An intelligent quality control system for peroxide raw materials

Abstract

The present invention relates to the technical field of peroxide raw material quality monitoring, and discloses an intelligent peroxide raw material quality control system. The system comprises: a differential heat flow sensing unit for continuously measuring the temperature gradient between the raw material barrel wall and the ambient air; a posture sensing unit for monitoring the physical motion state; an edge processing unit for performing a time-series integration of the temperature gradient when the raw material barrel is stationary to calculate a degradation degree value, and triggering an early warning when a threshold is exceeded. The present invention achieves continuous quantitative tracking of the peroxide raw material degradation process, avoiding the limitations of traditional discrete detection, ensuring measurement authenticity through the dual mechanisms of motion state recognition and signal filtering, and ensuring long-term monitoring reliability in combination with self-heating pulse calibration technology, thereby providing a new process monitoring method for chemical storage safety.

Description

Technical Field

[0001] The invention relates to an intelligent peroxide raw material quality control system, belonging to the technical field of peroxide raw material quality monitoring. Background Art

[0002] In the field of peroxide raw material storage safety monitoring, existing technologies mainly rely on periodic manual spot checks and fixed temperature threshold alarm mechanisms. The former analyzes degradation indicators through discrete sampling, and the latter triggers an alarm when the barrel wall temperature exceeds a preset threshold. In the densely stacked scenarios of typical chemical warehouses, the inherent defects of this method become more prominent with the monitoring duration and environmental complexity: on the one hand, the static threshold mechanism can only capture instantaneous overtemperature conditions, and has a perception blind spot for the continuous degradation process of raw materials that are at subcritical temperatures for a long time, and cannot quantify the cumulative decomposition risk; on the other hand, the thermal disturbances and mechanical vibrations caused by the start and stop of warehouse ventilation and cargo handling cause single-point temperature monitoring data to contain a large amount of high-frequency noise. The existing technology lacks an effective means to separate the raw material's own heat generation trend from the mixed signals.

[0003] The industry has attempted to introduce multi-sensor fusion strategies, such as adding vibration monitoring modules to filter out handling interference or using thermal imaging technology to improve spatial resolution. However, these improvements have failed to resolve the core contradictions: First, vibration sensors can only identify physical motion states and are powerless to address changes in natural convection patterns and resulting heat transfer deviations caused by the tilt of the barrels during static stacking; second, thermal imaging equipment is expensive and relies on centralized power supply, making it difficult to achieve large-scale and universal deployment.

[0004] A thorough analysis reveals the fundamental bottlenecks of existing technologies: 1. The inability to continuously track cumulative degradation effects—discrete detection cannot capture the continuous entropy increase at subcritical temperatures; 2. Signal distortion due to multi-source interference coupling—mechanical vibration, environmental thermal noise, and static posture deviations contaminate the core monitoring signal; and 3. The lack of a mechanism to ensure long-term reliability—sensor drift and the evolution of material thermal properties cause the measurement model to gradually fail. Therefore, the technical challenges addressed by this invention are how to achieve continuous, quantitative tracking of the degradation of peroxide raw materials with limited hardware resources and ensure long-term measurement reliability in complex industrial environments. Summary of the Invention

[0005] The present invention provides an intelligent quality control system for peroxide raw materials, the main purpose of which is to solve the problem that the existing technology cannot continuously and reliably quantify and track the cumulative deterioration degree of peroxide raw materials in complex storage environments.

[0006] To achieve the above objectives, the present invention provides an intelligent peroxide raw material quality control system, comprising:

[0007] A differential heat flow sensing unit is provided, which is configured to continuously measure the temperature gradient between the wall temperature of the raw material barrel and the ambient air temperature around the raw material barrel;

[0008] A posture sensing unit is provided, and the posture sensing unit is configured to monitor the physical motion state of the raw material barrel;

[0009] An edge processing unit is configured as follows: when the posture sensing unit indicates that the raw material barrel is in a physical steady state, the temperature difference gradient is accumulated over time according to a nonlinear weight function pre-set for the peroxide raw material, and a degradation time series integral value representing the degree of cumulative degradation of the peroxide in the raw material barrel is calculated; when the degradation time series integral value exceeds the safety threshold configured for the peroxide raw material, an early warning signal is sent to the external management system.

[0010] Preferably, the differential heat flow sensing unit includes a first thermistor close to the wall of the raw material barrel and a second thermistor exposed to the ambient air around the raw material barrel, and the temperature gradient is the difference between the measured temperature of the first thermistor and the measured temperature of the second thermistor.

