Intelligent early warning method and system for production state of carbon black reaction furnace
By performing piecewise linear interpolation and sliding cross-correlation on the temperature, flow rate, and pressure data of the carbon black reactor, the problems of misjudgment and missed reporting in the existing technology are solved, and high-precision early warning of early faults in the carbon black reactor is realized.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GONGYI XINKE REFRACTORIES CO LTD
- Filing Date
- 2026-04-09
- Publication Date
- 2026-06-16
AI Technical Summary
Existing technologies have problems with misjudgment and missed detection in carbon black reactor fault detection, and cannot effectively identify early faults, especially because they ignore the complex operating conditions caused by multi-parameter timing delays and thermal inertia.
By generating an aligned sequence through piecewise linear interpolation of furnace temperature, feed flow rate, and furnace pressure data, the shape drift rate and relative impedance are calculated. Combined with the sliding cross-correlation algorithm, early warning is provided to compensate for fluid transmission delay and thermal diffusion interference, thereby improving the accuracy of early fault warning.
It improves the accuracy and reliability of early airflow blockage warning in carbon black reactors, reduces the frequency of false alarms, and enhances the ability to detect local anomalies.
Smart Images
Figure CN122223922A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of monitoring and early warning technology. More specifically, this invention relates to an intelligent early warning method and system for the production status of a carbon black reactor. Background Technology
[0002] In carbon black industrial production, the stable operation of the carbon black reactor is crucial to ensuring production line continuity and equipment safety. This process requires real-time monitoring of key process parameters such as furnace temperature, feed flow rate, and furnace pressure. By identifying early signs of malfunction such as localized coking, airflow blockage, or abnormal air supply, timely and accurate alarm commands can be issued to guide manual intervention, thereby preventing physical damage to the reactor or production losses due to unplanned shutdowns.
[0003] In related technologies, for example, Chinese patent application document with publication number CN102568148A discloses an early warning method and early warning system, including: collecting current monitoring values; generating change values by comparing the current monitoring values with recorded historical monitoring values; and comparing the change values with preset threshold values according to an early warning strategy to generate an early warning signal.
[0004] In related technologies, routine anomaly warnings are mainly based on the difference comparison of a single parameter. However, if this is directly applied to the fault detection of carbon black reactors, the characteristics of multi-parameter time delays and thermal inertia of the reactor are ignored. During the operation of carbon black reactors, due to the influence of fluid transmission mechanisms and internal heat diffusion, the process parameters in the reactor may sometimes exhibit complex conditions such as time differences between flow rate and pressure, and local temperature drops being smoothed out by the whole process. This makes it easy to misjudge normal flow adjustment lag as pipeline blockage or miss the real deterioration trend when making routine judgments.
[0005] Furthermore, the operating conditions of reactors with hidden faults have similar parameter deviation characteristics to those of reactors with normal fluctuations, making it impossible to effectively identify sudden abnormalities in pipelines and slow temperature deviations. It is difficult to predict early faults in reactors by comparing the differences of a single parameter. Therefore, there are certain limitations to using existing technologies for fault prediction in carbon black reactors. Summary of the Invention
[0006] To address the technical problems of false alarms caused by data deviation and lag, missed fault reports due to heat masking, and response delays and hierarchical deviations caused by the rigidity of early warning mechanisms, this invention provides solutions in the following aspects.
[0007] In a first aspect, the present invention provides an intelligent early warning method for the production status of a carbon black reactor, comprising: performing piecewise linear interpolation on furnace temperature data, feed flow rate data, and furnace pressure data to obtain an aligned temperature sequence, an aligned flow rate sequence, and an aligned pressure sequence; using the median of the aligned temperature sequence within a reference time window as a temperature reference; extracting the values at the 90th percentile, 50th percentile, and 10th percentile of the aligned temperature sequence sorted from smallest to largest within a preset statistical window, and obtaining the morphological drift rate in conjunction with the temperature reference; obtaining the dynamic window length based on the morphological drift rate; and obtaining the alignment flow rate sequence based on the morphological drift rate. The standard deviations of the flow rate sequence and the aligned pressure sequence within the reference time window are used to obtain the optimal number of delay points, combined with the excitation threshold. The difference between the sampled values of the aligned flow rate sequence at the current time and the starting time of the dynamic window is calculated to obtain the flow rate difference value. The aligned pressure sequence is shifted forward along the time axis by the optimal number of delay points to obtain the time-shifted pressure sequence, and the difference between its sampled values at the current time and the starting time of the dynamic window is calculated to obtain the delay pressure difference value. The relative impedance is obtained based on the standard deviation of the flow rate of the aligned flow rate sequence, the standard deviation of the pressure of the aligned pressure sequence, the flow rate difference value, and the delay pressure difference value, and an early warning command is output in combination with the impedance alarm threshold.
