CBS hydrogen peroxide oxidation endpoint detection method

By employing in-situ multi-dimensional signal acquisition and online lightweight verification methods, combined with a dynamic self-correcting model, real-time and accurate detection of the oxidation endpoint of CBS hydrogen peroxide was achieved, solving the problems of detection lag and misjudgment in existing technologies and making it suitable for industrial continuous production.

CN121762649APending Publication Date: 2026-03-31HEBI RUIDA CHEM TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511996930.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-27
Publication Date
2026-03-31

AI Technical Summary

Technical Problem

In the existing technology, the CBS hydrogen peroxide oxidation endpoint detection method has the problems of strong lag, low accuracy, weak anti-interference ability and low degree of automation, which cannot meet the needs of industrial continuous production.

Method used

A method combining in-situ multi-dimensional real-time signal acquisition with online lightweight verification was adopted. The signal was acquired in real time using a platinum-saturated calomel composite electrode, an in-situ ultraviolet fiber optic sensor, and an in-situ fluorescence dissolved oxygen sensor. The characteristic functional group signals of CBS were detected by surface-enhanced Raman spectroscopy (SERS) and a dynamic self-correcting model was constructed for joint endpoint determination.

Benefits of technology

It enables real-time and accurate detection of oxidation endpoints, reduces the risk of misjudgment, improves automation, reduces production costs, and adapts to continuous production needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 31A6BFCF-2E6D-4CF2-B68F-B3BDE276F50C
    Figure 31A6BFCF-2E6D-4CF2-B68F-B3BDE276F50C
  • Figure 4B489DD3-3283-4EE2-9DE6-5BEFB0CF6AF4
    Figure 4B489DD3-3283-4EE2-9DE6-5BEFB0CF6AF4
  • Figure 81A4DCD6-274D-4463-9647-112D0E2C78FE
    Figure 81A4DCD6-274D-4463-9647-112D0E2C78FE
Patent Text Reader

Abstract

A CBS hydrogen peroxide oxidation endpoint detection method belongs to a chemical synthesis detection technology, realizes direct, real-time and accurate determination of an oxidation endpoint through a closed-loop system of in-situ potential signal, ultraviolet absorption signal and dissolved oxygen concentration multi-dimensional signal acquisition, on-line lightweight verification and dynamic model self-correction, does not need off-line sampling in a reaction process, and is simple in operation and convenient to operate. The total detection response time is not longer than 5 s, the end point judgment accuracy is high, the misjudgment risk is low, the automation degree is high, the process stability is high, and the method is suitable for CBS industrial continuous production.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Figure 1 The results show the dynamic changes of potential, 254 nm UV absorption, and solubility concentration with reaction time in this scheme. The three curves tend to stabilize at the 72 min endpoint, intuitively demonstrating the strong correlation between multi-dimensional signals and the reaction process. This proves that potential, UV absorption, and dissolved oxygen can be used as core indicators for determining the oxidation endpoint, supporting the scientific validity of multi-dimensional joint determination.

[0002] Figure 2 The study demonstrated the dynamic changes in dissolved oxygen and hydrogen peroxide concentrations. In the later stages of the reaction, hydrogen peroxide accumulated in trace amounts while dissolved oxygen rose rapidly. After the endpoint, hydrogen peroxide concentration dropped sharply while dissolved oxygen stabilized, verifying the kinetic correlation between the two and explaining the scientific basis for using dissolved oxygen as an auxiliary detection indicator for the oxidation endpoint.

[0003] Figure 3 The SERS spectral curves of the early, middle and final stages of the reaction are shown in the figure. The intensity of the SN bond characteristic peak increases significantly with the progress of the reaction, while the intensity of the cyclohexylamine internal standard peak remains constant, clearly demonstrating the positive correlation between the SN bond peak intensity and the amount of CBS generated, thus verifying the effectiveness of the online lightweight Raman verification module in directly detecting the final state characteristics.

[0004] Figure 4 Using the potential signal as an example, the changes in the upper and lower limits of the threshold for 20 production batches are shown. The temporary correction process of the abnormal batch 12 is marked, presenting the law of the model adaptively adjusting the threshold with each batch. This proves that the dynamic self-correcting model can adapt to production fluctuations, effectively reduce the risk of misjudgment in long-term operation, and demonstrate the stability and adaptability of the solution.

[0005] Example 2 This embodiment designs a control experiment. Example 1 is used as the experimental group, and the data from the first batch is used for comparison. Existing technology is selected as the control group for comparison.

