A real-time monitoring system and method for loss on ignition of hazardous waste in rotary kiln incineration.

Abstract

This invention discloses a real-time monitoring system and method for the loss on ignition (LOI) of hazardous waste in a rotary kiln incineration, relating to the field of hazardous waste rotary kiln incineration treatment and process control technology. The real-time monitoring system for LOI of hazardous waste in a rotary kiln incineration includes: a data acquisition module, a soft sensor model construction module, an online calibration module, and a DCS integration and display module. Through these modules, the system achieves second-level prediction, dynamic display, and early warning of LOI exceeding limits. The DCS integration and display module sets two-level alarm thresholds to achieve graded early warning for high alarms and very high alarms. Simultaneously, all predicted values ​​are stored in a historical database, supporting post-event traceability and model optimization. The system avoids the traditional "overburning" strategy, and in practical applications, it can save approximately 4.1% of natural gas consumption and significantly reduce thermal stress damage to refractory materials. It extends the service life of the rotary kiln and solves the problems of lagging, high energy consumption, and low control accuracy of traditional detection methods.

Description

Technical Field

[0001] A real-time monitoring system and method for the loss on ignition of hazardous waste in a rotary kiln incineration, relating to the field of hazardous waste rotary kiln incineration treatment and process control technology. Background Technology

[0002] Loss on ignition (LOI) is a key indicator for evaluating the effectiveness of hazardous waste incineration, reflecting the content of unburned organic matter in the residue. Current technologies for detecting LOI in rotary kiln incineration involve manual sampling and laboratory drying and ignition testing, which takes 8-9 hours. However, the incineration process is continuous and dynamic, with parameters such as kiln temperature, residence time, and disposal volume changing in real time. This results in a significant lag in test results, making real-time control impossible for operators, often forcing them to adopt an "overburning" strategy to ensure compliance.

[0003] Existing methods for detecting the loss on ignition (LOI) during hazardous waste rotary kiln incineration suffer from the following problems: First, LOI detection relies on manual sampling, making real-time monitoring impossible. Second, overburning leads to a significant increase in energy consumption, such as natural gas, resulting in high incineration energy costs and difficulty in effectively reducing natural gas consumption per unit area. Third, existing methods rely heavily on experience-based judgment, lacking scientific real-time data support, leading to insufficient detection accuracy and a lack of reliable early warning mechanisms. Furthermore, overburning keeps the rotary kiln refractory material at high temperatures for extended periods, reducing equipment lifespan. Finally, existing technologies lack a solution for real-time, accurate, and low-cost monitoring of LOI and its integration with industrial control systems.

[0004] It is evident that existing methods for detecting loss on ignition during the rotary kiln incineration of hazardous waste suffer from several drawbacks, including reliance on manual sampling, inability to monitor in real time, high incineration energy consumption, difficulty in effectively reducing natural gas consumption per unit, insufficient detection accuracy, lack of reliable early warning mechanisms, and lack of integration solutions with industrial control systems. Summary of the Invention

[0005] In view of this, the main objective of the present invention is to provide a real-time monitoring system and method for the loss on ignition of hazardous waste in rotary kiln incineration, which can solve the problems of existing technologies such as reliance on manual sampling and detection, inability to monitor in real time, high incineration energy consumption, difficulty in effectively reducing natural gas consumption per unit, insufficient detection accuracy, lack of reliable early warning mechanism, and lack of integration scheme with industrial control system.

[0006] To achieve the above objectives, the technical solution of the present invention is implemented as follows: The real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration includes: The data acquisition module is used to collect key process parameters in the rotary kiln combustion process in real time through sensors. The key process parameters include kiln tail temperature T, rotary kiln speed n, and disposal volume Q. The soft measurement model construction module is connected to the data acquisition module and is used to automatically calculate the predicted value of loss on ignition Y_raw based on the key process parameters. The soft measurement model construction module has a nonlinear mathematical model built based on the heat and mass transfer mechanism. The coefficients of the nonlinear mathematical model are determined by fitting the historical DCS process data and the corresponding measured value of loss on ignition Y_lab obtained by the laboratory through the nonlinear least squares method. The online calibration module is connected to the soft measurement model construction module. It is used to receive the measured value of the loss on ignition Y_lab obtained by the laboratory through the physical burning method, and to dynamically correct the predicted value of the loss on ignition Y_raw using the sliding window algorithm to generate the calibrated display value of the loss on ignition Y_cal. The sliding window algorithm is calculated based on the median of the deviation or the median of the ratio of the most recent N sets of paired data of predicted loss on ignition Y_raw and measured loss on ignition Y_lab, where N is the length of the sliding window; The DCS integration and display module is connected to the online calibration module. It is used to write the calibrated loss on ignition (Y_cal) display value into the DCS operation interface in real time and display it. It is also configured with an audible and visual alarm function to trigger an alarm when the Y_cal display value exceeds a preset threshold.

[0007] In a preferred embodiment, a material residence time logic control module is further established on the DCS system. This module automatically calculates the material residence time τ based on the rotary kiln's geometric parameters and its rotational speed n. The calculation formula is as follows:

[0008] In a preferred embodiment, the material residence time τ is in minutes. In a preferred embodiment, L / D is the length-to-diameter ratio of the rotary kiln; In a preferred embodiment, S is the kiln body inclination ratio, in m / m; In a preferred embodiment, n is the rotary kiln rotation speed, in r / min; In a preferred embodiment, A is an empirical coefficient related to the tribological properties of hazardous waste and can be corrected via the DCS operating interface.

[0009] In a preferred embodiment, the nonlinear mathematical model embedded in the soft measurement model construction module is:

[0010] In a preferred embodiment, the unit of the kiln tail temperature T is °C, and the unit of the processing volume Q is t; In a preferred embodiment, Q0, T0, and τ0 are pre-calibrated baseline operating conditions. In a preferred embodiment, α, β, γ, and δ are model coefficients, determined by fitting the historical DCS process data with the measured value of the heat loss on ignition Y_lab using the nonlinear least squares method; In a preferred embodiment, Y_ref is the typical loss on ignition value under the baseline operating condition.

[0011] In a preferred embodiment, the kiln tail temperature T, the material residence time τ, and the disposal volume Q are defined as follows: In a preferred embodiment, the kiln tail temperature T ranges from 600 to 1100°C. In a preferred embodiment, the material residence time τ ranges from 60 to 200 min; In a preferred embodiment, the disposal quantity Q ranges from 0.2 to 0.4 t.

[0012] In a preferred embodiment, the sliding window algorithm in the online calibration module is calibrated using a deviation correction method, and the calibration formula is: Y_cal = Y_raw + Δ; In a preferred embodiment, Δ is the median deviation of the sliding window algorithm based on the most recent N sets of paired data of the predicted loss on ignition Y_raw and the measured loss on ignition Y_lab.

[0013] In a preferred embodiment, the value of N ranges from 30 to 50.