[0008] This invention generates aligned temperature, flow, and pressure sequences through piecewise linear interpolation, measuring the evolution benchmark of multi-source data over a unified time dimension and improving the data loss phenomenon at time boundaries. By combining the percentile values of the temperature sequences with the temperature benchmark to calculate the morphological drift rate, it measures the deviation between local thermal resistance distribution and overall operating conditions, reducing the smoothing interference of conventional heat diffusion on abnormal coking characteristics. Using the morphological drift rate to construct a dynamic window length measures the interception scale for capturing abrupt airflow changes, preventing the dilution of high-frequency blockage peaks by long-period fixed windows. By using a sliding cross-correlation algorithm to shift the pressure sequence forward and integrating the flow difference value to calculate the relative impedance to trigger early warning commands, it measures the divergence between the feed flow rate and furnace pressure change trends. It compensates for the judgment interference caused by fluid transmission delays and flow adjustment operations, reducing the frequency of false alarms caused by background thermal disturbances and improving the accuracy and reliability of early airflow blockage warnings for carbon black reactors.
[0009] Preferably, the morphological drift rate satisfies the following relationship: ; In the formula, The shape drift rate at the current moment; The value at the 90th percentile of the sorted temperature sequence within the current time window is calculated. The value at the 50th percentile of the sorted temperature sequence within the current time window is calculated. The value at the tenth percentile of the sorted temperature sequence within the current time window is used for statistical analysis. The temperature reference is the reference time window at the current moment.
[0010] This invention calculates the morphological drift rate by extracting the sorted values of the aligned temperature sequence at a preset percentile and performing a difference operation with the temperature benchmark. This measures the degree of unevenness in the local thermal resistance distribution and the proportion of deviation from the overall operating conditions, thereby improving the capture resolution of the overall temperature drop or temperature rise characteristics of the reactor and reducing the frequency of false alarms caused by normal background thermal disturbances.
[0011] Preferably, the dynamic window length satisfies the following relationship: ; In the formula, The current dynamic window length; It is a function for maximizing the value; To cut off the bottom line; This is the floor function; To calculate the length of the statistical window; Sensitivity coefficient; The shape drift rate at the current moment; It is an exponential function with the natural constant as the base.
[0012] This invention utilizes the shape drift rate combined with the sensitivity coefficient to perform scaling operations to construct a dynamic window length, which measures the optimal observation intercept scale for high-frequency sudden changes in fluid flow when obstructed. It optimizes the hysteresis defects caused by data accumulation, reduces the under-detection of pressure spikes caused by blockage by the smoothing of large-sample stable baselines, and enhances the response capability to early airflow blockage characteristics.
[0013] Preferably, the relative impedance satisfies the following relationship: ; In the formula, The relative impedance at the current moment; It is a function for maximizing the value; This represents the delayed pressure difference value of the time-shifted pressure sequence at the current moment; The standard deviation of the aligned flow sequence within the current reference time window; The standard deviation of the aligned pressure series within the current reference time window; This is the flow difference value for aligning the flow sequence at the current moment.
[0014] This invention uses a sliding cross-correlation algorithm to shift the pressure sequence forward to compensate for the delay, and integrates the flow rate difference value and the delayed pressure difference value to calculate the relative impedance. It measures the deviation between the feed flow rate and the furnace pressure change trend and the degree of actual airflow stagnation, providing a joint judgment standard for output early warning action. This reduces the normal pressure rise phenomenon under high flow rate regulation conditions and improves the working condition fit of early warning of internal resistance rise.