[0006] 1. Control group design Control group 1 was determined using the starch-sodium hypochlorite titration method: ① Sampling: Manually collect 5-10 mL of reaction solution from the CBS hydrogen peroxide oxidation reactor every 5-10 min, and quickly cool to room temperature to terminate the local reaction; ②Pretreatment: Take 2 mL of sample, add 1 mL of 10% sulfuric acid for acidification, then add 0.5 g of KI solid, and let stand in the dark for 5 min to allow hydrogen peroxide to fully react with iodide ions to generate elemental iodine; ③ Titration: Titrate with 0.1 mol / L sodium thiosulfate standard solution until pale yellow, add 2 drops of 1% starch indicator, and continue titrating until the blue color disappears as the endpoint; ④ Calculation: Calculate the residual hydrogen peroxide concentration by the volume of sodium thiosulfate consumed. The endpoint is determined when the concentration is below 0.05 mol / L.

[0007] Control group 2 used the in-situ potential detection method: ①In-situ setup: Platinum-saturated calomel composite electrodes are set up in the reactor to collect the redox potential signal of the system in real time, with a sampling frequency of 1 time / min; ②Endpoint determination: The preset potential threshold is -50~-30 mV. When the potential stabilizes in this range and the rate of change is ≤0.1% for three consecutive cycles, it is determined to be the oxidation endpoint. ③ Operation: No sampling is required; direct in-situ detection is performed; the detection response time is ≤10 s; and the data is automatically transmitted to the control system.

[0008] 2. Comparative test results Table 1 compares the test data of this solution with existing technologies. Comparison Dimensions experimental group Control group 1 Control group 2 Detection type In-situ multidimensional joint detection + online lightweight verification Offline titration detection In-situ single signal detection Response time ≤5 s ≥30 min ≤10 s Sampling interval None (remains in place throughout) 5-10 min 1 min (no sampling required) Criteria for determining the finish line Multidimensional signal stability, characteristic peak ≥0.85 Residual hydrogen peroxide concentration ≤ 0.05 mol / L The potential is stable between -50 and -30 mV Finish line determination time 72 min 68 min 70 min CBS purity of finished product 99.3% 97.2% 98.0% Excessive consumption rate of hydrogen peroxide ≤1.0% 8.5% 4.2% Risk of misjudgment ≤0.5% 15.3% 10.2% Testing cost (per test / batch) 320 250 180 Automation level Fully automated (high level) Manual operation (low) Semi-automation (intermediate) Adapt to continuous production capacity Fully compatible Unable to adapt Partial compatibility (prone to misjudgment) Among them, the endpoint determination time, the purity of the finished CBS, and the excess hydrogen peroxide consumption rate were measured results of the hydrogen peroxide oxidation reaction of the same batch of CBS, with a reaction temperature of 60℃ and a hydrogen peroxide dropping rate of 0.5 mL / min. Figure 5 As shown; the testing cost includes reagent consumption, equipment maintenance, and labor costs; the misjudgment risk is the statistical result of 100 batches of parallel tests, misjudgment risk = (number of batches whose purity does not reach 98.5% after the endpoint is determined / total number of batches) × 100%; hydrogen peroxide excess consumption rate = (actual hydrogen peroxide consumption volume - theoretical volume required for complete reaction) / theoretical volume × 100%.

[0009] 3. Results Analysis 3.1 Real-time performance Control group 1 relied on offline sampling, manual titration, and chemical reaction, resulting in a response time exceeding 30 minutes and a Quest sampling interval of 5-10 minutes, leading to a severe lag in endpoint determination. At this point, a large amount of unreacted precursors remained in the system, directly causing incomplete oxidation. Although control group 2 achieved in-situ detection, it was essentially still an intermittent detection, unable to capture instantaneous signal changes in the later stages of the reaction. The implementation group adopted continuous in-situ detection throughout the entire process, with real-time synchronous reaction, accurately capturing the optimal endpoint at 72 minutes, solving the detection lag problem and fully matching the real-time monitoring needs of continuous industrial production.

[0010] 3.2 Accuracy of Judgment Regarding the purity of the finished product, due to the delayed judgment, the purity of the finished product CBS in control group 1 was only 97.2%, which did not meet the industrial standard of ≥98%. Control group 2 relied on a single potential signal, which was easily affected by pH fluctuations and impurities in the system. The endpoint misjudgment resulted in a purity of 98.0%, which barely met the standard but had poor stability. The experimental group directly locked the amount of the target product generated by the combined judgment logic of multi-dimensional signal stabilization and direct verification of SN bond functional groups, and the purity of the finished product reached 99.3%.

[0011] Regarding raw material consumption, the control group 1 was judged to have an excessive hydrogen peroxide consumption rate of 8.5%, and the control group 2 was judged to have an excessive consumption rate of 4.2%, both of which caused significant raw material waste; the experimental group accurately matched the reaction endpoint, and the excessive hydrogen peroxide consumption was ≤1.0%, which greatly reduced production costs.

[0012] 3.3 Reliability The human error in control group 1 and the weak anti-interference ability of single signal in control group 2 resulted in a misjudgment risk of 15.3% and 10.2% respectively. Misjudgment directly led to product batch non-conformity and production interruption. The implementation group, through multi-dimensional signal complementary verification and dynamic self-correction model to adapt to production fluctuations, had a misjudgment risk of ≤0.5%, and the long-term operational stability was greatly improved, ensuring the consistency of the production process.