[0014] In a preferred embodiment, the sliding window algorithm in the online calibration module is calibrated using a proportional correction method, and the calibration formula is: Y_cal=K×Y_raw; In a preferred embodiment, K is the median of the ratio of the sliding window algorithm based on the most recent N sets of paired data of the predicted loss on ignition value Y_raw and the measured loss on ignition value Y_lab. In a preferred embodiment, the value of N ranges from 30 to 50.

[0015] In a preferred embodiment, the online calibration module further includes a sliding window management unit, which is used for: In a preferred embodiment, each time the measured value of the loss on ignition Y_lab is entered, the predicted value of the loss on ignition Y_raw for the corresponding sampling time period is simultaneously recorded to form paired data and stored in the database. In a preferred embodiment, a fixed-length first-in-first-out queue is maintained, the capacity of which is equal to the sliding window length N; In a preferred embodiment, when the number of paired data groups stored in the queue reaches a preset start threshold M for the first time, calibration parameters are automatically calculated and calibration function is enabled. The preset start threshold M is a positive integer less than or equal to the sliding window length N. In a preferred embodiment, the calibration parameters are automatically recalculated and updated for each additional preset number of paired data. In a preferred embodiment, when the number of stored paired data sets is less than a preset minimum number, calibration is not enabled, and the predicted loss on ignition value Y_raw is displayed; In a preferred embodiment, when the absolute value of the calculated calibration parameter exceeds a preset deviation alarm threshold, an excessive deviation alarm is triggered, wherein the calibration parameter is the median of the deviation or the median of the ratio.

[0016] In a preferred embodiment, the DCS integration and display module further includes: In a preferred embodiment, the real-time computing unit performs a model calculation once per second and outputs the latest predicted value of the loss on ignition, Y_raw. In a preferred embodiment, the average value is calculated and output every 15 minutes to suppress fluctuations in the predicted loss on ignition value Y_raw caused by parameter anomalies. In a preferred embodiment, the human-machine interface unit is used to display the predicted value of heat loss on ignition Y_raw, the measured value of heat loss on ignition Y_lab, and the displayed value of heat loss on ignition Y_cal in real time, and supports viewing the historical trend curves of each value. The human-machine interface unit also displays the number of times the measured value of heat loss on ignition Y_lab has been entered, and automatically performs calibration when the set number of times is reached. In a preferred embodiment, the alarm unit sets two alarm thresholds. When the heat loss rate display value Y_cal exceeds the first threshold, a high alarm is triggered, and when it exceeds the second threshold, a high-high alarm is triggered, and an audible and visual alarm is issued on the DCS interface. In a preferred embodiment, the historical data storage unit stores all the predicted loss on ignition values ​​Y_raw into the DCS historical database for tracing and analysis.

[0017] A monitoring method for a real-time monitoring system of loss on ignition (LOI) in a rotary kiln for hazardous waste incineration includes the following automated steps: S1: Key process parameters during the rotary kiln combustion process are collected in real time by sensors. The key process parameters include the kiln tail temperature T, the rotary kiln rotation speed n, and the disposal volume Q. S2: Based on the rotary kiln rotation speed n and the pre-stored kiln body geometric parameters and material experience coefficients, automatically calculate the actual material residence time τ of waste in the kiln; S3: Input the kiln tail temperature T, the disposal volume Q, and the material residence time τ into the soft measurement model constructed based on the heat and mass transfer mechanism. The model automatically outputs the predicted value of loss on ignition Y_raw. S4: Periodically obtain the measured value of heat loss on ignition Y_lab obtained by the laboratory through the physical ignition method, and use the sliding window calibration algorithm to dynamically correct the predicted value of heat loss on ignition Y_raw to generate the calibrated displayed value of heat loss on ignition Y_cal. The sliding window algorithm is calculated based on the median of the deviation or the median of the ratio of the most recent N sets of paired data of predicted loss on ignition Y_raw and measured loss on ignition Y_lab, where N is the length of the sliding window; S5: Write the calibrated heat loss rate display value Y_cal into the DCS operation interface in real time and display it. When the heat loss rate display value Y_cal exceeds the preset threshold, trigger an audible and visual alarm.

[0018] In a preferred embodiment, the method for constructing the soft measurement model in S3 includes: In a preferred embodiment, continuous process data containing at least the kiln tail temperature T, the rotary kiln speed n, and the disposal volume Q are extracted from the DCS historical database and time-aligned with the measured value of the loss on ignition Y_lab at the corresponding time to form a paired sample set; In a preferred embodiment, correlation analysis is used to select the kiln tail temperature T, the rotary kiln speed n, and the disposal volume Q from candidate variables as model inputs; In a preferred embodiment, a model structure with an exponential main term plus a secondary coupling term is preset according to the heat and mass transfer mechanism; In a preferred embodiment, with the objective of minimizing the sum of squares of the differences between the predicted heat loss rate Y_raw and the measured heat loss rate Y_lab, a nonlinear least squares algorithm is used to solve for the model coefficients α, β, γ, δ and the typical heat loss rate value Y_ref under the reference operating condition.

[0019] In a preferred embodiment, the sliding window calibration algorithm in S4 includes two modes: deviation correction method and proportional correction method. The operator can select and enable the mode through the DCS human-machine interface. The system defaults to using the deviation correction method. When the absolute value of the difference between the predicted value Y_raw of the heat loss rate and the measured value Y_lab of the heat loss rate corresponding to three consecutive measured values ​​Y_lab of the heat loss rate is greater than 0.2%, the system automatically recommends switching the calibration mode and prompts the operator to confirm. In a preferred embodiment, the system further includes a model rolling optimization step: the system automatically monitors the cumulative number of paired data in the sliding window, and when the number of effective paired data exceeds a preset threshold, it automatically uses the data from the most recent three months as the training set to re-execute nonlinear least squares fitting, generates updated model coefficients, and replaces the original coefficients in the soft sensor model, thereby achieving continuous learning and optimization of the model.

[0020] The present invention provides a real-time monitoring system and method for the loss on ignition of hazardous waste in a rotary kiln incineration, which has the following beneficial effects: This real-time monitoring system for the loss on ignition (LOI) of hazardous waste in rotary kiln incineration includes: a data acquisition module, a soft measurement model construction module, an online calibration module, and a DCS integration and display module. Through these modules, it achieves second-level prediction, dynamic display, and early warning of LIO exceeding limits, suitable for real-time DCS environments, and solves the problems of slow response, high energy consumption, and low control accuracy associated with traditional detection methods.

[0021] A monitoring method for a real-time monitoring system of loss on ignition (LOI) of hazardous waste in a rotary kiln incineration includes: S1: Real-time acquisition of key process parameters during rotary kiln combustion via sensors; S2: Automatically calculate the actual material residence time τ of waste in the kiln based on the rotary kiln rotation speed n; S3: Input the kiln tail temperature T, the disposal volume Q, and the material residence time τ into the soft measurement model constructed based on the heat and mass transfer mechanism. The model automatically outputs the predicted value of loss on ignition Y_raw. S4: Periodically obtain the measured value of heat loss on ignition Y_lab obtained by the laboratory through the physical ignition method, and use the sliding window calibration algorithm to dynamically correct the predicted value of heat loss on ignition Y_raw to generate the calibrated displayed value of heat loss on ignition Y_cal. S5: Write the calibrated heat loss rate display value Y_cal into the DCS operation interface in real time and display it. When the heat loss rate display value Y_cal exceeds the preset threshold, trigger an audible and visual alarm.