[0015] Preferably, piecewise linear interpolation is performed on the furnace temperature data, feed flow rate data, and furnace pressure data to obtain aligned temperature sequences, aligned flow rate sequences, and aligned pressure sequences. This includes: collecting furnace temperature data and temperature sampling time series, feed flow rate data and flow rate sampling time series, and furnace pressure data and pressure sampling time series during the continuous operation of the carbon black reactor; extracting the initial moment when the temperature sampling time series, flow rate sampling time series, and pressure sampling time series all have sampling values for the first time as the start moment; generating a timestamp sequence between the start moment and the current moment with a resampling period as the time interval, and performing piecewise linear interpolation on the furnace temperature data, feed flow rate data, and furnace pressure data accordingly to obtain aligned temperature sequences, aligned flow rate sequences, and aligned pressure sequences.
[0016] This invention extracts and generates timestamp sequences and performs piecewise linear interpolation to generate aligned temperature, flow, and pressure sequences. It measures the evolution benchmark of furnace temperature, feed flow, and furnace pressure under a unified time dimension, optimizes the data missing phenomenon at the time axis boundary, and provides synchronized data for subsequent extraction of operating status features.
[0017] Preferably, the step of obtaining the optimal number of delay points based on the standard deviations of the aligned flow sequence and the aligned pressure sequence within a reference time window, combined with an excitation threshold, includes: calculating the standard deviation of the aligned flow sequence within the reference time window to obtain the flow standard deviation; calculating the standard deviation of the aligned pressure sequence within the reference time window to obtain the pressure standard deviation; when the flow standard deviation is greater than a preset excitation threshold, using a sliding cross-correlation algorithm to process the aligned flow sequence and the aligned pressure sequence to obtain a correlation coefficient sequence, and extracting the number of time lag points corresponding to the maximum value of the correlation coefficient sequence as the optimal number of delay points.
[0018] Preferably, the step of obtaining the optimal delay point by combining the incentive threshold further includes: when the standard deviation of the flow is not greater than the preset incentive threshold, the historical best delay point is used as the optimal delay point; when running for the first time and the historical best delay point has not yet been generated, the optimal delay point is set to 0 by default.
[0019] Preferably, the step of calculating the difference between the sampled values of the aligned flow sequence at the current time and the starting time of the dynamic window to obtain the flow difference value includes: taking the current time as the endpoint, backtracking the length of the dynamic window to determine the starting time of the dynamic window; and subtracting the sampled value of the aligned flow sequence at the starting time of the dynamic window from the sampled value of the aligned flow sequence at the current time to obtain the flow difference value.
[0020] Preferably, the step of combining the impedance alarm threshold to output the early warning command includes: comparing the relative impedance ratio with the impedance alarm threshold; outputting a normal status command when the relative impedance ratio is less than the impedance alarm threshold; and outputting an early warning command when the relative impedance ratio is not less than the impedance alarm threshold.
[0021] Secondly, the present invention provides an intelligent early warning system for the production status of a carbon black reactor, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned intelligent early warning method for the production status of a carbon black reactor is implemented.
[0022] By adopting the above technical solution, a computer program is generated from the above-mentioned intelligent early warning method for the production status of a carbon black reactor and stored in a memory so that it can be loaded and executed by a processor. Terminal equipment can then be made based on the memory and processor for convenient use.
[0023] The beneficial effects of this invention are as follows: This invention generates aligned temperature, flow, and pressure sequences using a piecewise linear interpolation algorithm, optimizing the data gaps at both ends of the time axis; it calculates the morphological drift rate by fusing sorted values at different percentiles of the temperature sequence, measuring the evolution of local temperature distribution and overall temperature deviation, thus eliminating the smoothing effect of heat diffusion on local anomalies; it constructs a dynamic window length using the morphological drift rate, measuring the optimal observation scale for capturing high-frequency characteristics of airflow blockage, preventing the dilution of transient mutation characteristics by long-period data accumulation; it uses a sliding cross-correlation algorithm to shift the pressure sequence forward, and combines the flow difference value and delayed pressure difference value to calculate the relative impedance rate to trigger early warning commands, measuring the divergence between fluid obstruction and pressure surge in the feed pipeline; it compensates for the interference of flow adjustment operations and fluid transmission delays, improving the accuracy of early warnings for reactor pipeline siltation and local coking. Attached Figure Description
[0024] Figure 1 This is a flowchart of an intelligent early warning method for the production status of a carbon black reactor. Detailed Implementation
[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0026] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0027] This invention discloses an intelligent early warning method for the production status of a carbon black reactor, referring to... Figure 1 This includes steps S1-S4: S1. Collect furnace temperature data, feed flow rate data and furnace pressure data of the carbon black reactor, and use piecewise linear interpolation algorithm to generate aligned temperature sequence, aligned flow rate sequence and aligned pressure sequence.