[0013] 3.4 Industrial Adaptability Control group 1 relied entirely on manual operation with extremely low automation, and the cost of each batch of testing, including reagents and labor, reached 250 yuan, resulting in high long-term maintenance costs. Control group 2 achieved semi-automation, but required additional manual calibration of electrode potentials, with a comprehensive cost of 180 yuan per batch. The experimental group operated fully automatically without human intervention. Although the initial investment in the Raman module was slightly higher, it saved on reagent consumption and labor costs. Furthermore, by reducing hydrogen peroxide waste and improving product qualification rate, the long-term comprehensive cost will be reduced. At the same time, it is fully compatible with continuous production control systems and has value for large-scale industrial applications.

[0014] In summary, compared with existing technologies, the present invention achieves a comprehensive breakthrough in real-time performance, accuracy, reliability, and industrial adaptability, providing an efficient and stable technical solution for the detection of CBS hydrogen peroxide oxidation endpoints, and possesses significant technical advantages and industrialization prospects. Technical Field

[0015] This invention relates to the field of chemical synthesis detection technology, and more particularly to CBS, specifically a method for detecting the endpoint of hydrogen peroxide oxidation during CBS production. Background Technology

[0016] CBS, short for N-cyclohexyl-2-benzothiazole sulfenamide, is an important rubber vulcanization accelerator. It is characterized by fast vulcanization speed and excellent vulcanized rubber performance, and is widely used in the production of rubber products such as tires and rubber seals.

[0017] The mainstream synthesis process of CBS adopts a two-step synthesis method of 2-mercaptobenzothiazole (MBT, promoter M)-cyclohexylamine. In the first step, MBT and cyclohexylamine are used as core raw materials to carry out a condensation reaction in an aqueous solution system. The mercapto-SH group of MBT and the amino-NH2 group of cyclohexylamine undergo an acid-base neutralization reaction to generate the precursor cyclohexylamine salt of N-cyclohexyl-2-benzothiazole sulfenamide. The precursor exists in a suspended state in the aqueous solution. After the condensation reaction is completed, hydrogen peroxide is added dropwise to the system. The precursor is converted into CBS through an oxidation reaction. Hydrogen peroxide acts as an oxidant to oxidize the sulfur element in the precursor from -2 to +1, realizing the formation of SN bond and generating the target product CBS: C6H11NH-S(O)-C7H4NS. At the same time, hydrogen peroxide is reduced to water. In this optimal reaction, to avoid hydrogen peroxide decomposition and side reactions, the system temperature needs to be controlled at a relatively slow reaction temperature of 55-65℃. High local hydrogen peroxide concentrations can cause over-oxidation, where the CBS product is further oxidized into sulfone and sulfoxide impurities, such as N-cyclohexyl-2-benzothiazole sulfonamide, while simultaneously damaging the already formed CBS molecular structure. These oxygen-containing impurities in CBS can lead to shortened scorch time and premature vulcanization, resulting in a significant decrease in the quality of the finished rubber. On the other hand, if oxidation is insufficient, a large amount of unconverted precursors remain in the system, leading to reduced CBS purity and excessive impurity content in the finished product. When used for rubber vulcanization, this can cause uneven vulcanization rates and insufficient crosslinking density, also affecting rubber quality. Therefore, hydrogen peroxide must be added slowly dropwise during the oxidation reaction, and the oxidation endpoint must be strictly monitored. The degree of reaction directly determines the purity, yield, and stability of subsequent processes for the CBS product. Accurate and timely determination of the hydrogen peroxide oxidation endpoint is a core element in ensuring CBS product quality and reducing production costs.

[0018] In existing technologies, the detection methods for the oxidation endpoint of CBS hydrogen peroxide mainly include starch-sodium hypochlorite titration, offline chromatography, and single-signal detection. Among them, starch-sodium hypochlorite titration is currently the most widely used method. It involves oxidizing KI with hydrogen peroxide to generate iodine, and then titrating the starch with sodium hypochlorite after color development to indirectly calculate the remaining hydrogen peroxide concentration and determine the endpoint. However, this method has a detection cycle of 25-35 minutes, which is extremely lagging. Furthermore, cyclohexylamine in the CBS reaction system can interfere with the acidification reaction, resulting in incomplete iodine formation and low detection accuracy. It also relies on manual operation, which leads to large human errors, low automation, and is not suitable for continuous production, easily causing over-oxidation or incomplete oxidation.

[0019] In addition, there is the chemical titration method, which involves periodically sampling and titrating the remaining hydrogen peroxide with standard solutions of potassium permanganate or sodium thiosulfate. The titration results are used to determine whether the oxidation reaction is complete. This method is also cumbersome, has a long detection cycle, exhibits significant lag, and is also susceptible to human error.