[0022] The present invention provides a real-time monitoring system and method for the loss on ignition of hazardous waste in a rotary kiln incineration, which has the following beneficial effects: This real-time monitoring system for the loss on ignition (LOI) of hazardous waste in rotary kiln incineration includes: a data acquisition module, a soft measurement model construction module, an online calibration module, and a DCS integration and display module. Through these modules, it achieves second-level prediction, dynamic display, and early warning of LIO exceeding limits, suitable for real-time DCS environments, and solves the problems of slow response, high energy consumption, and low control accuracy associated with traditional detection methods.

[0023] A monitoring method for a real-time monitoring system of loss on ignition (LOI) of hazardous waste in a rotary kiln incineration includes: S1: Real-time acquisition of key process parameters during rotary kiln combustion via sensors; S2: Automatically calculate the actual material residence time τ of waste in the kiln based on the rotary kiln rotation speed n; S3: Input the kiln tail temperature T, the disposal volume Q, and the material residence time τ into the soft measurement model constructed based on the heat and mass transfer mechanism. The model automatically outputs the predicted value of loss on ignition Y_raw. S4: Periodically obtain the measured value of heat loss on ignition Y_lab obtained by the laboratory through the physical ignition method, and use the sliding window calibration algorithm to dynamically correct the predicted value of heat loss on ignition Y_raw to generate the calibrated displayed value of heat loss on ignition Y_cal. S5: Write the calibrated heat loss rate display value Y_cal into the DCS operation interface in real time and display it. When the heat loss rate display value Y_cal exceeds the preset threshold, trigger an audible and visual alarm. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of a real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration according to one embodiment of the present disclosure. Figure 2 This is a system overall module structure diagram of a real-time monitoring system for the loss on ignition of hazardous waste in a rotary kiln according to one embodiment of the present disclosure. Figure 3 A schematic diagram of the structure of an incinerator for a real-time monitoring system of loss on ignition (LOI) of hazardous waste rotary kiln incineration according to an embodiment of this disclosure. Figure 4 This is a schematic diagram of the internal structure of the soft measurement model building module of a real-time monitoring system for the loss on ignition of hazardous waste rotary kiln incineration according to an embodiment of the present disclosure. Figure 5 This is a schematic diagram of a monitoring method for a real-time monitoring system for loss on ignition (LOI) of hazardous waste rotary kiln incineration, according to one embodiment of the present disclosure. Figure 6 This is a schematic diagram of the data flow and processing hierarchy of a real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration according to one embodiment of this disclosure. Detailed Implementation

[0026] The following description, in conjunction with the accompanying drawings and embodiments of the invention, provides a more detailed account of a real-time monitoring system and method for loss on ignition (LOI) of hazardous waste in a rotary kiln.

[0027] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0028] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0029] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0030] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0031] according to Figures 1-6As shown, the real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration includes: a data acquisition module, a soft sensor model construction module, an online calibration module, and a DCS integration and display module. Through these modules, the system achieves second-level prediction, dynamic display, and early warning of LIO, solving the problems of slow response, high energy consumption, and low control accuracy associated with traditional detection methods. Further explanation of each module follows: I. Data Acquisition Module: This module collects key process parameters in real time during the rotary kiln combustion process using sensors. These parameters include the kiln tail temperature (T), the rotary kiln rotation speed (n), and the throughput (Q). Specifically, the kiln tail temperature (T) is collected in real time using a thermocouple sensor installed at the kiln tail; the rotation speed (n) is collected in real time using a speed sensor installed on the main drive shaft of the rotary kiln; and the throughput (Q) is collected in real time using a feed belt scale. The sampling frequency is 1Hz, and all data is written to the DCS real-time database in real time.

[0032] II. The soft sensor model building module, connected to the data acquisition module, is used to automatically calculate the predicted loss on ignition (Y_raw) value based on key process parameters. This connection refers to the communication link established between the data acquisition module and the soft sensor model building module via the DCS system's internal data bus or shared memory. Specifically, the data acquisition module writes real-time acquired key process parameters such as kiln tail temperature T, rotary kiln speed n, and throughput Q into the DCS real-time database as variable tags. The soft sensor model building module reads the variable tag values ​​from the database through periodic scanning or event triggering, and automatically calculates the predicted loss on ignition value Y_raw accordingly. The output of the soft sensor model building module is connected to the input of the online calibration module, used to send the calculated predicted loss on ignition value Y_raw to the online calibration module. This connection can be implemented using the DCS internal data bus, OPC interface, or inter-process communication to ensure that the predicted loss on ignition value Y_raw can be transmitted to the online calibration module in real time for subsequent dynamic correction.

[0033] The soft measurement model construction module embeds a nonlinear mathematical model based on heat and mass transfer mechanisms. This nonlinear mathematical model establishes a mathematical mapping relationship between process parameters that can be measured in real time and the loss on ignition (LOI), which is difficult to measure directly in real time, thereby achieving second-level prediction of LOI. The coefficients of the nonlinear mathematical model are determined by fitting historical DCS process data and the corresponding measured LOI value Y_lab from the laboratory using the nonlinear least squares method. This allows for the determination of the optimal values ​​of the unknown coefficients (α, β, γ, δ, Y_ref) in the nonlinear mathematical model using historical DCS process data and the corresponding measured LOI value Y_lab from the laboratory, minimizing the sum of squared errors between the model-predicted LOI and the measured LOI value Y_lab, thus ensuring the model's prediction accuracy under actual operating conditions.

[0034] Furthermore, the nonlinear mathematical model embedded in the soft measurement model building module is as follows:

[0035] In this embodiment, the kiln tail temperature T is in °C, and the throughput Q is in tons (t). It's important to note that in this specific implementation, the throughput is defined as Q in tons (t), and the material residence time is defined as τ in minutes (min). Q0, T0, and τ0 are pre-calibrated baseline operating conditions. These baseline operating conditions refer to the typical reference or central values ​​of key process parameters such as throughput Q, kiln tail temperature T, and material residence time τ when the rotary kiln is operating stably. The baseline operating conditions serve as the zero point for calculating deviations in the soft sensor model, quantifying the degree of deviation of the current operating condition from the normal operating condition. α, β, γ, and δ are model coefficients, determined by fitting historical DCS process data with the measured loss on ignition value Y_lab using the nonlinear least squares method. Y_ref is the typical loss on ignition value under the baseline operating conditions. The model, based on the heat and mass transfer mechanism, uses an exponential principal term plus a quadratic coupling term, with model coefficients fitted from historical data, resulting in high accuracy.