[0011] Preferably, the edge processing unit is further configured to: accumulate the pause time when the posture sensing unit indicates that the raw material barrel is in a physical motion state.

[0012] Preferably, the edge processing unit is further configured to: when the posture sensing unit indicates that the raw material barrel is in a physical motion state, reduce the weight of the temperature difference gradient in the cumulative calculation.

[0013] Preferably, the posture sensing unit is a micro-electromechanical system accelerometer, and the physical motion state is determined based on the frequency spectrum characteristics or variance changes of the output data of the micro-electromechanical system accelerometer.

[0014] Preferably, the edge processing unit is further configured to: perform a digital low-pass filtering algorithm on the time series data of the temperature difference gradient to obtain a low-frequency trend signal generated by the heat generated by the decomposition of the peroxide raw material itself; and only use the low-frequency trend signal to calculate the degradation time series integral value.

[0015] Preferably, the edge processing unit is further configured to: when the posture sensing unit indicates that the raw material barrel is in a physical steady state, analyze the output data of the posture sensing unit to determine the static tilt angle and direction of the raw material barrel; and according to the static tilt angle and direction, query and obtain a compensation coefficient from a compensation model established based on the heat exchange model of the raw material barrel, which is used to correct the value of the temperature difference gradient to eliminate the measurement bias introduced by the static tilt.

[0016] Preferably, the edge processing unit is further configured to: periodically enter a self-diagnosis mode, in which, by executing a preset high-computational load program, it generates a standardized internal heat pulse, which is transmitted to the wall of the raw material barrel through the structure of the micro wireless node; and captures the transient response curve of the differential heat flow sensing unit to the internal heat pulse, and converts the thermal response parameters of the transient response curve into the thermal response parameters of the internal heat pulse. With a stored reference thermal response parameter A comparison is performed to determine the health status of the differential heat flow sensing unit. When the comparison result indicates an abnormality, a sensor health alarm signal is sent. The thermal response parameter R is calculated as follows: in, is the peak temperature difference of the transient response curve, The equivalent instantaneous power consumption of the edge processing unit when a high computational load program is running.

[0017] Preferably, the edge processing unit is further configured to: periodically execute a preset high-computational load program so that it generates a standardized internal thermal pulse and uses it as an active detection signal; capture the transient response curve of the differential heat flow sensing unit to the internal thermal pulse, and analyze the peak height, time to reach the peak, and decay half-cycle duration of the transient response curve to invert the thermal fingerprint that characterizes the equivalent thermal resistance of the peroxide raw material and its packaging; and generate a dynamic calibration factor based on the difference between the thermal fingerprint and a reference fingerprint measured when the micro wireless node leaves the factory, which is used to correct the calculation of the degradation time series integral value.

[0018] Preferably, the micro wireless node uses low-power wide area network technology, such as LoRaWAN technology or NB-IoT technology for data communication, and is powered by a button battery to achieve a maintenance-free operating time of more than three years.

[0019] Compared with the prior art, the present invention has the following beneficial effects:

[0020] 1. The differential heat flow sensing unit captures the temperature gradient between the barrel wall and the environment. Combined with the posture sensing unit's precise judgment of physical steady state, the edge processing unit can perform time-series integration of the temperature difference signal while the raw material barrel is stationary. This mechanism, which converts the weak heat generation effect of chemical decomposition into a cumulative quantifiable value, enables continuous tracking of the cumulative degradation degree of peroxide raw materials throughout their life cycle, fundamentally changing the passive mode of traditional discrete sampling or single-point alarms, and providing dynamic process data support for risk management. The linkage mechanism between the posture sensing unit and the edge processing unit automatically pauses or downgrades the integral calculation when the physical motion state is identified, effectively isolating the noise introduced by handling vibration. At the same time, through low-pass filtering of the temperature difference signal, the high-frequency fluctuations of environmental thermal disturbances are removed, retaining only the low-frequency trend signal of the raw material's own decomposition heat generation. These two mechanisms collaboratively construct a signal discrimination system from the dimensions of mechanical disturbance and thermal noise, respectively, to ensure that the DTI integral value always reflects the true chemical degradation process.

[0021] 2. The edge processing unit generates standardized excitation through periodic self-heating pulses and uses the transient response curve of the differential heat flow sensing unit to achieve dual self-verification. On the one hand, the sensor health diagnosis is completed by comparing the response parameters with the reference values ​​to avoid the risk of silent failure. On the other hand, the changes in the thermal parameters of the raw materials are inverted by analyzing the morphological characteristics of the response curve, and the DTI calculation model is dynamically calibrated. This mechanism of converting processor heating defects into self-diagnostic resources gives the system the inherent ability to continuously combat component aging and the evolution of material physical properties.