[0020] Offline chromatography methods, such as high-performance liquid chromatography (HPLC), offer high detection accuracy, but require the sample to be removed from the reaction system for offline pretreatment and detection. This also results in long detection cycles, lack of real-time feedback, high equipment costs, and complex operation, making them unsuitable for online monitoring in large-scale industrial production.

[0021] Single-signal detection methods, such as potentiometric or ultraviolet absorption methods, determine the endpoint by detecting changes in a specific physicochemical signal in the reaction system. However, the CBS hydrogen peroxide oxidation reaction system has a complex composition and is subject to interference from side reactions. The single signal is easily affected by factors such as temperature, pressure, and pH of the system, leading to inaccurate endpoint determination and a high false-judgment rate.

[0022] Therefore, in view of the existing mainstream starch sodium hypochlorite titration method, which has problems such as strong lag, low accuracy, weak anti-interference ability, and low degree of automation, as well as the problems that exist in other methods, the industry urgently needs a real-time, accurate, and anti-interference-strong CBS hydrogen peroxide oxidation endpoint detection method to meet the needs of industrial continuous production. Summary of the Invention

[0023] To address the shortcomings of existing technologies, this invention proposes a method for detecting the oxidation endpoint of CBS hydrogen peroxide, enabling rapid online determination of the oxidation endpoint. This method is accurate, has strong anti-interference capabilities, ensures product quality, and reduces production costs.

[0024] The technical problem to be solved by the present invention is achieved through the following technical solution: A method for detecting the oxidation endpoint of CBS hydrogen peroxide includes the following steps: (1) Real-time acquisition of in-situ multidimensional signals Three sets of integrated in-situ detection units were installed in the middle of the CBS hydrogen peroxide oxidation reactor, near the drip inlet, and at the bottom of the reactor to ensure the representativeness of the signal acquisition. Each set of integrated in-situ detection units includes: In-situ platinum-saturated calomel composite electrode: used to collect potential signals of the reaction system, with a detection range of -500~500mV and an accuracy of ±0.5 mV. The electrode probe is protected by a polytetrafluoroethylene sleeve to prevent signal drift caused by the adhesion of suspended particles in the system. In-situ ultraviolet fiber optic sensor: The probe is inserted into the reaction system, with a fixed detection wavelength of 254 nm, an optical path of 0.5 cm, a response time of ≤2 s, and real-time acquisition of ultraviolet absorption signals; In-situ fluorescence dissolved oxygen sensor: Real-time acquisition of dissolved oxygen concentration signal in the system, detection range 0-20 mg / L, accuracy ±0.05 mg / L, sampling frequency 10 Hz, avoiding interference of oxygen release on the detection results.

[0025] After the hydrogen peroxide oxidation reaction is started, the three detection units synchronously collect signals in real time, and take the average of the signals from each group as the effective signal to eliminate signal deviation caused by uneven local reaction.

[0026] (2) Online lightweight verification In the flow stability zone of the reactor's circulation pipeline, an online lightweight verification module is connected in series. Based on surface-enhanced Raman spectroscopy (SERS), the characteristic functional group signals of the target product CBS in the reaction system are directly detected. Specific settings are as follows: Detection parameters: excitation wavelength 785 nm to avoid fluorescence interference, laser power 50 mW to avoid sample damage, detection time ≤3 s, spectral acquisition range 500-1800 cm⁻¹ -1 ; Feature identification: using the SN bond characteristic peak of CBS at 1075 cm⁻¹ -1 The target peak is the characteristic peak of cyclohexylamine at 1452 cm⁻¹. -1 The target peak area is the internal standard peak. The ratio of the target peak area to the internal standard peak area is calculated, which is the relative intensity of the SN bond characteristic peak. This intensity directly reflects the amount of CBS generated in the system and is the core direct indicator of the final state of the reaction.

[0027] (3) Signal preprocessing and dynamic model construction Real-time preprocessing of the in-situ multidimensional signals acquired in step (1): Smoothing and Denoising: Wavelet thresholding is used to remove random noise and motor fluctuation interference from the signal; Standardization and normalization: z-score standardization is adopted, with the formula x'=(x-μ) / σ, where x is the real-time signal value, μ is the real-time mean of the signal, and σ is the real-time standard deviation, eliminating the influence of differences in signal dimensions and slight temperature fluctuations.

[0028] Construct a dynamic self-correcting endpoint determination model based on LSTM: Model training: Collect more than 100 batches of qualified CBS in-situ multi-dimensional signal time series data and SN bond feature peak relative intensity data. Use multi-dimensional preprocessed signals and feature peak relative intensity as inputs, and use reaching the endpoint and not reaching the endpoint as output labels to train the LSTM time series prediction model and optimize the model parameters. Initial threshold setting: Based on the training data, the initial joint judgment thresholds are set as follows: point signal -50~-30 mV, 254 nm ultraviolet absorption signal 0.35-0.45 AU, dissolved oxygen concentration 10-12 mg / L, with 10 s as one detection cycle, the absolute value of the change rate of each signal over 5 consecutive detection cycles ≤0.05%, and the relative intensity of the SN bond characteristic peak ≥0.85.