[0036] Of course, in order to convert the real-time rotational speed n, length-to-diameter ratio L / D, inclination rate S, and empirical coefficient A of the rotary kiln into the actual residence time τ of the material in the kiln, so as to provide key input variables for the nonlinear mathematical model, a material residence time logic control module is also established on the DCS system. The time logic control module is used to automatically calculate the material residence time τ based on the geometric parameters of the rotary kiln and the rotational speed n of the rotary kiln. The calculation formula is as follows:

[0037] Wherein, the material residence time τ is in min, L / D is the length-to-diameter ratio of the rotary kiln, S is the kiln inclination rate in m / m, and n is the rotary kiln rotation speed in r / min. A is an empirical coefficient related to the frictional characteristics of hazardous waste and can be corrected through the DCS operating interface. By automatically calculating the material residence time τ, the soft-sensor model can accurately reflect the motion characteristics of the material in the kiln, improving the model's prediction accuracy.

[0038] To avoid unpredictable prediction accuracy due to ambiguous input parameter ranges, the kiln tail temperature T, material residence time τ, and throughput Q are defined as follows: kiln tail temperature T ranges from 600 to 1100℃, material residence time τ ranges from 60 to 200 min, and throughput Q ranges from 0.2 to 0.4 t. In this specific implementation, the coefficients of the soft sensor model are obtained by fitting historical DCS data and the measured value of loss on ignition Y_lab in the laboratory within a specific operating range. Limiting the parameter range is equivalent to defining the effective working interval within which the model can guarantee a prediction error ≤ ±0.3%.

[0039] III. The online calibration module, connected to the soft sensor model building module, receives the measured loss on ignition (Y_lab) value from the laboratory using the physical ignition method. Correspondingly, the online calibration module and the soft sensor model building module establish a communication connection via the DCS system's internal data bus or shared memory. Specifically, the soft sensor model building module writes the calculated predicted loss on ignition value Y_raw into the DCS real-time database as variable labels or sends it directly to the shared memory area. The online calibration module reads the predicted loss on ignition value Y_raw from the database or shared memory through periodic scanning or event triggering, while simultaneously receiving the measured loss on ignition value Y_lab entered by the laboratory. It then executes the sliding window calibration algorithm to generate the calibrated loss on ignition display value Y_cal. This connection can also be implemented using the DCS's internal OPC interface, inter-process communication, or API calls to ensure that the predicted loss on ignition value Y_raw can be reliably and in real-time transmitted to the online calibration module.

[0040] A sliding window algorithm is used to dynamically correct the predicted loss on ignition (Y_raw) value, generating a calibrated displayed Y_cal value. The sliding window algorithm compensates for the current prediction value in real time based on the median or median ratio of the most recent N pairs of predicted and measured loss on ignition values ​​Y_raw and Y_lab. This method effectively corrects long-term systematic deviations in the soft sensor model caused by changes in operating conditions, raw material fluctuations, and equipment aging, making the final displayed Y_cal value closer to the true value and controlling the overall prediction error within ±0.3%. Here, N is the sliding window length, i.e., the number of pairs of most recent paired data (Y_raw, Y_lab) retained in the window. It determines the number of historical samples used to calculate the calibration parameters.

[0041] The following further explains the use of the deviation correction method in the sliding window algorithm of the online calibration module. The calibration formula is: Y_cal = Y_raw + Δ, where Δ is the median deviation of the sliding window algorithm based on the most recent N sets of paired data of predicted and measured loss on ignition values ​​Y_raw and Y_lab. The calculation formula is: Δ = median(Y_lab - Y_raw). Using the median to calculate Δ effectively resists the influence of individual laboratory test errors, sampling time deviations, or data entry errors on the calibration parameters. Even if a small amount of abnormal paired data is mixed in the window, the median remains robust, ensuring the reliability of the calibration. The deviation correction method only needs to maintain the queue and calculate the median, with low computational complexity. It can be executed on the DCS system at a rate of seconds without affecting the response speed of real-time monitoring.

[0042] The following further explains the proportional correction method used for calibration of the sliding window algorithm in the online calibration module. The calibration formula is: Y_cal = K × Y_raw, where K is the median of the ratio of the most recent N sets of paired data of predicted and measured loss on ignition (Y_raw) values ​​Y_lab. The calculation formula is: K = median( The proportional correction method compensates for the current prediction value by multiplying the median K of the ratios of the most recent N pairs of predicted and measured ignition loss rates (Y_raw and Y_lab) within a sliding window. When the prediction error of the soft sensor model is proportional to the size of the predicted ignition loss rate (Y_raw), for example, if the predicted value is 5% too high or 5% too low, fixed bias is difficult to compensate for effectively. The proportional correction method, by calculating the median K of the ratios of the measured and predicted ignition loss rates (Y_lab and Y_raw), reflects the scaling factor of the predicted Y_raw relative to the true value, thus more accurately eliminating proportional bias. It is suitable for calibration when the ignition loss rate fluctuates over a large range. The proportional correction method is suitable for scenarios where the error changes with the size of the predicted Y_raw, while the bias correction method is suitable for scenarios with constant offset. Both can be switched; the system defaults to using the bias correction method, and automatically recommends switching when the absolute value of the bias is greater than 0.2% for three consecutive times, enhancing the system's adaptability to different error characteristics.

[0043] Based on the above-mentioned deviation correction method and proportional correction method, the sliding window length N is defined as follows: When the sample size is insufficient, N < 30, the median is easily affected by random errors, the calibration parameters fluctuate greatly, which may lead to frequent jumps in the displayed value, affecting the operator's judgment; at the same time, the deviation of individual test values ​​will excessively affect the calibration results, reducing robustness. When the sample size is too large, N > 50, the window is too long and contains too much outdated historical data, which cannot reflect the changes in the current working conditions in a timely manner, resulting in a delayed calibration response, the system is not sensitive to recent deviations, and the prediction accuracy decreases. In this specific implementation, the value of N is in the range of 30~50. 30~50 sets of data are sufficient to make the median have good statistical representativeness, which can smooth short-term random fluctuations, while maintaining keen tracking of systematic drift. Industrial tests have verified that this range can make the displayed value of the calibrated loss on ignition stable in the long term, with the error always within ±0.3%.

[0044] The online calibration module also includes a sliding window management unit, which is used for: To avoid errors in calibration parameter calculations due to time misalignment between the measured loss on ignition (Y_lab) and the predicted loss on ignition (Y_raw), and to ensure that the paired data used for calibration accurately reflects the incineration effect of the same batch of material, the predicted loss on ignition (Y_raw) for the corresponding sampling time period is simultaneously recorded and stored in the database each time the measured loss on ignition (Y_lab) is entered. This ensures the consistency of the calibration data, allowing the median deviation or median ratio calculated subsequently to accurately reflect the systematic error of the model under the current operating conditions.

[0045] To avoid memory bloat caused by unlimited storage of historical data, or inconsistent window lengths that prevent calibration parameters from reflecting recent operating conditions, a fixed-length first-in-first-out (FIFO) queue is maintained, with a capacity equal to the sliding window length N. This fixed-length queue automatically evicts older data, ensuring the window always retains the most recent N pairs of paired data. This allows the calibration parameters Δ or K to dynamically track slow shifts in operating conditions, such as changes in raw materials or equipment aging, without manual intervention.

