrimination unit 302. In the case of a Level 1 warning, only a prompt is given on the human-machine interface, and diagnostic group data is activated to the relevant tank-top intelligent radar monitoring terminal 100 via the star-flash gateway node 200. In the case of a Level 2 alarm, the on-site audible and visual alarm is activated, event window data is recorded for approximately 10 minutes before and after the alarm time, and the alarm signal is output to the programmable logic controller or distributed control system at the well site in the form of dry contacts or communication messages. When a Level 2 alarm is further triggered, and the conditions for a single tank exceeding its limit are met, the alarm linkage unit 303 will perform graded actions. When the rate of change of the normalized total liquid volume continuously exceeds the preset severe deviation threshold and other superimposed conditions, a Level 3 severe alarm is detected. At this time, the alarm linkage unit 303 may optionally output a pump valve interlock suggestion signal or a control signal awaiting manual confirmation. It should be noted that the protection subject of this system is a radar level monitoring system, and the above-mentioned interlock controls are all optional external linkages. Whether to execute them is determined by the on-site process side and does not change the essential attribute of the system as a monitoring system.

[0059] The historical model management unit 304 archives the false echo spectrum versions, threshold parameters, and event samples of each tank-top intelligent radar monitoring terminal 100 according to tank number, operating condition label, and environmental segment. The environmental segments are divided by season or by environmental temperature in 15-degree Celsius increments. The daily evolution of the spectrum is completed autonomously by the tank-top intelligent radar monitoring terminal 100 according to stable window conditions, without the need for sequential permission from the host. The historical model management unit 304 is responsible for post-event supervision and parameter management: each tank-top intelligent radar monitoring terminal 100 sends a full snapshot of the current spectrum for the four operating conditions through the diagnostic group message at a fixed period, such as every 24 hours, or when the cumulative change of the spectrum exceeds the preset range. The historical model management unit 304 registers each snapshot along with its time period's operating condition label and environmental segment as a spectrum version. The spectrum increments in the diagnostic group are only used for short-term evidence collection, and version management is based on the full snapshot. During the observation period after the new version is registered, for example, within 24 hours, if the average confidence level of the tank drops by more than a preset range compared to the previous version period, the historical model management unit 304 will send the previous version's map back to the intelligent radar monitoring terminal 100 on the tank top via the StarSpark gateway node 200 for restoration, and temporarily tighten its update parameters, such as raising the lower limit of the forgetting factor or pausing map updates for a certain period of time to prevent abnormal operating conditions from being mislearned. The environmental segment dimension only exists at the archiving and tuning level on the host side. The intelligent radar monitoring terminal 100 on the tank top always maintains only the current maps for four operating conditions. When the season or temperature segment changes, the historical model management unit 304 will send back the map version and threshold parameters that match the new segment from the archive to achieve experience reuse across segments. Once a false alarm or missed alarm event is confirmed by manual verification, the historical model management unit 304 will review the data based on the archived candidate peak sequence, confidence curve, and time slot allocation record, and set the scoring confidence threshold. Difference threshold The proposed adjustments to the scoring weights are then written back to the corresponding intelligent radar monitoring terminal 100 on the tank top via configuration commands from the StarSpark gateway node 200. This establishes a two-layer model structure: edge adaptation on the terminal side and historical governance on the host side. The former ensures real-time performance, while the latter guarantees long-term consistency.