[0029] (4) Joint endpoint determination The preprocessed in-situ multidimensional signal is input into the dynamic self-correcting model, and combined with the online lightweight validation results, a joint endpoint determination is performed. If all four of the following conditions are met simultaneously, the oxidation endpoint is directly determined to have been reached, and the addition of hydrogen peroxide should be stopped immediately: a) The potential signal is stable in the range of -50 to -30 mV, and the absolute value of the change rate over five consecutive detection cycles is ≤0.05%; b) The 254 nm ultraviolet absorption signal is stable in the range of 0.35-0.45 AU, and the absolute value of the change rate over five consecutive detection cycles is ≤0.05%; c) The dissolved oxygen concentration changes from a rapidly rising state to a stable state, with a stable value in the range of 10-12 mg / L, and the fluctuation over five consecutive detection cycles is ≤0.2 mg / L; d) The relative intensity of the SN bond characteristic peak is ≥0.85. If the above conditions are not met simultaneously, the endpoint is not reached, and the oxidation reaction continues. If conditions a, b, and c are met for three consecutive testing cycles, but condition d is not met, the model triggers an early warning and automatically reduces the hydrogen peroxide dripping rate to 50% of the original rate while continuing to monitor. If condition d is still not met for the next five testing cycles, it is determined to be an abnormal reaction, such as substandard raw material purity or reaction temperature deviation. The hydrogen peroxide dripping is stopped and a fault prompt is output, indicating that abnormal intervention should be performed.

[0030] (5) Model dynamic self-correction After each batch of reactions is completed, the model automatically initiates a correction process: Normal batch correction: If, after the endpoint determination of the batch, the CBS purity of the system sample is detected by HPLC to be ≥98.5%, then the multi-dimensional signal feature values ​​and the relative intensity of the SN bond feature peaks at the endpoint of the batch are extracted, and weighted averaged with the historical qualified batch data, with the weight of the most recent 3 batches at 60% and the weight of other historical batches at 40%, and the joint determination threshold range of the model is updated to achieve cross-batch adaptation. Abnormal batch shielding: If a batch has signal abnormalities or product purity does not meet the standards, the model automatically marks the batch data as abnormal and shields its impact on threshold updates. The average threshold of the three most recent qualified batches is used as the temporary judgment threshold for the next batch. After the abnormality is eliminated, such as after changing raw materials or calibrating the temperature, the normal correction logic is restored.

[0031] In this invention, the endpoint of the oxidation reaction is detected using in-situ multi-dimensional signal acquisition. The CBS hydrogen peroxide oxidation reaction is a typical redox-driven functional group transformation reaction. The precursor cyclohexylamine salt, under the action of hydrogen peroxide, forms an SN bond through electron transfer. During the reaction, the redox potential and characteristic wavelength ultraviolet absorption of the system exhibit dynamic changes until they stabilize at the endpoint. Furthermore, these macroscopic physicochemical signals are completely synchronized with the microscopic reaction level; that is, the electron transfer rate equals the potential change rate, and the CBS production amount equals the ultraviolet absorption intensity. The in-situ sensor used in this method directly contacts the reaction system. The sensor probe captures the system's potential, ultraviolet absorption, and dissolved oxygen signals in real time. The signal transmission delay is ≤5 s, limited only by the response speed of the data acquisition card, effectively eliminating the lag in sampling and detection, and ensuring complete synchronization between the detection signal and the microscopic state of the reaction system.

[0032] The three dimensions employed in this scheme essentially provide complementary characterization of the CBS oxidation process from different fundamental reaction perspectives, avoiding interference and bias from a single signal: The potential signal is based on electron transfer in the redox reaction. The core of the CBS oxidation reaction is the oxidation of -2 S in the precursor to +1 by hydrogen peroxide. During the reaction, the redox potential of the system continuously increases with the amount of electron transfer. When the precursor is completely converted to CBS, electron transfer terminates, and the potential tends to stabilize (-50~-30 mV). The potential signal directly reflects the completion degree of the redox reaction and serves as a benchmark for electron transfer during the reaction. The 254 nm ultraviolet absorption signal is based on Beer-Lambert's law. The benzothiazole ring in the CBS molecule exhibits characteristic absorption at a wavelength of 254 nm, with a molar absorption coefficient ε = 1.2 × 10⁻⁶. 4 L·mol -1 ·cm -1 The precursor and byproducts show no significant absorption at this wavelength. As the amount of CBS generated increases, the UV absorption intensity continues to rise. At the endpoint, the concentration stabilizes and the absorption intensity tends to be constant (0.35-0.45 AU). This signal directly reflects the amount of target product generated and is a concentration scale at the final state of the reaction. Dissolved oxygen, on the other hand, is reacted from a different perspective.