[0046] To avoid undefined calibration parameters or unreliable statistics if calibration is initiated directly when the system is first put into operation due to insufficient data in the window, calibration parameters are automatically calculated and the calibration function is activated when the number of paired data sets stored in the queue first reaches a preset activation threshold M. The preset activation threshold M is a positive integer less than or equal to the sliding window length N. In this way, the first calibration is automatically triggered only after a sufficient number of paired data sets M have been accumulated, ensuring that the calibration parameters are statistically representative and achieving automated activation of the calibration function without manual intervention.

[0047] To avoid the system becoming unable to adapt to changes in operating conditions due to the long-term use of the same set of calibration parameters, resulting in an accumulation of deviations in the predicted loss on ignition (Y_raw), the calibration parameters are automatically recalculated and updated every time a preset number of paired data sets are added. This achieves rolling updates of calibration parameters, enabling the system to continuously track recent operating conditions and maintain long-term stability of the calibration results. The update frequency is adjustable, such as for every 1 or 5 sets of data added, balancing response speed and stability.

[0048] To avoid excessive fluctuations in calibration parameters calculated based on a small amount of data when the sample size in the window is too small due to reduced laboratory testing frequency or data entry delays, which could cause distortion of the displayed loss on ignition value Y_cal, calibration is not enabled when the number of stored paired data sets is less than the preset minimum, and the predicted loss on ignition value Y_raw is displayed.

[0049] To prevent the system from outputting an incorrect loss on ignition (Y_cal) value due to abnormal calibration parameters caused by severely aged model coefficients, laboratory testing errors, or drastic changes in operating conditions, a large deviation alarm is triggered when the absolute value of the calculated calibration parameter exceeds a preset deviation alarm threshold. The calibration parameter is either the median deviation or the median ratio. This large deviation alarm promptly alerts operators to check the model or data source, preventing the system from remaining in an abnormal state for an extended period, thus enhancing system security and maintainability. Furthermore, this alarm can serve as a prerequisite for triggering rolling model optimization.

[0050] IV. The DCS integration and display module connects to the online calibration module. Specifically, the DCS integration and display module and the online calibration module establish a communication connection through the DCS system's internal data bus or shared memory. Specifically, the online calibration module writes the generated calibrated ignition loss rate display value Y_cal into the DCS real-time database as a variable label or sends it directly to the shared memory area. The DCS integration and display module reads the ignition loss rate display value Y_cal from the database or shared memory through periodic scanning or event triggering, then updates the displayed value on the DCS operation interface in real time and determines whether to trigger an over-limit audible and visual alarm. Furthermore, this connection can also be implemented using the DCS system's standard OPC interface, inter-process communication (IPC), or API calls to ensure that the ignition loss rate display value Y_cal can be transmitted to the display and alarm units.

[0051] The DCS integration and display module is used to write the calibrated ignition loss rate (Y_cal) value into the DCS operating interface in real time and display it. It is also configured with an audible and visual alarm function; when the Y_cal value exceeds a preset threshold, an alarm is triggered. The calibrated Y_cal value is refreshed every second, allowing operators to view the current ignition loss rate in real time on the DCS operating interface. The feedback time is reduced to the second level, providing data support for timely adjustments.

[0052] The DCS integration and display module further includes: a real-time computing unit, a human-machine interface unit, an alarm unit, and a historical data storage unit.

[0053] The real-time computing unit performs model calculations once per second and outputs the latest predicted value of heat loss rate Y_raw, or calculates the average value every 15 minutes and outputs it to suppress fluctuations in the predicted value of heat loss rate Y_raw caused by abnormal parameters. This ensures real-time performance and avoids severe fluctuations in the displayed value caused by instantaneous changes in the sensor.

[0054] The human-machine interface unit is used to display the predicted value of heat loss on ignition Y_raw, the measured value of heat loss on ignition Y_lab, and the displayed value of heat loss on ignition Y_cal in real time. It also supports viewing the historical trend curves of each value. The human-machine interface unit also displays the number of times the measured value of heat loss on ignition Y_lab has been entered. When the set number of times is reached, it will automatically perform calibration without manual intervention, which improves work efficiency.

[0055] The alarm unit has two levels of alarm thresholds. When the heat loss rate display value Y_cal exceeds the first threshold, a high alarm is triggered. When it exceeds the second threshold, a high-high alarm is triggered, and an audible and visual alarm is issued on the DCS interface. Two alarm thresholds can be set, for example, a first threshold of 4.7% and a second threshold of 4.9%. When the loss on ignition (Y_cal) exceeds the first threshold, a high alarm is triggered, indicated by a flashing yellow light. When it exceeds the second threshold, a very high alarm is triggered, indicated by a flashing red light and an audible alarm. These tiered alarms help operators quickly distinguish the severity of problems and prioritize handling high-risk situations.

[0056] The historical data storage unit stores all predicted loss on ignition values ​​Y_raw into the DCS historical database. This data can be used for post-event tracing, analyzing changes in model accuracy, and serving as training samples for rolling model optimization, enabling continuous learning and optimization of the system.

[0057] according to Figure 3 As shown, the hazardous waste incineration system used in this specific embodiment mainly includes a rotary kiln body and a secondary combustion chamber. Material is continuously fed in from the kiln tail inlet via a conveyor, flowing parallel to the high-temperature hot air generated by the hot blast stove. Under the rotation and tilting action of the kiln body, the material is evenly tumbled and dried. Large particles are discharged from the drying port of the slag discharge machine, while fine particles are conveyed by the airflow to a cyclone separator via an induced draft fan. Under centrifugal force, larger particles are separated and collected. The remaining dust-laden gas enters a bag filter through a flue, where ultrafine dust is intercepted by the filter bags. The purified gas is finally discharged from the system. A waste heat recovery device uses the heat energy in the high-temperature exhaust gas to preheat the air. The preheated air is mixed with the newly generated hot air from the hot blast stove and reinjected into the kiln body, forming a circulating heat source to improve energy utilization efficiency.

[0058] It should be noted that the throughput Q is collected by installing an electronic belt scale on the feed conveyor at the inlet. The weighing bridge is located below the belt and rigidly connected to the conveyor frame. In this specific embodiment, a pressure-resistant ICS-17A type electronic belt scale is selected. The load cell measures the weight of the material per unit length in real time, while the speed sensor measures the belt speed. The weighing instrument calculates the instantaneous flow rate Q(t) and transmits it to the analog input module of the DCS via a standard 4-20mA signal or RS485 communication using the Modbus RTU protocol. After being converted into an engineering value, the program stores it in the real-time database.

[0059] A speed sensor is used to acquire the rotary kiln speed n, and is installed on the non-shaft extension end of the main drive motor of the rotary kiln, the input / output end of the reducer, or the auxiliary transmission device. In this specific embodiment, an incremental photoelectric rotary encoder E6B2-CWZ6C is selected. The encoder generates two-phase orthogonal pulse signals A and B as the drive shaft rotates. The signals are acquired by a high-speed counting module, and the controller calculates the rotary kiln speed n (r / min) based on the number of pulses per unit time. The speed value n is then written into the DCS real-time database after range conversion.