[0060] The system commissioning process is as follows. During the pre-installation survey phase, the opening locations of each tank are determined, and the distance ranges between the agitator, heating coil, ladder, and fixed components such as internal welds and other fixed components and the antenna aperture are recorded. After installation, the intelligent radar monitoring terminal 100 on the tank top is powered on and registers with the star-flash gateway node 200 via the star-flash communication unit 104, obtaining all static configurations, including the priori marking range for obstruction, and the associated emergency request micro-time slot mapping. Subsequently, the initial mapping phase begins, with each of the four operating conditions accumulating, for example, over two hundred frames of echoes. The initial base maps for the three modes—stop agitation, low-speed agitation, and high-speed agitation—are acquired under conditions of empty tanks or known and stable liquid levels. The nozzle impact mode is inevitably accompanied by feeding, and its initial base map is acquired under controlled feeding conditions. The initial liquid level is provided by manual measurement or a comparison channel accessed via the measurement input channel. During acquisition, the real-time liquid level is calculated based on the feeding volume, and the protection range rotates with the calculated liquid level. During mapping, the distance corresponding to the actual reflective surface is determined first: under empty tank conditions, the actual reflective surface is the tank bottom, and the corresponding distance is the installation reference height. Under liquid conditions, the recorded liquid level is used for... The result is obtained through reverse calculation. A protection zone is set with this distance as the center. During the mapping period, the protection zone is taken as 0.2 meters on each side, wider than the four distance units on each side during operation. This is because the position of the reflector during the mapping period is given by manual measurement or comparison channels, and its positioning accuracy is lower than the interpolation positioning of the locked peak during operation. The protection zone is widened accordingly to cover the input deviation. Only the echo amplitude between the antenna aperture and the near end of the protection zone is recorded in the initial base map for the corresponding operating condition. Distance units within the protection zone and on the side farther away are not written in. The base map amplitude of the uncovered sections is set to zero by default, and candidate peaks falling within these sections do not generate false echo penalties. With this shielding, the real liquid surface echo will not be mistakenly learned as a false echo, and when the liquid level falls back to near the mapping level, it will not be incorrectly suppressed due to base map overlap. Deeper sections not covered by the base map are gradually learned and filled in as the liquid level decreases during subsequent operation, the corresponding sections are exposed to the steam space, and the stable window conditions are met. The initial base map only needs to be collected once during commissioning, and then evolves automatically through forgetting updates within a stable window. This differs from the traditional radar level gauge method, which requires manual relearning of the empty tank spectrum every time operating conditions change. After the map is completed, the intelligent radar monitoring terminal 100 on the top of each tank establishes an initial liquid level lock in recapture mode, and then switches to normal sampling mode.

[0061] During the routine monitoring phase, radar measurement unit 101 periodically transmits frequency-modulated continuous waves and receives echoes. Echo analysis unit 102 locks the liquid level according to the aforementioned scoring mechanism. When the liquid level is stable and the confidence level is high, edge decision unit 105 only sends status group messages periodically. At this time, the overall network air interface occupancy and terminal power consumption are maintained at very low levels. Power consumption control is achieved through time-sharing power supply: the RF front-end is powered on only for a period of approximately 16 milliseconds during 16 frequency sweeps per frame. Signal processing is completed by a low-power microcontroller using fixed-point arithmetic. The total wake-up time, including the satellite flash transceiver, is tens of milliseconds. The entire unit is in microampere-level sleep mode for the rest of the time. Based on a 5-second sampling period, the average operating current can be controlled to within 1 mA. Combined with the aforementioned capacity lithium battery, it can support more than two years of maintenance-free operation. This conclusion is based on the premise of normal sampling configuration. Encrypted observation with a shortened sampling period under suspicious or alarm conditions is a short-term event-triggered behavior with limited impact on overall battery life. When the liquid level in a tank fluctuates abnormally or the confidence level decreases but has not yet exceeded the limit, the corresponding intelligent radar monitoring terminal 100 on the tank top enters a suspicious state and sends a diagnostic group message in addition to the status group. The StarGateway node 200 arranges it into the elastic time slot pool. When the liquid level exceeds the limit, the rate of change is abnormal, or the continuous low confidence condition is met, the intelligent radar monitoring terminal 100 on the tank top enters an alarm state. The terminal that is in sleep mode is woken up first and synchronizes with the next beacon. It sends a request in the corresponding request micro-time slot of the emergency request window that follows. Even if the emergency sub-time slot has been conditionally authorized to lend to the message in the elastic time slot pool, the group scheduling unit 202 also announces the time slot transfer in the emergency indication micro-time slot. The lending node gives up sending, and the alarm group message is sent in the emergency sub-time slot of this scheduling cycle, ensuring that the alarm information arrives at the field control host 300 with an approximately certain time delay.