[0033] Because hydrogen peroxide is added slowly and uniformly, the concentration of hydrogen peroxide in the system is always in a dynamic equilibrium of addition and consumption. Hydrogen peroxide exhibits two reactions: the main reaction of CBS oxidation and the side reaction of spontaneous decomposition (2H2O2→2H2O+O2↑). In the early stage of the reaction (0-40% progress), the precursor concentration in the system is high, and the consumption rate of hydrogen peroxide is much greater than the dropping acceleration rate. Therefore, the instantaneous concentration of hydrogen peroxide in the system remains extremely low, almost consumed immediately upon addition. The dissolved oxygen concentration in the solution slowly increases from the initial 2-3 mg / L to 5-6 mg / L. In the middle stage of the reaction (40-80% progress), the precursor concentration decreases, and the consumption rate and dropping acceleration rate tend to balance. The instantaneous concentration of hydrogen peroxide in the system stabilizes at a very low constant value, usually 50-100 mg / L. The decomposition side reaction rate is constant, and the amount of O2 generated and released tends to balance. The dissolved oxygen concentration fluctuates slightly but remains stable at 6-8 mg / L overall. In the late stage of the reaction (80% to the endpoint), the precursor is nearly exhausted, and the consumption rate of the main reaction decreases sharply. If the addition continues at the original rate, a trace amount of hydrogen peroxide will accumulate in the system, and the instantaneous concentration will rise to 200-300 mg / L. At a concentration of mg / L, the rate of decomposition side reactions accelerates with increasing hydrogen peroxide concentration, leading to a surge in O2 generation and a rapid increase in dissolved oxygen concentration from 8 mg / L to 10⁻¹² mg / L. Once the endpoint is determined, the addition is immediately stopped. The remaining trace amounts of hydrogen peroxide are completely consumed by the residual precursor within 1-2 minutes, resulting in a final hydrogen peroxide content ≤50 mg / L. At this point, the decomposition side reactions terminate, O2 generation approaches zero, and the dissolved oxygen concentration stabilizes at 10⁻¹² mg / L, consistent with the O2 dissolution equilibrium under the system's temperature and pressure. Therefore, while dissolved oxygen concentration does not directly reflect hydrogen peroxide concentration, it is an indirect but reliable indicator of the reaction's progress. In-situ dissolved oxygen sensors offer fast response and strong anti-interference capabilities, making them more suitable for industrial online monitoring compared to direct hydrogen peroxide concentration detection. The rapid rise to a stable dissolved oxygen concentration can be cross-validated with the fluctuations in potential and UV absorption signals, further eliminating false endpoints.

[0034] In this invention, an online lightweight verification based on SERS is employed to directly detect the final state characteristics, replacing HPLC. The characteristic peak of the SN bond of CBS is used as the target peak. This peak corresponds to the stretching vibration mode of the SN bond and is the characteristic functional group peak of CBS. The precursor and byproduct do not have this vibration mode, and there is no interference peak overlap. Its peak intensity is linearly positively correlated with the CBS concentration. The characteristic peak of the CH bond of cyclohexylamine is used as the internal standard peak. Cyclohexane is used as the reaction raw material. Its concentration in the system is stable. It is in excess and does not participate in the oxidation reaction. Its CH bond bending vibration peak intensity is constant. It is used as the internal standard peak to eliminate interference from instrument fluctuations, system concentration changes, etc. The amount of CBS generated is directly detected by the ratio of the characteristic functional group peak and the internal standard peak, realizing direct verification of the reaction final state and solving the problems of lag and high cost of traditional verification methods.

[0035] Therefore, the joint determination of the endpoint in this invention is based on two independent chains of evidence: the stability of the physicochemical signals of the reaction process and the attainment of characteristic functional groups of the target product. Stable physicochemical signals are a necessary condition for the reaction to reach its endpoint; signal stability indicates that the redox reaction, product formation, and oxidant consumption have all reached a steady state. However, this stability may be a false signal due to interference from impurities. Attainment of the SERS characteristic peak is a sufficient condition for the reaction to reach its endpoint; the relative intensity of the SN bond characteristic peak is ≥0.85. Based on the quantitative relationship between functional group formation and product purity, the attainment of the target product is directly verified.