[0022] 3. When the posture sensing unit determines that the raw material barrel is in a static tilt, the edge processing unit analyzes the gravity vector and calls the heat exchange compensation model to dynamically correct the temperature gradient measurement value. This mechanism extends the perception dimension of the same accelerometer from the binary judgment of motion / stillness to static tilt angle analysis, avoiding the heat transfer model offset caused by non-ideal stacking. It enables the core monitoring function to maintain measurement consistency in the posture differences in the real warehouse environment, significantly reducing the deployment accuracy requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of data sampling and degradation integration processing of the peroxide raw material quality intelligent control system of the present invention;

[0024] Figure 2 A comparison curve diagram of the degradation time series integral value changing with time in the present invention;

[0025] Figure 3 Schematic diagram of temperature difference gradient acquisition and signal processing of the differential heat flow sensing unit of the present invention.

[0026] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0027] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0028] An embodiment of the present invention provides an intelligent peroxide raw material quality control system, whose overall architecture is based on a set of micro wireless nodes deployed on a single raw material barrel. The node is highly integrated with a differential heat flow sensing unit responsible for core data acquisition, a posture sensing unit responsible for physical state perception, and an edge processing unit serving as the local data processing and decision-making center; these units work together in function to convert the collected raw physical quantity data stream into structured information that can characterize the degree of risk accumulation in real time at the node, and ultimately report key decision results to an external management system through low-power wide area network technology.

[0029] In view of the inherent perception blind spot of the traditional fixed-point temperature alarm mechanism for the long-term weak decomposition process of raw materials at subcritical temperature, the monitoring mechanism of the present invention is conceived as a quantitative tracking of the cumulative effect. Its specific procedures begin with the continuous measurement of the differential heat flow sensing unit. The unit continuously obtains the temperature difference gradient between a first thermistor tightly attached to the outer wall of the raw material barrel and a second thermistor exposed to the ambient air around the barrel through a first thermistor. This gradient value accurately reflects the real-time heat flow exchange intensity between the barrel wall and the environment; at the same time, the posture sensing unit continuously monitors the physical movement state of the raw material barrel and transmits this state information to the real-time monitoring unit. It is synchronized to the edge processing unit in real time. When and only when the output of the posture sensing unit clearly indicates that the raw material barrel is in a physical steady state, the edge processing unit starts a time accumulation calculation process. It performs weighted integration on the collected temperature difference gradient time series data based on a nonlinear weight function pre-set for the thermal decomposition kinetics of the target peroxide, thereby generating a quantitative indicator that can dynamically characterize the degree of cumulative degradation of the raw material, namely the degradation time series integral value; it is through this strategy of cleverly converting instantaneous weak heat flow signals into cumulative values ​​in the time dimension that the system is able to achieve continuous quantitative monitoring of slow degradation processes that traditional methods cannot effectively capture.

[0030] In order to deal with the combined pollution of mechanical vibration and environmental thermal disturbance to the monitoring signal caused by activities such as cargo handling and equipment start-up and shutdown in the storage environment, the system has built-in a set of efficient multi-source interference collaborative suppression mechanisms. First, at the mechanical disturbance suppression level, the posture sensing unit is specifically implemented as a micro-electromechanical system accelerometer. The edge processing unit accurately determines whether the raw material barrel is in a physical motion state caused by handling or collision by real-time analysis of the spectral characteristics or drastic changes in the statistical variance of the accelerometer output data. Once it is determined to be in a motion state, the edge processing unit immediately executes the preset avoidance logic, that is, completely suspends the cumulative calculation of the degradation time series integral value, or as an alternative, significantly reduces the weight of the current temperature difference gradient data point in the integral calculation, so as to ensure that the instantaneous and drastic heat exchange changes caused by mechanical vibration are fundamentally eliminated. It will not be mistakenly included in the cumulative value representing chemical decomposition; secondly, at the level of environmental thermal noise filtering, the edge processing unit enforces a digital low-pass filtering algorithm on the received raw temperature difference gradient time series data before performing the integral calculation. The cutoff frequency of this algorithm is carefully set to an extremely low value based on the inherent time constant of the peroxide's own chemical decomposition and heat generation process. Its purpose is to accurately remove temperature fluctuations caused by external high-frequency or medium-frequency factors such as ventilation, sunlight, and people walking, and only allow the ultra-low-frequency trend signal that represents the slow decomposition and heat generation of the raw material itself to pass through, and specifically use this highly purified low-frequency trend signal to calculate the degradation time series integral value; through the close coordination of the signal processing mechanisms in the above-mentioned mechanical and thermal dimensions, the system is able to extract the core chemical degradation process signal with high fidelity from the complex noise background.