[0062] After receiving the data, the on-site control host 300 converts the liquid level of each tank into volume using the volume conversion unit 301, and the multi-tank fusion discrimination unit 302 calculates the total liquid volume change and fusion anomaly index. The alarm linkage unit 303 executes a graded action based on the direct-access criterion, and the historical model management unit 304 simultaneously archives the original candidate peaks, confidence sequences, time slot allocation records, and total volume curves before and after the alarm for post-event review, parameter optimization, and implementation verification. Taking the overflow condition as an example, after the formation fluid invades the wellbore, the return flow rate is greater than the pump inflow flow rate, and the liquid level in the circulation tanks participating in the circulation continues to rise at an abnormal rate. The relevant intelligent radar monitoring terminal 100 on the tank top detects the rate of change exceeding the limit and switches to an alarm state. After a request is sent through the emergency request window, the alarm group message is sent up in the emergency sub-time slot of the same scheduling cycle. When multiple circulation tanks alarm simultaneously, they are sent up successively in the adjacent scheduling cycle according to concurrent arbitration. At the same time, the rate of change of the total liquid volume of multiple tanks deviates significantly from the reference rate calculated by the pump flush count, and the fusion anomaly index is established. If the level exceeds the secondary threshold and remains there, the direct-access criterion is also established due to the circulation tank exceeding the limit alarm. The alarm linkage unit 303 triggers an audible and visual alarm, archives the event window, and outputs an alarm signal to the distributed control system. On-duty personnel then take timely action based on this information. Whether an alarm is established is no longer directly determined by the instantaneous exceeding of the limit at a single liquid level measuring point, but rather by a combination of four dimensions: liquid level value, confidence level, total liquid volume, and operating condition reference. This is a key operational feature that distinguishes this embodiment from conventional liquid level telemetry schemes.

[0063] Let's take the interference scenario of switching operating conditions as an example. After a certain agitator switches from low speed to high speed, the blade echo is significantly enhanced and a new secondary peak appears in the range spectrum. In some frames, the score of this secondary peak is close to the liquid surface peak, and the difference between the two is less than 1. In these frames, the echo analysis unit 102 does not switch the locked position but maintains the previous reliable liquid level. The node briefly enters a suspicious state and sends a diagnostic group message. The StarScan gateway node 200 arranges it into the elastic time slot pool. In most other frames, the liquid surface peak still has a significant advantage due to its continuity, prediction consistency, and the occlusion penalty of the secondary peak, and the reliable locking is maintained. On the field control host 300 side, due to the phased decrease in the confidence of the tank, the corresponding component is recorded in the fusion anomaly index. However, the total liquid volume is stable and there are no other tank anomalies. Therefore, only a single point suspicious prompt is given and the diagnostic data is confirmed to be up, without generating a process alarm. Under high-speed stirring conditions, the vibration index falls into the preset normal range of this level. As the continuous reliable locking is restored, the stable window is established one after another. The false echo spectrum of this level gradually converges in the section above the liquid surface through forgetting update. The false echo overlap term at the secondary peak position increases and its score decreases. Ambiguous frames no longer appear, and the node stabilizes in the normal state. No manual adjustment is required throughout the process. If the review reveals that the tank repeatedly enters a suspicious state during similar switching, the historical model management unit 304 can correct the tank's scoring weight or map update parameters based on the archived samples and write them back to the corresponding intelligent radar monitoring terminal 100 on the tank top.