[0036] Compared with the prior art, the present invention has the following advantages: (1) This application adopts an in-situ detection combined with online lightweight verification mode, which does not require offline sampling and the total detection response time is no more than 5 s, realizing real-time synchronous monitoring of the reaction and completely solving the detection lag of traditional methods; (2) This application uses in-situ multidimensional signal stabilization and direct detection of characteristic functional groups of target products as the joint judgment logic, rather than relying on model prediction. The endpoint judgment is more accurate and reliable. At the same time, the dynamic self-correcting model can adapt to the differences in raw material batches and the fluctuation of reaction conditions, further reducing the risk of misjudgment. (3) This application uses an online lightweight Raman verification module to replace offline HPLC, which reduces equipment maintenance costs by more than 60%, and eliminates the need for titration reagents and chromatographic reagents, reducing reagent consumption and waste liquid treatment costs, and significantly reducing detection costs; (4) The entire detection, judgment and correction process of this application is fully automated and requires no manual intervention. It can be directly connected to the CBS continuous production control system, with a high degree of automation. At the same time, the early warning and intervention mechanism can also avoid production accidents and ensure process stability. (5) This application uses a self-correcting model to adapt to the differences in raw materials and reactor specifications of different manufacturers, without the need for frequent manual adjustment of parameters. It has a wide range of applications and strong practicality. Attached Figure Description

[0037] Figure 1 This is a multi-dimensional in-situ signal change curve during the reaction process of this application; Figure 2 The curves showing the relationship between dissolved oxygen and hydrogen peroxide concentration at different stages of the reaction process in this application are shown. Figure 3 Comparison of SERS spectra at different stages of the online lightweight verification process in this application; Figure 4 Potential threshold update curves for different batches of LSTM models in this application; Figure 5 This is a bar chart comparing the detection method of this application with existing technologies. Detailed Implementation

[0038] The present invention will be further described below with reference to the accompanying drawings and specific preferred embodiments, but this does not limit the scope of protection of the present invention.

[0039] Example 1 1. Preparation A CBS hydrogen peroxide oxidation reactor was constructed, including a 50 L reactor, a stirrer, a constant temperature water bath, a hydrogen peroxide dripping device, and a circulation pipeline, with a stable flow rate of 5 L / min. An integrated in-situ detection unit was installed in the middle of the reactor, near the dripping port, and at the bottom of the reactor. This unit included a platinum-saturated calomel composite electrode, an ultraviolet fiber optic sensor, and a fluorescence dissolved oxygen sensor. All three sensors were connected to an industrial control computer via a data acquisition card. An online lightweight Raman verification module was connected in series in the circulation pipeline. The module's output was connected to the same industrial control computer, which had built-in signal preprocessing software, an LSTM dynamic self-correcting model, and a judgment program.

[0040] 2. Detection 2.1 Add 10 kg of 2-mercaptobenzothiazole, 6.5 kg of cyclohexylamine, and 20 L of deionized water to the reactor. Turn on the stirrer at 300 r / min and control the system temperature at 60℃ using a constant temperature water bath. Adjust the pH to 8.5 and start the hydrogen peroxide dropping device to add 30% hydrogen peroxide at an initial dropping rate of 0.5 mL / min. 2.2 Three integrated in-situ detection units and online lightweight Raman verification modules were started to collect potential signals, 254nm ultraviolet absorption signals, dissolved oxygen concentration signals and relative intensities of SN bond characteristic peaks in real time. The dissolved oxygen sampling frequency was 10 Hz and the SERS detection cycle was 10 s / time. 2.3 The industrial control computer performs real-time wavelet threshold denoising and z-score normalization preprocessing on the in-situ signal, and inputs the preprocessed signal into the LSTM dynamic self-correcting model. 2.4 When the reaction proceeded for 70 min, the in-situ signal showed: potential -42 mV (change rate of 0.03% over 5 consecutive cycles), UV absorption 0.40 AU (change rate of 0.02% over 5 consecutive cycles), and dissolved oxygen concentration 10.5 mg / L (fluctuation of 0.1 mg / L over 5 consecutive cycles). However, the relative intensity of the SN bond characteristic peak was 0.78 (<0.85). The model determined that the endpoint had not been reached, and the reaction continued. 2.5 When the reaction proceeds to 72 min, the in-situ signal becomes stable and unchanged, and the relative intensity of the SN bond characteristic peak reaches 0.88 (≥0.85). At the same time, all the joint judgment conditions are met, and the computer immediately outputs a signal that the endpoint has been reached, controlling the hydrogen peroxide dripping device to stop dripping. 2.6 A small amount of the reaction system sample was taken for HPLC verification. The detection conditions were C18 column, methanol-water (70:30) mobile phase, flow rate 1.0 mL / min, and detection wavelength 254 nm. The results showed that the purity of CBS was 99.3%, and the batch was judged to be normal. The model automatically extracted the endpoint signal feature value of the batch and updated the threshold range. 2.7 When reacting in the next batch, the model used the updated threshold for judgment. Due to slight differences in raw material batches, the model automatically fine-tuned the potential signal threshold to -48~-32 mV. The final determination of the endpoint time was 71.5 min, and the CBS purity was 99.1%, showing good adaptability.