[0028] It should be noted that the carbon black reactor generates temperature, flow rate, and pressure data simultaneously during operation. Due to differences in the activation and data upload times of different sensors on site, there are data gaps at the beginning and end of the timeline in the multi-source data. Therefore, this invention eliminates these gaps by extracting the overlapping time intervals present in the multi-source data, providing a unified temporal basis for subsequent extraction of state features.
[0029] Specifically, the furnace temperature data and temperature sampling time series, feed flow data and flow sampling time series, and furnace pressure data and pressure sampling time series are collected during the continuous operation of the carbon black reactor. The initial moment when all three sampling time series (temperature, flow, and pressure) have a sampling value for the first time is extracted as the start moment. A timestamp sequence is generated between the start moment and the current moment with a preset resampling period. The furnace temperature data is then segmented and linearly interpolated based on the timestamp sequence to obtain an aligned temperature sequence. The feed flow data is then segmented and linearly interpolated based on the timestamp sequence to obtain an aligned flow sequence. The furnace pressure data is then segmented and linearly interpolated based on the timestamp sequence to obtain an aligned pressure sequence.
[0030] In this embodiment, the preset resampling period is set to 5 seconds. The implementer can set the preset resampling period according to the difference in data acquisition frequency of the multi-source sensors.
[0031] S2. Based on the numerical distribution characteristics of the aligned temperature sequence, calculate the morphological drift rate of the current observation window.
[0032] It should be noted that during the operation of the carbon black reactor, the initial stage of temperature loss caused by air supply failure manifests as a localized temperature decay. Due to heat diffusion within the reactor, this localized decay is homogenized by the heat diffusion, resulting in a stable overall temperature curve and making it impossible to capture the early temperature loss state, thus leading to missed detections. Therefore, this invention measures the deviation between the local reactor temperature distribution and the overall temperature by calculating the morphological drift rate, thus eliminating the smoothing effect of heat diffusion on local anomalies and providing a basis for dynamically adjusting the observation window length.
[0033] Specifically, the median value of the aligned temperature sequence within a preset reference time window is extracted as the temperature reference; the values of the aligned temperature sequence within the preset statistical window are sorted from smallest to largest, and the 90th percentile value of the sorted temperature sequence is extracted; the 50th percentile value of the sorted temperature sequence is extracted; the 10th percentile value of the sorted temperature sequence is extracted; and the morphological drift rate is obtained based on the 90th percentile, 50th percentile, and 10th percentile values of the sorted temperature sequence and the temperature reference.
[0034] Specifically, the morphological drift rate satisfies the following relationship: ; In the formula, The shape drift rate at the current moment; The value at the 90th percentile of the sorted temperature sequence within the current time window is calculated. The value at the 50th percentile of the sorted temperature sequence within the current time window is calculated. The value at the tenth percentile of the sorted temperature sequence within the current time window is used for statistical analysis. The temperature reference is the reference time window at the current moment.
[0035] in, The larger the value, the greater the difference between the 90th percentile and the 50th percentile of the sorted temperature sequence. Similarly, the greater the asymmetry between the difference between the 50th percentile and the 10th percentile of the sorted temperature sequence, the more uneven the local thermal resistance distribution becomes due to localized coking blocking the symmetrical diffusion of heat within the reactor. The larger the value, the more the temperature change characteristics caused by local coking are stretched and preserved, thereby improving the early warning sensitivity of the local deterioration state of the reactor and reducing missed detections. The smaller the value, the smaller the difference between the 90th percentile and the 50th percentile of the sorted temperature sequence, and the closer the difference between the 50th percentile and the 10th percentile of the sorted temperature sequence is to being. This reflects that the heat diffusion inside the reactor is less hindered and the local thermal resistance distribution is more uniform. The smaller the value, the easier it is to smooth out and dilute the characteristics of normal temperature fluctuations, thereby reducing the frequency of warnings caused by normal thermal disturbances and avoiding false alarms.