[0060] A kiln tail temperature sensor is used to acquire the kiln tail temperature T. It is installed at the flue gas outlet of the kiln tail flue and extends vertically or obliquely into the cavity. In this specific embodiment, a WRN or SLD-RFT series assembled wear-resistant and high-temperature armored thermocouple is selected. The thermocouple wire is inserted vertically or obliquely into the temperature measurement interface of the kiln tail flue, utilizing the thermoelectric effect to generate a weak thermoelectric potential. The compensating wire sends the thermoelectric potential signal differentially to the dedicated channel of the thermocouple input analog input module of the DCS system. The module has built-in cold junction compensation, converts the thermoelectric potential into an engineering temperature value, and then writes it into the DCS real-time database via the system bus for model access.

[0061] A monitoring method for a real-time monitoring system based on the loss on ignition (LOI) of hazardous waste rotary kiln incineration includes the following automated steps: S1: Key process parameters during the rotary kiln combustion process are collected in real time using sensors. These parameters include the kiln tail temperature T, the rotary kiln rotation speed n, and the throughput Q. Specifically, the kiln tail temperature T is collected in real time using a thermocouple sensor installed at the kiln tail, the rotary kiln rotation speed n is collected in real time using a rotation speed sensor installed on the main drive shaft, and the throughput Q is collected in real time using a feed belt scale. The sampling frequency is 1Hz, and all data is written to the DCS real-time database in real time.

[0062] S2: Based on the rotary kiln rotation speed n, pre-stored kiln geometric parameters, and material experience coefficients, the actual material residence time τ in the kiln is automatically calculated. Based on the rotary kiln rotation speed n acquired in step S1, combined with the pre-stored kiln geometric parameters (length-to-diameter ratio L / D, inclination rate S, and material experience coefficient A), the material residence time logic control module calculates the actual material residence time τ according to the formula... The actual material residence time τ of waste in the kiln is automatically calculated. This calculation is performed in real time on the DCS system, and the results are used for subsequent soft-sensor models.

[0063] S3: Input the kiln tail temperature T, throughput Q, and material residence time τ into the soft sensor model built based on the heat and mass transfer mechanism. The model automatically outputs the predicted loss on ignition value Y_raw. Input the kiln tail temperature T, throughput Q, and material residence time τ obtained in step S1 into the nonlinear mathematical model embedded in the soft sensor model building module. This model is built based on the heat and mass transfer mechanism, and the model coefficients α, β, γ, δ, and Y_ref have been determined by fitting historical DCS process data with the corresponding measured laboratory loss on ignition value Y_lab using the nonlinear least squares method. The model automatically calculates and outputs the predicted loss on ignition value Y_raw.

[0064] S4: Periodically acquire the measured loss on ignition (Y_lab) value from the laboratory using the physical ignition method. Use a sliding window calibration algorithm to dynamically correct the predicted loss on ignition value Y_raw, generating the calibrated displayed loss on ignition value Y_cal. The sliding window algorithm is calculated based on the median deviation or median ratio of the most recent N pairs of predicted and measured loss on ignition values ​​Y_raw and Y_lab, where N is the sliding window length. The online calibration module uses the median deviation or median ratio of the most recent N pairs of paired data to dynamically correct the predicted loss on ignition value Y_raw using a deviation correction method or a proportional correction method, generating the calibrated displayed loss on ignition value Y_cal. The sliding window length N ranges from 30 to 50 pairs, Δ is the median deviation, and K is the median ratio. The system defaults to prioritizing the deviation correction method. When the absolute value of the predicted loss on ignition (Y_raw) corresponding to three consecutive measured loss on ignition (Y_lab) values ​​is greater than 0.2%, it automatically recommends switching to the proportional correction method. The sliding window management unit automatically maintains a first-in-first-out queue to ensure that calibration parameters are always based on the most recent N sets of valid data. When data is insufficient, it automatically backtracks to display the predicted loss on ignition (Y_raw) value. When calibration parameters are abnormal, it triggers an alarm for excessive deviation.

[0065] S5: The calibrated ignition loss rate (Y_cal) is written to and displayed in the DCS operation interface in real time. An audible and visual alarm is triggered when the Y_cal exceeds a preset threshold. The calibrated Y_cal generated in step S4 is written to and displayed in the DCS operation interface in real time, with a refresh rate of 1 second / time or a 15-minute average. Simultaneously, the alarm unit in the DCS integration and display module compares the Y_cal with two preset alarm thresholds: a first threshold and a second threshold. When the Y_cal exceeds the first threshold, a high alarm is triggered with a yellow flashing indicator; when it exceeds the second threshold, a high-high alarm is triggered with a red flashing indicator and an audible alarm. Operators can view historical trend curves through the human-machine interface unit. The historical data storage unit stores all predicted Y_raw ignition loss rates into the DCS historical database for traceability and model rolling optimization.

[0066] The methods for constructing the soft measurement model in S3 include: Continuous process data containing at least the kiln tail temperature T, rotary kiln speed n, and throughput Q are extracted from the DCS historical database and time-aligned with the corresponding measured loss on ignition value Y_lab to form a paired sample set.

[0067] Correlation analysis was used to select kiln tail temperature T, rotary kiln speed n, and throughput Q from candidate variables as model inputs. It should be noted that candidate variables refer to the set of all process parameters potentially related to the loss on ignition (LOI) extracted from the DCS historical database during the initial stage of soft sensor model construction. These parameters include, but are not limited to, kiln tail temperature T, rotary kiln speed n, and throughput Q, and may also include kiln head temperature, auxiliary fuel flow rate, primary air volume, secondary air volume, and kiln tail pressure. Correlation analysis, such as Pearson correlation coefficient and principal component analysis, was used to select parameters strongly correlated with LOI—namely, kiln tail temperature T, rotary kiln speed n, and throughput Q—as the final model input variables.

[0068] The model structure is pre-defined based on the heat and mass transfer mechanism, with an exponential main term and a secondary coupling term. The objective is to minimize the sum of squares of the differences between the predicted and measured loss on ignition (Y_raw) and the actual loss on ignition (Y_lab). A nonlinear least squares algorithm is used to solve for the model coefficients α, β, γ, δ, and the typical loss on ignition value Y_ref under the baseline operating condition.

[0069] The sliding window calibration algorithm in S4 includes two modes: deviation correction method and proportional correction method. Operators can select and enable the mode through the DCS human-machine interface. The system defaults to using the deviation correction method. When the absolute value of the difference between the predicted value of the heat loss on ignition Y_raw and the measured value of the heat loss on ignition Y_lab corresponding to three consecutive measured values ​​of heat loss on ignition Y_lab is greater than 0.2%, the system automatically recommends switching the calibration mode and prompts the operator to confirm. It also includes a model rolling optimization step: the system automatically monitors the cumulative number of paired data in the sliding window. When the number of effective paired data exceeds the preset threshold, it automatically uses the data from the most recent three months as the training set, re-executes nonlinear least squares fitting, generates updated model coefficients, and replaces the original coefficients in the soft measurement model, thereby realizing continuous learning and optimization of the model.