[0031] Taking into account that during long-term deployment, the performance drift and aging of the sensor itself, as well as the potential evolution of the physical properties of the raw material packaging, may cause the measurement model to gradually become invalid, the system further integrates a set of self-diagnosis and self-calibration closed-loop systems based on active detection technology. Specifically, the edge processing unit is configured to automatically enter the self-diagnosis mode periodically. In this mode, it cyclically executes an internal preset high-computational load program to enable its microprocessor to generate a standardized internal heat pulse with known power and stable waveform. As an active detection signal, this heat pulse is transmitted to the wall of the raw material barrel through the physical structure of the node, and its complete transient response curve is captured by the differential heat flow sensing unit; the edge processing unit then performs an in-depth analysis of the curve. On the one hand, it calculates the temperature difference peak of the response curve Equivalent instantaneous power consumption of the processor program running The ratio of the thermal response parameter is obtained. and compare it with a reference thermal response parameter that is calibrated at the factory and stored locally. A precise comparison is performed. If the deviation between the two exceeds the preset tolerance range, it is immediately determined that the heat transfer path or sensitive element of the differential heat flow sensing unit has a malfunction, and the sensor health alarm signal is actively sent, thereby proactively solving the risk of silent failure of the sensor; on the other hand, it analyzes the overall morphological characteristics of the transient response curve more comprehensively, including but not limited to peak height, time to peak value and decay half-cycle duration, and reversely deduces a thermal fingerprint that can accurately characterize the equivalent thermal resistance and thermal capacity of the current peroxide raw material, packaging and installation interface, and compares it with the reference fingerprint measured at the factory. According to the difference between the two, a dynamic calibration factor is generated in real time for online correction of the calculation model of the subsequent degradation time series integral value, so as to dynamically compensate for the system thermal parameter drift caused by material aging or environmental factors, in order to adapt to the irregular stacking of raw material barrels in real storage environments. To solve the static tilt problem that may be caused by the fan, the system is also equipped with a set of attitude adaptive compensation mechanism. When the attitude sensing unit confirms that the raw material barrel is in a physical steady state, the edge processing unit will further analyze the static output data of the MEMS accelerometer to accurately calculate the projection of the gravity vector in the sensor coordinate system, thereby determining the current specific static tilt angle and direction of the raw material barrel; then, the edge processing unit uses this angle and direction information to perform a quick query or interpolation calculation in a pre-loaded compensation model established based on the standard heat exchange model of the raw material barrel to obtain a corresponding compensation coefficient, and use this coefficient to make an instant correction to the currently measured temperature difference gradient value. This is intended to systematically eliminate the measurement bias introduced by the tilt of the barrel changing the natural convection pattern between its surface and the surrounding air, thereby ensuring the high consistency and accuracy of the monitoring data under various non-ideal deployment postures.