[0036] The larger the value, the higher the proportion of the overall reactor temperature deviating from the temperature reference, making... The increase in temperature amplifies the overall temperature rise or fall characteristics, thereby enhancing the ability to detect overall deterioration caused by abnormal air supply and preventing dangerous operating conditions. The smaller the value, the closer the overall temperature of the reactor is to the temperature reference. Reducing the temperature drift allows the interference of global temperature drift characteristics on early warning judgments to be further eliminated, thereby ensuring the baseline stability under normal operation and avoiding misjudgments.
[0037] In this embodiment, the preset reference time window length is set to 2000 sampling points. The implementer can set the preset reference time window length according to the steady-state cycle of the reactor reaching thermal equilibrium after startup.
[0038] In this embodiment, the statistical window length is set to 500. The implementer can set the statistical window length according to the stable cycle of the production formula.
[0039] It should be added that the preset reference time window mainly serves a calibration function. It extracts a relatively long period of historical operating data, which is specifically used to determine the noise level and normal baseline of the reactor under stable operating conditions. The statistical window is specifically used to observe the temperature. Since the temperature change of the reactor is relatively slow, a fixed observation period must be set in order to accurately determine whether the temperature is undergoing a slow abnormal deviation, thereby calculating the morphological drift rate.
[0040] S3. Calculate the dynamic window length for the current observation period based on the morphological drift rate and basic observation parameters.
[0041] It should be noted that conventional data monitoring relies on fixed statistical windows to extract historical sequences to filter operational noise. However, airflow blockage caused by deteriorating reactor temperature distribution manifests as transient, high-frequency abrupt changes. Long-term data accumulation dilutes high-frequency fluctuations and smooths out abrupt changes, leading to a lag in status response and thus delaying early warning. Therefore, this invention adaptively calculates the dynamic window length using morphological drift rate, providing a suitable detection scale for capturing airflow blockage.
[0042] Specifically, the truncation baseline and sensitivity coefficient are obtained; the dynamic window length is obtained based on the morphological drift rate, statistical window length, truncation baseline, and sensitivity coefficient.
[0043] Specifically, the dynamic window length satisfies the following relationship: ; In the formula, The current dynamic window length; It is a function for maximizing the value; To cut off the bottom line; This is the floor function; To calculate the length of the statistical window; Sensitivity coefficient; The shape drift rate at the current moment; It is an exponential function with the natural constant as the base.
[0044] in, The larger the value, the more severe the overall temperature loss condition. Reducing the size of the window allows for greater compression of the dynamic window length, thereby improving the early warning sensitivity of extracting high-frequency blocking features and preventing high-frequency anomalies from being missed by long-cycle smoothing. The smaller the value, the more stable the temperature distribution in the thermal field, making... The increase makes the dynamic window length close to the statistical window length, thereby preserving the stable baseline characteristics of a large sample, ensuring the reliability of early warning under stable operating conditions, and avoiding local noise interference.
[0045] It should be added that the dynamic window is specifically designed to capture instantaneous changes when the fluid is obstructed. In the early coking process of the reactor, slow deviations in local temperature often indicate that the pipeline is about to be blocked. Therefore, this invention uses the morphological drift rate to adjust the length of the dynamic window. The more severe the temperature deviation, the more likely the internal blockage is to be close to the limit. At this time, the dynamic window is shortened simultaneously to reduce the occurrence of pressure spikes generated instantaneously when there is a real blockage being diluted by the previous stable data.
[0046] For example, the truncation threshold determines the minimum sample size that early warning monitoring should retain under deteriorating conditions. Empirically, this threshold ranges from 20 to 100; in this embodiment, it is set to 50. Implementers can set the truncation threshold based on the underlying computing power's anti-spread capability. For instance, when the underlying hardware computing power is weak and prone to errors due to insufficient samples, the truncation threshold can be appropriately increased to force the retention of a safe sample size; when it is necessary to capture transient anomalies, the truncation threshold can be appropriately decreased to compress the observation field.
[0047] In this embodiment, the sensitivity coefficient is set to 20. The implementer can set the sensitivity coefficient according to the degree of sudden polarization when fluid blockage occurs.