[0064] Based on the above embodiments, there are several alternative implementation methods. The operating frequency band of the radar measurement unit 101 is not limited to 80GHz; it can also use other millimeter-wave frequency bands such as 60GHz or 120GHz. The antenna type can be selected between a lens antenna and a horn antenna, as long as the beamwidth is compatible with the tank geometry. The number of levels for the operating condition label can be expanded from four levels. For example, special operating conditions can be added for special operation stages such as cementing and tripping, along with independent false echo base maps. For the most critical circulating tank, a hydrostatic level gauge can be installed in parallel at the bottom of the tank as a comparison channel. It is connected to the measurement input channel of the auxiliary sensing unit 103, and its data is reported along with the status group. The field control host 300 performs dual-channel consistency verification. The readings of this comparison channel can also serve as a source of known liquid levels during the initial mapping stage. The existence of the comparison channel does not change the system configuration that is mainly based on radar measurement. For well sites with high continuous monitoring requirements, two StarScan gateway nodes 200 can be configured for mutual hot standby. Both share node registration information and request micro-timeslot mapping. If the primary StarScan gateway node 200 fails, the backup StarScan gateway node 200 takes over the scheduling. For well sites with a large number of tanks or a large distribution area, multiple StarScan gateway nodes 200 can be set up in a domain-specific network, each managing the intelligent radar monitoring terminals 100 on the tank top within its jurisdiction. Beacon transmission and timeslot arrangement in each domain are uniformly coordinated by the field control host 300, using different working channels or staggered superframe time offsets to avoid inter-domain interference. This ensures that collision-free access and emergency timeslot mechanisms within each domain are unaffected, and the data required for judgment is uniformly aggregated to the field control host 300. The intelligent radar monitoring terminal 100 on the tank top can have a ring buffer. Status data during link interruptions is cached periodically and sent back using the aforementioned low-priority retransmission method after link recovery, ensuring the integrity of historical data without crowding out real-time message air interface resources. In addition to alarm groups, status groups, and diagnostic groups, a configuration management group can be added for parameter distribution and firmware upgrades. This group only occupies the idle portion of the normal time slots and does not affect the aforementioned scheduling mechanism. The functional division within the field control host 300 is also not limited to the aforementioned forms. The volume conversion function can be integrated into the multi-tank fusion discrimination unit 302. The merging or splitting of functional units does not change the aforementioned discrimination logic and the system's protection theme. All the above variations are based on the technical concept defined by two core logics: dynamic false echo patterns with operating condition layering and scoring peak selection with confidence output, and deterministic transmission scheduling of packets prioritizing anomalies.

[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art should understand that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A radar level monitoring system based on intelligent sensors and data transmission, characterized in that, The system includes several intelligent radar monitoring terminals installed on the top of the monitored tank, a StarScan gateway node wirelessly connected to the terminals, and a field control host connected to the StarScan gateway node. The intelligent radar monitoring terminal on the tank top includes: The radar measurement unit is used to periodically transmit frequency-modulated continuous wave signals and receive echoes to generate a range spectrum containing several candidate peaks. The auxiliary sensing unit is used to sense and acquire the current operating status label of the monitored tank; The echo analysis unit is used to independently maintain the corresponding dynamic pseudo-echo spectrum according to the operating condition label. When the stable window condition is met, it performs a forgetting update on the dynamic pseudo-echo spectrum for the amplitudes within the update upper bound range of the current frame distance spectrum. The update upper bound is determined based on the smaller distance between the current locked liquid surface peak position and the candidate peak position with the highest signal-to-noise ratio in the current frame. A protection interval is reserved on both sides of the corresponding peak position without performing an update. The unit calculates the true liquid surface score of the candidate peak based on the signal-to-noise ratio, locking continuity, prediction consistency, pseudo-echo overlap with the dynamic pseudo-echo spectrum, and the occlusion prior determined according to the fixed component position. The unit locks the liquid surface based on the highest score and the score difference, and outputs the liquid surface measurement result including the confidence level calculated based on the highest score. The edge decision unit is used to classify the operating status of the terminal into normal, suspicious or alarm status based on the liquid level measurement results, and to classify the data to be sent into at least a status group containing the confidence level and an alarm group generated in the alarm status. The StarFlash communication unit is used to send the alarm group within a reserved emergency time slot by sending a request with an alarm indication to the StarFlash gateway node. The StarFlash gateway node includes a group scheduling unit, which is used to broadcast a time slot table containing emergency request windows and emergency sub-time slots at fixed intervals. When the request is received, the emergency sub-time slot is assigned to the corresponding alarm terminal to send alarm group data. Under normal circumstances, the remaining time slots are allocated to the remaining data groups. The field control host includes a multi-tank fusion discrimination unit, which is used to convert the liquid level of each monitored tank into volume, calculate the fusion anomaly index based on the estimation deviation of the total liquid volume change rate of the multi-tank tanks, the average confidence loss normalized according to the number of monitored tanks, and the proportion of terminal nodes in a suspicious state. When the fusion anomaly index exceeds a preset threshold, an alarm linkage is triggered.