[0032] Finally, when the accumulated degradation time series integral value obtained after the above-mentioned filtering, calibration and compensation exceeds the safety threshold configured for the specific peroxide raw material, the edge processing unit will use its integrated communication module and low-power wide area network technologies such as LoRaWAN or NB-IoT to send a high-priority early warning signal to the external management system. The entire system is powered by a button battery, and its highly optimized low-power design with coordinated software and hardware can ensure more than three years of maintenance-free and reliable operation time; the nonlinear weight function for specific types of peroxide raw materials in the system , safety threshold of degradation time series integral value and dynamic calibration factors The determination procedure is as follows: The cutoff frequency of the digital low-pass filter algorithm in this system With nonlinear weight function The determination procedure begins with the spectrum analysis of the original temperature gradient time series data collected in the accelerated aging experiment, which contains the real degradation trend and environmental noise; by analyzing the fast Fourier transform results of the data, the lower limit of the noise signal frequency corresponding to the environmental thermal disturbance and mechanical vibration is identified. and set the cutoff frequency Set to one tenth of the lower limit, that is, The low-frequency trend signal obtained is consistent with the degradation rate measured by gas chromatograph. Construct a data pair set; then, the data pair set is fitted using the least squares method with constraints to generate a third-order polynomial weight function , where the constraint condition is that within the effective operating temperature difference range of the system, the function must also satisfy the non-negativity, i.e. And the monotone non-decreasing property is that its first-order derivative , thus uniquely determining the coefficient ; For the judgment of the physical motion state of the raw material barrel, the system can be configured with two mutually exclusive integral processing logics, one is integration suspension, the other is weighted integration. When the weighted integration logic is used, the edge processing unit will use a locally stored fixed motion impact factor with a value between 0 and 1. Multiply the integral calculation of the time step to correct the degradation time series integral value; and the system's self-calibration function, its dynamic calibration factor The benchmark value is the transient response temperature difference peak value obtained by triggering a field baseline calibration procedure after the node is successfully installed in the raw material barrel and passes the installation quality verification. , the peak temperature difference measured in real time during subsequent periodic self-diagnosis The comparison will be made with this on-site installation benchmark value, i.e. , which standardizes the thermal response parameters at the factory It is only used to determine the health status of the sensor, and the dynamic calibration is tied to the actual heat conduction path after installation, so as to achieve accurate compensation for sensor aging or changes in the barrel wall interface. In addition, the compensation model used to correct the measurement bias introduced by static tilt is constructed by fixing a raw material barrel with a built-in 2W constant power heat source on a dual-axis goniometer. In a windless environment at 25°C, the goniometer is allowed to traverse all angle combinations of the pitch axis and azimuth axis within the range of -90 degrees to +90 degrees in 5-degree steps. , at each attitude point, wait for the temperature gradient The stable value is recorded when the fluctuation is less than 0.005°C within 60 seconds, thereby generating a two-dimensional lookup table. , the compensation coefficient in actual use is The system's operating time of more than three years is based on its power consumption model, which uses a 1000mAh CR2477 battery and defines the operating current and duration of each state of the system as: deep sleep current 2.5μA, data acquisition current 4mA for 60ms, local calculation current 6mA for 30ms, LoRaWAN communication current 35mA for 1.2s, high-load self-diagnosis current 12mA for 4s, and the execution cycles of each state are continuous, every 10 minutes, every 10 minutes, every 24 hours and every 30 days. The average daily power consumption is calculated as the sum of the power consumption of each state multiplied by its average daily execution times. The result is less than 0.22mAh, so the calculated total operating life is more than three years.

[0033] Example 1: In a large, densely stacked chemical warehouse, several batches of peroxide raw material barrels from different suppliers are stored. The warehouse's operating environment is extremely complex. The frequent movement of handling equipment such as forklifts causes continuous, random mechanical vibrations. At the same time, the opening and closing of the large warehouse doors causes violent and irregular ambient air temperature fluctuations. Against this backdrop, a batch of newly entered peroxide raw materials, due to slight differences in their formulation, exhibits a characteristic of continuous and slow decomposition at room temperature, far below traditional alarm temperatures. This micro-heat generation process is completely imperceptible to traditional monitoring systems that rely on fixed temperature threshold alarms, posing a potential and difficult-to-quantify safety hazard.

[0034] After the intelligent management and control system was deployed on this batch of raw material barrels, the raw temperature gradient data collected during an initial monitoring period of several weeks showed highly chaotic and noisy characteristics, including a large amount of vibration noise caused by the passage of forklifts and high-frequency thermal disturbances introduced by changes in ambient airflow. Under this operating condition, the preset collaborative mechanism between the system's posture sensing unit and edge processing unit began to operate. The posture sensing unit accurately identified each physical motion state caused by physical handling and instructed the edge processing unit to pause time accumulation calculation during these periods. This provided a pure time window free of mechanical vibration interference for subsequent signal processing. It was during these selected physical steady-state periods that the digital low-pass filtering algorithm executed by the edge processing unit was able to operate under optimal conditions, effectively separating the extremely weak low-frequency trend signal generated by the decomposition of the raw materials themselves from the mixed thermal signals.

[0035] As time went on, although the instantaneous barrel wall temperature of this batch of raw material barrels remained normal, the corresponding degradation time series integral value began to show a clear trajectory of slow but continuous unidirectional growth. This cumulative value, as the result of integrating the weak low-frequency trend signal in the time dimension, objectively and quantitatively revealed the irreversible cumulative chemical degradation process occurring inside the batch of raw materials. When the integral value finally reached the preset safety threshold, the system automatically sent an early warning signal to the external management system, allowing management personnel to isolate and prioritize the high-risk batch long before reaching the critical point of thermal runaway, thereby avoiding a potential serious safety accident. This process shows that the operation of this technical solution transforms the core issue of safety monitoring from the passive and noise-susceptible instantaneous state judgment of whether the raw material barrel is currently overheated to the quantifiable and predictable process management issue of how much degradation the raw material barrel has accumulated since the beginning.