[0048] S4. Use the sliding cross-correlation algorithm to time-align the aligned pressure sequence and the aligned flow sequence, calculate the relative impedance based on the relative pressure gain and the relative flow gain, and output a warning command in combination with the preset impedance alarm threshold.
[0049] It should be noted that the fluid transport process in the feed pipeline of the carbon black reactor often involves transmission delays, causing a time misalignment between furnace pressure fluctuations and feed flow rate fluctuations. This leads to timing interference when extracting fluid characteristics for early warning, resulting in misjudgments. Therefore, this invention utilizes a sliding cross-correlation algorithm to extract the delayed pressure difference value and construct a relative impedance ratio to eliminate timing misalignment interference. Furthermore, the relative impedance ratio, which characterizes the degree of airflow blockage, is compared with a preset impedance alarm threshold, providing an execution standard and judgment output for airflow blockage early warning.
[0050] Specifically, the standard deviation of the aligned flow sequence within the reference time window is calculated to obtain the flow standard deviation; the standard deviation of the aligned pressure sequence within the reference time window is calculated to obtain the pressure standard deviation; when the flow standard deviation is greater than the preset excitation threshold, the aligned flow sequence and the aligned pressure sequence are processed using a sliding cross-correlation algorithm to obtain a correlation coefficient sequence, and the time lag point corresponding to the maximum value of the correlation coefficient sequence is extracted as the optimal delay point; when the flow standard deviation is not greater than the preset excitation threshold, the historical best delay point is used as the optimal delay point; when running for the first time and the historical best delay point has not yet been generated, the optimal delay point is set to 0 by default.
[0051] Using the current time as the endpoint, the starting time of the dynamic window is determined by tracing back the length of the dynamic window; the flow difference value is obtained by subtracting the sampled value of the aligned flow sequence at the starting time of the dynamic window from the sampled value of the aligned flow sequence at the current time; the time-shifted pressure sequence is obtained by shifting the aligned pressure sequence forward along the time axis by the optimal number of delay points; the delay pressure difference value is obtained by subtracting the sampled value of the time-shifted pressure sequence at the starting time of the dynamic window from the sampled value of the time-shifted pressure sequence at the current time.
[0052] It should be added that the time-shifted pressure sequence is obtained by shifting the aligned pressure sequence forward by the optimal number of delay points along the time axis. This is because the pressure will not rise in advance before the flow rate increases under normal circumstances. Therefore, the pressure must lag behind the flow rate in one direction in time. So the aligned pressure sequence needs to be shifted forward along the time axis.
[0053] It should be further explained that, considering the inherent delay in fluid transmission, this invention employs a data processing method of first shifting the entire sequence forward and then synchronously capturing the data. Specifically, the original aligned pressure sequence is shifted forward along the time axis by the optimal number of delay points, thereby generating a completely new time-shifted pressure sequence. When subsequently calculating the delayed pressure difference, the original pressure data is no longer called upon; instead, the newly generated time-shifted pressure sequence is directly subjected to a fixed-point capture and subtraction based on the exact same time window used for extracting the flow rate data—that is, the current moment is captured at the beginning of the dynamic window. This method of first shifting forward to generate a new sequence and then performing fixed-window value extraction ensures that the calculation steps for flow rate and pressure remain consistent in terms of instructions, compensating for the lag time in pressure response and eliminating false alarms caused by misalignment of normal timing.
[0054] The relative impedance is obtained based on the standard deviation of flow rate, standard deviation of pressure, flow rate difference score, and delayed pressure difference value; a preset impedance alarm threshold is obtained; the relative impedance is compared with the impedance alarm threshold; when the relative impedance is less than the impedance alarm threshold, a normal status command is output; when the relative impedance is not less than the impedance alarm threshold, a warning command is output.
[0055] Specifically, the relative impedance satisfies the following relationship: ; In the formula, The relative impedance at the current moment; It is a function for maximizing the value; This represents the delayed pressure difference value of the time-shifted pressure sequence at the current moment; The standard deviation of the aligned flow sequence within the current reference time window; The standard deviation of the aligned pressure series within the current reference time window; This is the flow difference value for aligning the flow sequence at the current moment.