2. The radar level monitoring system based on intelligent sensors and data transmission according to claim 1, characterized in that, The specific rules for updating the dynamic pseudo-echo spectrum by the echo analysis unit when the stable window condition is met are as follows: The upper bound for updating is the smaller of the current locked liquid surface peak position and the candidate peak position with the highest signal-to-noise ratio in the current frame. The protection interval is formed by extending a preset number of distance units to both sides of the corresponding peak position. The distance units within the upper bound of the update that do not belong to the protected areas are updated by weighting the amplitude of the current frame distance spectrum with the amplitude of the original spectrum according to the exponential forgetting factor. The distance cells within the protection zone and beyond the updated upper boundary maintain the original dynamic pseudo-echo spectrum amplitude unchanged.

3. The radar level monitoring system based on intelligent sensors and data transmission according to claim 1, characterized in that, When the echo analysis unit calculates the true liquid level score: The continuity score is calculated by the difference between the candidate peak position and the locked liquid surface position at the previous moment. The estimated rate of change of liquid surface distance is extracted by a reliable locking frame within a sliding time window to predict the current short-term liquid surface distance. The prediction consistency score is calculated based on the difference between the candidate peak position and the short-term liquid surface distance. The false echo overlap penalty term is calculated based on the amplitude of the preset neighborhood of the candidate peak in the corresponding dynamic false echo spectrum; The occlusion penalty is calculated based on whether the candidate peak falls within the pre-marked distance range of the fixed components inside the tank and the transition zones on both sides. When the short-term liquid level distance falls within a certain distance range of a fixed component or the transition zone, the corresponding occlusion penalty for the candidate peak whose deviation from the short-term liquid level distance is less than a preset scale parameter is temporarily exempted.

4. The radar level monitoring system based on intelligent sensors and data transmission according to claim 3, characterized in that, The liquid level locking logic of the echo analysis unit includes: Candidate peaks with scores not higher than zero are removed from the current frame, and the winning peak with the highest score and not lower than the score confidence threshold is selected from the remaining candidate peaks. When the winning peak deviates from the current locked position, and the difference between the winning peak score and the second highest score is not less than the score difference threshold, the switching confirmation process is initiated. After maintaining dominance for several consecutive frames, the locked position is switched. When the difference between the winning peak score and the second highest score is less than the difference threshold, the previous trusted locking position remains unchanged, the operating status of the corresponding terminal is marked as suspicious, and a request to send diagnostic group data is triggered. When the number of frames continuously in a suspicious state reaches a preset number, the sliding time window is cleared and the mode is switched to recapture mode. Within the full range, the peak value is reselected based on the capture score that does not include the continuity item and the prediction consistency item to establish a lock. After the lock is re-established, the liquid level change rate during the lock loss period is calculated and sent to the alarm judgment.

5. The radar level monitoring system based on intelligent sensors and data transmission according to claim 1, characterized in that, The auxiliary sensing unit includes a triaxial vibration sensor; the stability window condition requires that the following criteria be met simultaneously: The terminal continuously obtains a reliable liquid level lock within a preset number of sampling periods, and the liquid level change rate calculated based on the reliable lock sequence is lower than a preset change rate threshold. The vibration index obtained by the triaxial vibration sensor falls within the normal vibration range corresponding to the current working condition label; wherein, the upper limit of the normal vibration range corresponding to different working condition labels is set in stages according to the typical operating vibration level of the working condition and the impact transient level. There is no liquid replenishment or drainage action; among them, for the nozzle impact condition accompanied by feeding, the individual update conditions are constant feeding state, reliable locking is obtained in each frame and the liquid level change rate matches the feeding conversion rate.