[0036] Example 2: To conduct this verification, the test platform was built in a constant temperature environmental chamber and equipped with two groups of identical standard peroxide raw material barrels, marked as test group A and test group B. Both groups of raw material barrels were installed with intelligent management and control systems, and the systems were calibrated with the same benchmark. Among them, test group A served as the control group and maintained absolute physical steady state and stable ambient temperature throughout the test to characterize the natural deterioration process under ideal storage conditions; test group B served as the verification group and was subjected to a series of programmed external disturbances designed to simulate real storage conditions, including periodic application of wide-spectrum mechanical vibrations through a vibration table rigidly connected to the bottom of the raw material barrel, and intermittent irradiation of the barrel wall through a programmable power infrared heat source. For radioactive heating, the setting of key parameters in the experiment, such as the sampling period of the differential heat flow sensing unit in the system, lies in balancing the real-time performance of data acquisition with the energy consumption of the system. Given that the heat generated by the decomposition of peroxide itself is a slow-changing process at the level of hours, while the environmental thermal disturbance may be a fast-changing process at the level of minutes, in order to ensure that no signal aliasing occurs according to the Nyquist sampling theorem and to avoid unnecessary data redundancy, the sampling period is set to ten seconds in this experiment. Similarly, the cutoff frequency of the digital low-pass filtering algorithm in the edge processing unit is set based on the fact that the characteristic frequency of the heat generated by the decomposition of peroxide is much lower than one hertz, while the environmental thermal disturbance is usually at a higher frequency band. Therefore, the cutoff frequency is set to a value that can effectively separate the two to ensure the quality of the signal ultimately used for the integral calculation.

[0037] The test ran for 72 hours after initiation, during which system data was continuously recorded. Data from Test Group A showed that its raw temperature gradient fluctuated steadily within a very small range, and the corresponding low-frequency trend signal closely matched this fluctuation. Its degradation time-series integral value exhibited a smooth, continuous upward curve with a very low slope. The raw temperature gradient data from Test Group B exhibited large peak fluctuations synchronized with the external disturbance period. However, the low-frequency trend signal, obtained after filtering by the edge processing unit, successfully stripped away these high-frequency disturbances, and its variation curve was essentially parallel to that of Test Group A. Calculations of the integral value further revealed that whenever the vibration table was activated and the attitude sensing unit determined that the system was in physical motion, the integral accumulation process for Test Group B automatically paused and did not resume until vibration ceased, resulting in a step-like growth pattern. Table 1 summarizes the comparative data records at some key moments.

[0038] Table 1: Data comparison table.

[0039]

[0040] As shown in Table 1, at the 24th and 36th hours, the original temperature difference gradient of test group B increased significantly due to external disturbances, but its low-frequency trend signal and degradation time series integral value remained highly consistent with those of the undisturbed control group, test group A. This data shows that there is a synergy between the system's built-in digital low-pass filtering algorithm and the integral gating logic based on attitude sensing, which can effectively separate the core signal representing true chemical degradation from the high-intensity noise from the mechanical and thermal dimensions.

[0041] Example 3: This example combines Figures 1 to 3 , the realization of an intelligent quality control system for peroxide raw materials is described, such as Figure 1 As shown, Figure 1 The data sampling and degradation integration processing flow of the intelligent peroxide raw material quality control system of the present invention is demonstrated, which includes the collaborative interaction mechanism of four functional modules: differential heat flow sensing unit, attitude sensing unit, edge processing unit and external management system. The differential heat flow sensing unit is responsible for continuously outputting temperature gradient data with a sampling period of every 10 seconds ( ), the attitude sensing unit synchronously outputs acceleration data, and the two are transmitted to the edge processing unit together. The edge processing unit first judges the physical state based on the acceleration data. When it is confirmed that the raw material barrel is in a physically static state, it enters the degradation judgment process, executes the low-pass filtering algorithm in sequence to filter out high-frequency thermal noise, extracts the low-frequency trend signal, and then applies the nonlinear weight function to weight the trend signal and update the degradation time integral value (DTI); if the obtained DTI is greater than the safety threshold, the system triggers the sending of an early warning signal (LoRaWAN / NB-IoT) to the external management system. After the external management system confirms the receipt of the signal, the alarm process is completed; on the contrary, if the edge processing unit judges that the output data of the attitude sensing unit indicates that the raw material barrel is in a physical motion state, the integral calculation is suspended to avoid the interference caused by transportation being erroneously included in the DTI value calculation. The entire process is strictly cycled according to the time period to achieve continuous quantitative tracking and real-time early warning management of the peroxide raw material degradation process.