[0056] in, Characterized by the relative pressure gain based on the pressure standard deviation. The larger the value, the greater the pressure surge inside the carbon black reactor, exceeding the normal fluctuation baseline, which allows the blockage characteristics to be extracted, thus ensuring that airflow blockage can be detected by early warning. The smaller the value, the more likely the pressure change inside the carbon black reactor is within the normal physical fluctuation range, which makes it easier to block natural pressure fluctuations and thus avoid normal false alarms.
[0057] Characterizes the relative gain of flow rate based on the standard deviation of flow rate. The larger the value, the more the carbon black reactor feed pipeline is under high flow rate regulation, which increases the denominator of the relative impedance, weakening the effect of the pressure surge characteristic and thus avoiding operational misjudgment. The smaller the value, the more restricted the fluid in the carbon black reactor feed line, which reduces the denominator of the relative impedance, amplifies the pressure buildup characteristics, and thus improves the response capability to early fluid anomalies.
[0058] It should be added that in the actual operation of the reactor, there is a linkage between the feed flow rate and the furnace pressure. An increase in the feed rate will cause the internal pressure to rise synchronously. If only a single pressure index is used for monitoring, it is easy to misjudge the pressure fluctuation caused by normal flow adjustment operation as pipeline obstruction. In fact, the characteristic of airflow blockage is not simply a pressure increase, but a disconnect between the flow rate and pressure change trends. When early coking occurs inside the reactor, causing the exhaust channel to narrow, if the feed flow rate remains constant, the furnace pressure will passively rise due to airflow stagnation. This invention shifts the core of judgment from a single parameter to the correspondence between flow rate and pressure. Once a deviation phenomenon of pressure increase without flow rate increase is detected, it can be determined as an increase in internal resistance and trigger an early warning. For pressure increase caused by normal flow regulation, the interference of operational fluctuations can be eliminated, thereby providing a judgment basis that fits the actual working conditions for early airflow blockage.
[0059] For example, the excitation threshold determines the effective fluctuation excitation lower limit required for the fluid delay alignment operation to initiate the early warning monitoring. Empirically, the value range is [0.01, 0.1]. In this embodiment, the excitation threshold is set to 0.05. Implementers can set the excitation threshold according to the sensor's noise floor level. For instance, when there is severe electromagnetic interference and high sensor background noise, the excitation threshold can be appropriately increased to prevent electrical noise from being mistaken for fluid fluctuations; when there is good electromagnetic shielding and high sensor accuracy, the excitation threshold can be appropriately decreased to capture the true fluid excitation signal.
[0060] In this embodiment, the impedance alarm threshold is set to 15. The implementer can set the impedance alarm threshold according to the physical diameter of the exhaust pipe.
[0061] This invention also discloses an intelligent early warning system for the production status of a carbon black reactor, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent early warning method for the production status of a carbon black reactor according to this invention is implemented.
[0062] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.
Claims
1. A method for intelligent early warning of the production status of a carbon black reactor, characterized in that, include: Piecewise linear interpolation of furnace temperature data, feed flow rate data, and furnace pressure data yields aligned temperature sequence, aligned flow rate sequence, and aligned pressure sequence. The median of the aligned temperature series within the reference time window is used as the temperature reference. Extract the values at the 90th, 50th, and 10th percentiles of the aligned temperature sequence within the preset statistical window, sorted from smallest to largest, and obtain the morphological drift rate by combining it with the temperature baseline; then obtain the dynamic window length based on the morphological drift rate. The optimal number of delay points is obtained by combining the standard deviation of the aligned flow sequence and the aligned pressure sequence within the reference time window with the excitation threshold. The flow difference value is obtained by calculating the difference between the sampled values of the aligned flow sequence at the current time and at the start time of the dynamic window; The time-shifted pressure sequence is obtained by shifting the aligned pressure sequence forward by the optimal number of delay points along the time axis, and the delay pressure difference is calculated by comparing the sampled values at the current time with those at the start of the dynamic window. The relative impedance is obtained by combining the standard deviation of the flow rate of the aligned flow rate sequence, the standard deviation of the pressure of the aligned pressure sequence, the flow rate difference, and the delay pressure difference. The warning command is then output in conjunction with the impedance alarm threshold.