[0070] To illustrate the actual working process more clearly, a specific application example is given below: I. System Configuration Parameters: The rotary kiln structural parameters in this example are as follows: kiln body length 16m, inner diameter 3.4m, length-to-diameter ratio L / D=4.71, kiln body inclination rate S=0.035m / m (corresponding to a horizontal inclination angle of 2°), and empirical coefficient A is taken as 0.19.

[0071] The baseline operating conditions for the soft sensor model were set as follows: throughput Q0 = 0.3t, kiln tail temperature T0 = 850℃, material residence time τ0 = 130min, and baseline loss on ignition Y_ref = 3%. The model coefficients obtained by fitting using the nonlinear least squares method were: α = 10, β = 0.005, γ = 0.01, δ = 10^5.

[0072] The online calibration module uses a deviation correction method, with a sliding window length N=30, a preset start threshold M=20, a preset minimum effective data volume of 10 sets, and a deviation alarm threshold set to 2%. The two-level alarm thresholds of the DCS integration and display module are set as follows: first threshold 4.7% (high alarm), second threshold 4.9% (high-high alarm).

[0073] II. Data Acquisition and Residence Time Calculation: When the rotary kiln speed sensor measures the current speed n=0.3r / min, the material residence time logic control module automatically calculates the actual residence time of waste in the kiln according to the formula: .

[0074] III. Soft Sensor Model Prediction Calculation: Assuming the current kiln tail temperature T = 900℃ and the throughput Q = 0.3t obtained by the data acquisition module, and combining the material residence time τ = 85min calculated above, these three parameters are input into the soft sensor model:

[0075]

[0076]

[0077]

[0078] The model outputs a predicted loss on ignition (Y_raw) of approximately 3.66%, which is well below the alarm threshold of 4.7%, indicating that the current incineration effect is good. Subsequently, the online calibration module calculates the median deviation Δ based on the most recent 30 sets of test data, obtaining the final displayed value Y_cal = Y_raw + Δ, which is then refreshed in real-time on the DCS operation interface. Operators can use this information to determine whether adjustments to the incineration conditions are necessary.

[0079] IV. Online calibration process:

[0080] The above is a partial example of representative sampling data, showing the paired data between 30 sets of laboratory test values ​​Y_lab and the corresponding model predicted values ​​Y_raw at different times. In this specific implementation, the median of the 30 deviations (Y_lab - Y_raw) is Δ = 0.01. Taking the first set of data as an example, Y_raw = 2.37%, Y_lab = 2.42%, with a deviation of 0.05%; after calibration, Y_cal = Y_raw + Δ = 2.38%, and the relative deviation from the measured value is approximately 1.6%, meeting the requirement that the process deviation is less than 5%. The absolute values ​​of the deviations in the remaining sets are mostly within 0.1, with only one set on September 7th showing a larger deviation of 0.49. However, the median estimation effectively suppressed the influence of outliers, and the calibration parameters are stable and reliable.

[0081] V. Over-limit alarm process: During normal operation, when the rotary kiln speed increases, leading to a shorter material residence time, the Y_raw value predicted by the soft measurement model will rise accordingly, and consequently, the calibrated Y_cal will also increase. The DCS integration and display module monitors the Y_cal value in real time. Once it exceeds the first threshold of 4.7%, a high alarm is triggered: the value on the DCS operation interface begins to flash yellow, and an intermittent alarm sound is emitted. If Y_cal continues to rise and exceeds the second threshold of 4.9%, a high-high alarm is triggered: the value turns red and flashes, and the alarm sound frequency increases. Operators can manually adjust parameters such as the rotary kiln speed and throughput based on the alarm signal, or automatically adjust the throughput and rotary kiln speed through pre-set DCS interlock logic, thereby effectively controlling the loss on ignition within the standard range.

[0082] VI. Model Rolling Optimization:

[0083] The above are some representative data examples. After the system ran continuously for one month, it accumulated a large amount of effective paired data containing kiln tail temperature T, throughput Q, material residence time τ, and the corresponding measured loss on ignition Y_lab. The system automatically used the data from the most recent three months as the training set and re-performed nonlinear least squares fitting, obtaining updated model coefficients: α=2.41, β=0.0112, γ=0.0105, δ=2.45×10^5. The new model's coefficient of determination R²=0.927, and root mean square error RMSE=0.218%. The effective coverage of the model has been expanded to: kiln tail temperature 600~785℃, material residence time 60~200min, and throughput 0.20~0.39t. The optimized model is applicable to a wider range of operating conditions. Verification shows that the updated model continues to maintain high prediction accuracy and can adapt to a wider range of operating condition changes.

[0084] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention.

Claims

1. A real-time monitoring system for the loss on ignition (LOI) of hazardous waste in a rotary kiln incineration, characterized in that, include: The data acquisition module is used to collect key process parameters in the rotary kiln combustion process in real time through sensors. The key process parameters include kiln tail temperature T, rotary kiln speed n, and disposal volume Q. The soft measurement model construction module is connected to the data acquisition module and is used to automatically calculate the predicted value of loss on ignition Y_raw based on the key process parameters. The soft measurement model construction module has a nonlinear mathematical model built based on the heat and mass transfer mechanism. The coefficients of the nonlinear mathematical model are determined by fitting the historical DCS process data and the corresponding measured value of loss on ignition Y_lab obtained by the laboratory through the nonlinear least squares method. The online calibration module is connected to the soft measurement model construction module. It is used to receive the measured value of the loss on ignition Y_lab obtained by the laboratory through the physical burning method, and to dynamically correct the predicted value of the loss on ignition Y_raw using the sliding window algorithm to generate the calibrated display value of the loss on ignition Y_cal. The sliding window algorithm is calculated based on the median of the deviation or the median of the ratio of the most recent N sets of paired data of predicted loss on ignition Y_raw and measured loss on ignition Y_lab, where N is the length of the sliding window; The DCS integration and display module is connected to the online calibration module. It is used to write the calibrated loss on ignition (Y_cal) display value into the DCS operation interface in real time and display it. It is also configured with an audible and visual alarm function to trigger an alarm when the Y_cal display value exceeds a preset threshold.

2. The real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration according to claim 1, characterized in that, A material residence time logic control module is also established on the DCS system. This module automatically calculates the material residence time τ based on the geometric parameters of the rotary kiln and the rotary kiln rotation speed n. The calculation formula is as follows: Wherein, the material residence time τ is in min; L / D is the length-to-diameter ratio of the rotary kiln; S represents the kiln body inclination ratio, in m / m; n is the rotational speed of the rotary kiln, in r / min; A is an empirical coefficient related to the tribological properties of hazardous waste, and can be corrected through the DCS operating interface.