Claims

1. A method for detecting the oxidation endpoint of CBS hydrogen peroxide, characterized in that, Includes the following steps: (1) Real-time acquisition of multi-dimensional signals in situ: An integrated in-situ detection unit is set up in the CBS hydrogen peroxide oxidation reactor. The integrated in-situ detection unit includes an in-situ platinum-saturated calomel composite electrode, an in-situ ultraviolet fiber optic sensor and an in-situ fluorescence dissolved oxygen sensor. After the hydrogen peroxide oxidation reaction is started, the potential signal, the 254 nm characteristic ultraviolet absorption signal and the dissolved oxygen concentration signal of the reaction system are acquired in real time. (2) Online lightweight verification: An online lightweight verification module is connected in series in the circulation pipeline of the reactor. Based on surface-enhanced Raman spectroscopy (SERS) technology, the characteristic functional group signal of the target product CBS in the reaction system is directly detected, that is, the characteristic peak of the SN bond of CBS at 1075 cm⁻¹. -1 The relative intensity; (3) Signal preprocessing and dynamic model construction: Real-time preprocessing of the in-situ multidimensional signals acquired in step (1), including smoothing and denoising and normalization; and construction of a dynamic self-correcting endpoint determination model based on LSTM. (4) Joint endpoint determination: The pre-processed in-situ multidimensional signal is input into the dynamic self-correction model and combined with the online lightweight verification results to determine the joint endpoint: If the following four conditions are met at the same time, the oxidation endpoint is directly determined and the addition of hydrogen peroxide is stopped immediately: a) The potential signal is stable in the range of -50~-30 mV and the absolute value of the change rate is ≤0.05% for 5 consecutive detection cycles; b) The 254 nm ultraviolet absorption signal is stable in the range of 0.35-0.45 AU and the absolute value of the change rate is ≤0.05% for 5 consecutive detection cycles; c) The dissolved oxygen concentration changes from a rapid rising state to a stable state, with a stable value in the range of 10-12 mg / L and a fluctuation of ≤0.2 mg / L for 5 consecutive detection cycles; d) The relative intensity of the SN bond characteristic peak is ≥0.85; (5) Model dynamic self-correction: After each batch reaction is completed, the model automatically starts the correction process: If the endpoint of the batch is determined, the joint judgment threshold range of the model is updated to achieve cross-batch adaptation. If a batch has abnormal signal or product purity is not up to standard, the model automatically marks the data of the batch as abnormal, shields its impact on the threshold update, and uses the average threshold of the three most recent qualified data as the temporary judgment threshold for the next batch.

2. The detection method according to claim 1, characterized in that, The integrated in-situ detection unit in step (1) consists of three sets, which are respectively located in the middle of the oxidation reactor, near the drip inlet, and at the bottom of the reactor.

3. The detection method according to claim 1, characterized in that, In step (1), the detection range of the in-situ platinum-saturated calomel composite electrode is -500~500 mV, with an accuracy of ±0.5 mV, and the electrode probe is protected by a polytetrafluoroethylene sleeve; the in-situ ultraviolet fiber optic sensor has a fixed detection wavelength of 254 nm, an optical path of 0.5 cm, and a response time of ≤2 s; the in-situ fluorescence dissolved oxygen sensor has a detection range of 0-20 mg / L, an accuracy of ±0.05 mg / L, and a sampling frequency of 10 Hz.

4. The detection method according to claim 1, characterized in that, The detection parameters for the online lightweight verification module in step (2) are: excitation wavelength 785 nm to avoid fluorescence interference, laser power 50 mW to avoid damaging the sample, detection time ≤ 3 s, and spectral acquisition range 500-1800 cm⁻¹. -1 .

5. The detection method according to claim 4, characterized in that: In step (2), the characteristic peak of the SN bond of CBS is 1075 cm⁻¹. -1 The target peak is the characteristic peak of cyclohexylamine at 1452 cm⁻¹. -1 The internal standard peak is the characteristic peak of the SN bond of the CBS at 1075 cm⁻¹. -1 The relative intensity is the ratio of the target peak area to the internal standard peak area.

6. The detection method according to claim 1, characterized in that: Step (3) Signal preprocessing smoothing and denoising adopts wavelet threshold denoising method, and standardization and normalization adopt z-score standardization. The formula is x'=(x-μ) / σ, where x is the real-time signal value, μ is the real-time mean of the signal, and σ is the real-time standard deviation.

7. The detection method according to claim 1, characterized in that, Step (4) also includes an abnormal intervention step: if conditions a, b, and c are met for three consecutive detection cycles but condition d is not met, the model triggers an early warning and automatically reduces the hydrogen peroxide dripping speed to 50% of the original rate, and continues to detect; if condition d is still not met for the next five detection cycles, it is determined to be an abnormal reaction, the hydrogen peroxide dripping is stopped and a fault prompt is output, prompting abnormal intervention.