2. The intelligent early warning method for the production status of a carbon black reactor according to claim 1, characterized in that, The morphological drift rate satisfies the following relationship: ； In the formula, The shape drift rate at the current moment; The value at the 90th percentile of the sorted temperature sequence within the current time window is calculated. The value at the 50th percentile of the sorted temperature sequence within the current time window is calculated. The value at the tenth percentile of the sorted temperature sequence within the current time window is used for statistical analysis. The temperature reference is the reference time window at the current moment.
3. The intelligent early warning method for the production status of a carbon black reactor according to claim 1, characterized in that, The length of the dynamic window satisfies the following relationship: ； In the formula, The current dynamic window length; It is a function for maximizing the value; To cut off the bottom line; This is the floor function; To calculate the length of the statistical window; Sensitivity coefficient; The shape drift rate at the current moment; It is an exponential function with the natural constant as the base.
4. The intelligent early warning method for the production status of a carbon black reactor according to claim 1, characterized in that, The relative impedance satisfies the following relationship: ； In the formula, The relative impedance at the current moment; It is a function for maximizing the value; This represents the delayed pressure difference value of the time-shifted pressure sequence at the current moment; The standard deviation of the aligned flow sequence within the current reference time window; The standard deviation of the aligned pressure series within the current reference time window; This is the flow difference value for aligning the flow sequence at the current moment.
5. The intelligent early warning method for the production status of a carbon black reactor according to claim 1, characterized in that, The step of performing piecewise linear interpolation on furnace temperature data, feed flow data, and furnace pressure data to obtain aligned temperature, flow, and pressure sequences includes: collecting furnace temperature data and temperature sampling time series, feed flow data and flow sampling time series, and furnace pressure data and pressure sampling time series during continuous operation of the carbon black reactor; extracting the initial moment when all three sampling time series have a sample value for the first time as the start moment; generating a timestamp sequence between the start moment and the current moment with a resampling period as the time interval, and performing piecewise linear interpolation on the furnace temperature data, feed flow data, and furnace pressure data accordingly to obtain aligned temperature, flow, and pressure sequences.
6. The intelligent early warning method for the production status of a carbon black reactor according to claim 1, characterized in that, The step of obtaining the optimal number of delay points based on the standard deviations of the aligned flow sequence and the aligned pressure sequence within a reference time window, combined with an excitation threshold, includes: calculating the standard deviation of the aligned flow sequence within the reference time window to obtain the flow standard deviation; calculating the standard deviation of the aligned pressure sequence within the reference time window to obtain the pressure standard deviation; when the flow standard deviation is greater than a preset excitation threshold, using a sliding cross-correlation algorithm to process the aligned flow sequence and the aligned pressure sequence to obtain a correlation coefficient sequence, and extracting the number of time lag points corresponding to the maximum value of the correlation coefficient sequence as the optimal number of delay points.
7. The intelligent early warning method for the production status of a carbon black reactor according to claim 6, characterized in that, The method of obtaining the optimal delay point by combining the incentive threshold also includes: when the standard deviation of the traffic is not greater than the preset incentive threshold, the historical best delay point is used as the optimal delay point; when running for the first time and the historical best delay point has not yet been generated, the optimal delay point is set to 0 by default.
8. The intelligent early warning method for the production status of a carbon black reactor according to claim 1, characterized in that, The calculation of the difference between the sampled values of the aligned flow sequence at the current time and the starting time of the dynamic window to obtain the flow difference value includes: taking the current time as the end point, backtracking the length of the dynamic window to determine the starting time of the dynamic window; and subtracting the sampled value of the aligned flow sequence at the starting time of the dynamic window from the sampled value of the aligned flow sequence at the current time to obtain the flow difference value.
9. The intelligent early warning method for the production status of a carbon black reactor according to claim 1, characterized in that, The method of combining the impedance alarm threshold to output a warning command includes: comparing the relative impedance ratio with the impedance alarm threshold; outputting a normal status command when the relative impedance ratio is less than the impedance alarm threshold; and outputting a warning command when the relative impedance ratio is not less than the impedance alarm threshold.
10. An intelligent early warning system for the production status of a carbon black reactor, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a method for intelligent early warning of the production status of a carbon black reactor according to any one of claims 1-9.