6. The radar level monitoring system based on intelligent sensors and data transmission according to claim 1, characterized in that, The edge decision unit divides the data to be sent into the alarm group, the status group, and the diagnostic group; The alarm group includes a timestamp, liquid level value, confidence level, alarm type code, and liquid level change rate summary. The alarm group is generated only when the terminal liquid level exceeds the limit, the liquid level change rate exceeds the standard, or the number of consecutive suspicious periods reaches a preset upper limit. The status group includes node status, liquid level value, confidence level, temperature, vibration index, and link quality, and is sent during the normal sampling period. The diagnostic group includes a candidate peak list, local echo feature vectors, false echo spectrum increments, and echo summaries compressed by differential coding. It is generated and uploaded when the terminal is in a suspicious or alarm state for traceability and evidence collection.

7. The radar level monitoring system based on intelligent sensors and data transmission according to claim 6, characterized in that, The group scheduling unit of the StarGateway node executes an emergency time slot scheduling mechanism that combines reservation and borrowing: Each scheduling cycle includes a beacon, an emergency request window, an emergency indication micro-slot, a normal time slot, and the emergency sub-slot. The emergency request window consists of request micro-slots that are fixedly mapped and allocated to each terminal. When there are no alarm group messages on the entire network, the temporary conditional authorization of the emergency sub-slot is lent to the other messages with the highest priority in the system, and it is agreed that the authorization will only take effect when the current emergency request window is silent. When any terminal generates an alarm group message and sends a request in its corresponding request micro-slot, the StarFlash gateway node declares that the emergency sub-slot belongs to the alarm terminal in the immediately following emergency indication micro-slot. Terminals with conditional authorization immediately give up borrowing the transmission, thereby ensuring that the alarm group message obtains a deterministic transmission delay.

8. The radar level monitoring system based on intelligent sensors and data transmission according to claim 7, characterized in that, The remaining time slots not scheduled by the normal time slot polling are immediately incorporated into the elastic time slot pool on the network side; for the diagnostic group messages and low-priority retransmission messages after link recovery, the packet scheduling unit determines the transmission order in the elastic time slot pool based on the comprehensive priority sorting. The overall priority is calculated by normalizing and weighting the event level corresponding to the data group to which the message belongs, the waiting time of the message in the sending buffer, the criticality weight of the monitored tank to which the message belongs, the retransmission requirement flag, and the message length.

9. The radar level monitoring system based on intelligent sensors and data transmission according to claim 1, characterized in that, The calculation and linkage logic of the fusion anomaly index by the multi-tank fusion discrimination unit is as follows: Within the calculation period, the effective reported volume of each monitored tank is fitted using a sliding window least squares method to obtain the estimated value of the current total liquid volume change rate of multiple tanks; the reference change rate of total liquid volume calculated from the operating condition reference information is obtained, and the first term is obtained by normalizing the absolute difference between the two with the reference flow rate and limiting the amplitude. The second term is obtained by acquiring the average confidence loss of all currently monitored tanks. The third item is obtained by calculating the proportion of terminal nodes currently in a suspicious state out of the total number of monitored nodes in the system; The first, second, and third items are weighted and summed to output the fusion anomaly index; For terminal nodes in a suspicious state, the volume is extrapolated using the most recent reliable rate of change before they entered the suspicious state and included in the total volume. When the extrapolation continues for more than a preset time, it is removed from the self-fitted sequence and the summation of multiple tanks. When any terminal directly reports a single-point over-limit or rate of change alarm, and the monitored tank is a circulating tank participating in drilling fluid circulation or multiple tanks are alarmed concurrently, a level two alarm is directly triggered and the corresponding diagnostic data is activated.

10. The radar level monitoring system based on intelligent sensors and data transmission according to claim 6, characterized in that, The field control host also includes a historical model management unit, which is used to archive the dynamic false echo spectrum version, threshold parameters and event samples of each terminal according to tank number, working condition label and environmental section. When the terminal changes by a fixed period or the cumulative change in the spectrum exceeds a preset range, the diagnostic team sends a full snapshot of the current dynamic false echo spectrum. During the observation period after the new version snapshot is registered, if the average confidence level of the corresponding monitored tank decreases by more than a preset range compared to the previous version period, the historical model management unit sends the dynamic false echo spectrum of the previous version to the corresponding terminal through the starlight gateway node for full coverage restoration, and temporarily tightens the spectrum update parameters of the terminal.