[0042] like Figure 2 As shown, the vertical axis is the degradation time series integral value (unit), and the horizontal axis is time (hour). A safety threshold dashed line is set in the figure as the degradation risk judgment standard line. The figure contains two integral curves of test group A (no disturbance) and test group B (with disturbance). The former represents the integral trend without the influence of external disturbance, and the latter reflects the effect of the anti-interference mechanism of the system of the present invention. The vibration period, thermal radiation period and compound disturbance period marked in the figure respectively represent the mechanical vibration, infrared thermal radiation and compound disturbance applied to the test group B in different time periods. In these intervals, multiple integral pause marks appear on the curve of test group B (with disturbance), indicating that the system passes the attitude sensing unit It cooperates with the edge processing unit to determine whether it is in a disturbance state, thereby automatically executing the integral gating logic to suspend the calculation of the degradation time series integral value, avoiding the mistaken counting of high-frequency disturbances as heat generation caused by the degradation of the raw materials themselves; in contrast, test group A (no disturbance) continues to integrate throughout the entire process, and the curve shows a smooth upward trend. Although the curve of test group B (with disturbance) in the figure has experienced strong external interference, its integral value still grows steadily because the system accurately implements the integration suspension, and is always lower than that of test group A (no disturbance), which effectively proves that the intelligent management and control system of the present invention has excellent resistance to mechanical vibration and thermal disturbance, ensuring that the degradation time series integral value truly reflects the chemical degradation process of the raw materials, rather than being contaminated by noise.

[0043] like Figure 3As shown, the left side of the figure is the internal structure of the storage barrel containing peroxide raw materials. The raw materials may decompose and generate heat during long-term storage. The heat is conducted outward through the barrel wall, thereby forming a local heat flow. A first thermistor is set close to the outer wall of the barrel (close to the barrel wall) for real-time collection of the barrel wall temperature T1. At the same time, a second thermistor is arranged in the surrounding air (exposed to the environment) for collecting the ambient temperature T2. The signal processing part first performs a temperature difference calculation ΔT=T1-T2 to obtain the current temperature difference gradient between the barrel wall and the environment, and then outputs a temperature difference gradient output signal and forms a temperature difference gradient time series curve for subsequent degradation trend extraction and integral analysis. This structure and processing logic constitute the most basic heat flow perception module in the system, which can accurately convert the heat flow changes caused by the decomposition and heat generation of peroxide into a structured temperature difference gradient time series curve signal, thereby realizing the continuous and quantitative tracking of the slow degradation behavior of the peroxide raw material in the present invention; in terms of, Figure 3 The curve in the lower right corner shows that the temperature gradient time series curve output by the module conceptually contains two levels: one is the original signal superimposed with the ambient thermal noise (the curve with larger fluctuations in the figure); the other is the core trend representing the slow increase in temperature difference due to the continuous weak heat release of peroxide due to its own chemical decomposition. The smoothly rising bold curve in the figure indicates its growth direction.

[0044] Example 4: In this example, when the intelligent management and control system is applied to a new type of peroxide raw material, the decision-making models and thresholds in the system, including the nonlinear weight function representing the cumulative degree of degradation, the compensation model for correcting posture deviation, and the safety threshold for triggering the warning, must be calibrated through a set of experimental procedures to ensure the accuracy of subsequent applications. This calibration work is carried out in the laboratory. To determine the safety threshold of the degradation time series integral value, a sample of the new type of peroxide raw material is placed in a temperature-controlled reactor, and the intelligent management and control system is deployed outside it. The accelerated aging test is initiated by raising the ambient temperature to a constant high temperature that can induce accelerated degradation. During the test, the management and control system continuously records the growth of the degradation time series integral value and periodically detects the attenuation of key active components in the raw material through gas chromatography. When the chemical analysis results confirm that the degree of degradation of the raw material has reached a preset critical point, the test is terminated and the final reading of the degradation time series integral value recorded by the management and control system is read. This reading is defined as the safety threshold for this type of raw material.