3. The real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration according to claim 2, characterized in that, The nonlinear mathematical model embedded in the soft measurement model construction module is: The unit of the kiln tail temperature T is ℃, and the unit of the processing capacity Q is t; Q0, T0, 0 represents the pre-calibrated baseline operating condition value; α, β, γ, and δ are model coefficients, determined by fitting the historical DCS process data with the measured value of the loss on ignition Y_lab using the nonlinear least squares method; Y_ref is the typical loss on ignition value under the reference operating condition; The kiln tail temperature T, the material residence time τ, and the disposal volume Q are defined as follows: The kiln tail temperature T ranges from 600 to 1100℃. The material residence time τ ranges from 60 to 200 min; The disposal quantity Q ranges from 0.2 to 0.4 t.

4. The real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration according to claim 3, characterized in that, The sliding window algorithm in the online calibration module uses a deviation correction method for calibration, and the calibration formula is: Y_cal=Y_raw+Δ; Δ is the median deviation of the sliding window algorithm based on the most recent N sets of paired data of the predicted loss on ignition value Y_raw and the measured loss on ignition value Y_lab; The value of N ranges from 30 to 50.

5. The real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration according to claim 3, characterized in that, The sliding window algorithm in the online calibration module uses a proportional correction method for calibration, and the calibration formula is: Y_cal=K×Y_raw; K is the median of the ratio of the sliding window algorithm based on the most recent N sets of paired data of the predicted loss on ignition value Y_raw and the measured loss on ignition value Y_lab; The value of N ranges from 30 to 50.

6. The real-time monitoring system for loss on ignition (LOI) of hazardous waste rotary kiln incineration according to claim 1, characterized in that, The online calibration module further includes a sliding window management unit, which is used for: Each time the measured value of the loss on ignition Y_lab is entered, the predicted value of the loss on ignition Y_raw for the corresponding sampling time period is simultaneously recorded to form paired data and stored in the database. Maintain a fixed-length first-in-first-out queue, the capacity of which is equal to the sliding window length N; When the number of paired data groups stored in the queue reaches the preset start threshold M for the first time, the calibration parameters are automatically calculated and the calibration function is enabled. The preset start threshold M is a positive integer less than or equal to the sliding window length N. Subsequently, for each additional preset number of paired data, the calibration parameters are automatically recalculated and updated; When the number of stored paired data sets is less than a preset minimum, calibration is not enabled, and the predicted loss on ignition value Y_raw is displayed; When the absolute value of the calculated calibration parameter exceeds the preset deviation alarm threshold, an excessive deviation alarm is triggered. The calibration parameter is the median of the deviation or the median of the ratio.

7. The real-time monitoring system for loss on ignition (LOI) of hazardous waste rotary kiln incineration according to claim 1, characterized in that, The DCS integration and display module further includes: The real-time computing unit performs a model calculation once per second and outputs the latest predicted value of the heat loss rate Y_raw. Alternatively, the average value can be calculated every 15 minutes and then output to suppress fluctuations in the predicted loss on ignition value Y_raw caused by parameter anomalies; The human-machine interface unit is used to display the predicted value of heat loss on ignition Y_raw, the measured value of heat loss on ignition Y_lab, and the displayed value of heat loss on ignition Y_cal in real time, and supports viewing the historical trend curves of each value. The human-machine interface unit also displays the number of times the measured value of heat loss on ignition Y_lab has been entered, and automatically performs calibration when the set number of times is reached. The alarm unit is equipped with two alarm thresholds. When the displayed value of the loss on ignition, Y_cal, exceeds the first threshold, a high alarm is triggered. When it exceeds the second threshold, a high-high alarm is triggered, and an audible and visual alarm is issued on the DCS interface. The historical data storage unit stores all the predicted loss on ignition values ​​Y_raw into the DCS historical database for traceability and analysis.

8. A monitoring method based on a real-time monitoring system for the loss on ignition (LOI) of hazardous waste rotary kiln incineration according to any one of claims 1 to 7, characterized in that, The following steps are automated: S1: Key process parameters during the rotary kiln combustion process are collected in real time by sensors. The key process parameters include the kiln tail temperature T, the rotary kiln rotation speed n, and the disposal volume Q. S2: Based on the rotary kiln rotation speed n and the pre-stored kiln body geometric parameters and material experience coefficients, automatically calculate the actual material residence time τ of waste in the kiln; S3: Input the kiln tail temperature T, the disposal volume Q, and the material residence time τ into the soft measurement model constructed based on the heat and mass transfer mechanism. The model automatically outputs the predicted value of loss on ignition Y_raw. S4: Periodically obtain the measured value of heat loss on ignition Y_lab obtained by the laboratory through the physical ignition method, and use the sliding window calibration algorithm to dynamically correct the predicted value of heat loss on ignition Y_raw to generate the calibrated displayed value of heat loss on ignition Y_cal. The sliding window algorithm is calculated based on the median of the deviation or the median of the ratio of the most recent N sets of paired data of predicted loss on ignition Y_raw and measured loss on ignition Y_lab, where N is the length of the sliding window; S5: Write the calibrated heat loss rate display value Y_cal into the DCS operation interface in real time and display it. When the heat loss rate display value Y_cal exceeds the preset threshold, trigger an audible and visual alarm.

9. The monitoring method of the real-time monitoring system for the loss on ignition of hazardous waste rotary kiln incineration according to claim 8, characterized in that, The method for constructing the soft measurement model in S3 includes: Extract continuous process data containing at least the kiln tail temperature T, the rotary kiln speed n, and the disposal volume Q from the DCS historical database, and align them with the measured value of the loss on ignition Y_lab at the corresponding time to form a paired sample set; Correlation analysis was used to select the kiln tail temperature T, the rotary kiln speed n, and the disposal volume Q as model inputs from candidate variables; Based on the heat and mass transfer mechanism, a model structure with an exponential main term and a secondary coupling term is pre-defined; With the objective of minimizing the sum of squares of the differences between the predicted heat loss rate Y_raw and the measured heat loss rate Y_lab, a nonlinear least squares algorithm is used to solve for the model coefficients α, β, γ, δ and the typical heat loss rate value Y_ref under the reference operating condition.

10. The monitoring method of the real-time monitoring system for the loss on ignition of hazardous waste rotary kiln incineration according to claim 9, characterized in that, The sliding window calibration algorithm in S4 includes two modes: deviation correction method and proportional correction method. The operator can select to enable it through the DCS human-machine interface. The system defaults to using the deviation correction method. When the absolute value of the difference between the predicted value of the loss on ignition Y_raw and the measured value of the loss on ignition Y_lab corresponding to three consecutive measured values ​​of the loss on ignition Y_lab is greater than 0.2%, the system automatically recommends switching the calibration mode and prompts the operator to confirm. It also includes a model rolling optimization step: the system automatically monitors the cumulative number of paired data in the sliding window. When the number of effective paired data exceeds the preset threshold, it automatically uses the data from the most recent three months as the training set, re-executes nonlinear least squares fitting, generates updated model coefficients, and replaces the original coefficients in the soft measurement model, thereby realizing continuous learning and optimization of the model.

Images