[0045] By comparing the instantaneous values ​​of the temperature gradient at different times with the actual degradation rate measured by chemical analysis during the same accelerated aging test, a set of discrete data points describing the relationship between heat generation intensity and degradation rate can be obtained. The nonlinear weight function is the empirical curve obtained by fitting this set of data points. It reflects the temperature-dependent characteristics of the heat generation effect of this specific raw material and provides a weight basis for the integral calculation. At the same time, to construct a compensation model for correcting static tilt, another control system is fixed to a standard raw material barrel filled with an inert medium and maintained at a constant temperature slightly higher than the ambient temperature. This raw material barrel is then placed on a three-axis pan-tilt stage. The pan-tilt stage is controlled by a program to move the raw material barrel through multiple static tilt angles and directions from horizontal to vertical. At each posture, the system is allowed to rest for a sufficient time for the heat flow to stabilize, and the offset of the temperature gradient measurement caused by this posture is recorded. The data set of all posture points and their corresponding offsets constitutes a three-dimensional lookup table, which is the compensation model.

[0046] After completing the above-mentioned offline parameter calibration and model construction, the system is run in self-diagnosis mode under the standard experimental conditions. A standardized internal thermal pulse is generated by running a high-computational load program, and the transient response curve at this moment is captured. The characteristic parameters of the curve, especially the temperature difference peak, are used as part of the benchmark thermal fingerprint and compared with the above-mentioned calibrated thermal response parameters. The thermal fingerprint acquired in real time is compared with the stored reference fingerprint. A dynamic calibration factor is calculated by the formula Calculate, where is the peak temperature difference of real-time response, The factor is used to calculate the benchmark temperature difference peak value in real time to correct the degradation time series integral value to compensate for the change in heat transfer efficiency caused by sensor aging or barrel wall contamination. This completes a configuration process from offline calibration to online self-adaptation.

[0047] Example 5: This example describes a standardized debugging and baseline establishment procedure for the intelligent management and control system during the on-site deployment stage. After a set of intelligent management and control system nodes that have been offline calibrated are installed in a raw material barrel on site, an installation quality verification procedure must be performed to ensure effective heat conduction between the sensor unit and the barrel wall. After the program is started, the edge processing unit enters a one-time debugging mode. In this mode, the system executes a preset high-computational load program that is the same as the self-diagnosis mode to generate a standardized internal thermal pulse; the edge processing unit captures and analyzes the complete transient response curve of the differential heat flow sensing unit to the pulse, and compares its morphological characteristics, especially the height of the temperature difference peak and the time to reach the peak, with the benchmark thermal fingerprint measured and stored for the node under standard thermal conditions before leaving the factory. If the difference between the two exceeds the preset tolerance, the system determines that the current installation has poor heat conduction and sends an installation failure signal to the installer's terminal device, prompting them to readjust the node position or tightening status.

[0048] After the installation quality verification is passed, the debugging procedure enters the environmental noise adaptive calibration stage to set the physical motion state discrimination threshold that conforms to the current storage environment. In this stage, the edge processing unit continuously collects the output data of the posture sensor unit within a preset time window of fifteen minutes. At this time, the storage environment is in its normal operating state, covering background ground vibration and equipment operation noise; the edge processing unit performs statistical analysis on the acceleration data stream collected within the time window, calculates the mean and standard deviation of the output data variance, and automatically generates a threshold for judging physical motion based on this statistical result. The dynamic threshold of the state is set as the sum of the mean and three standard deviations of the data variance within the statistical period. This procedure enables the system to reliably distinguish the violent vibration caused by close-range cargo handling from the background environmental noise, providing a decision-making basis for the precise calculation of the subsequent degradation time series integral value; the identification of the physical motion state adopts a threshold judgment mechanism based on the variance of the acceleration time series. After the system is deployed, it enters a 15-minute initialization monitoring period. During this period, the attitude sensor unit continuously collects three-axis acceleration data at a frequency of 100Hz, and the edge processing unit calculates the variance value of the acceleration module every 10 seconds. , forming 90 groups of initial sample sets {V1, V2, ..., V 90}, then perform distribution analysis on the sample set and take its sample mean and standard deviation , calculate the physical motion determination threshold ; During the operation period, before entering each integration period, the system calculates the acceleration modulus variance in real time within the last 10 seconds ,like > , it is determined that the raw material barrel is in a physical motion state and the cycle integration operation is suspended. Otherwise, the time series integration operation is continued. The threshold It is written into the non-volatile memory after initialization and remains unchanged during the operation of the entire machine without adaptive update to avoid the impact of long-term drift on the consistency of judgment.

[0049] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0050] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An intelligent peroxide raw material quality control system, characterized in that: include: A differential heat flow sensing unit is provided, which is configured to continuously measure the temperature gradient between the wall temperature of the raw material barrel and the ambient air temperature around the raw material barrel; A posture sensing unit is provided, and the posture sensing unit is configured to monitor the physical motion state of the raw material barrel; An edge processing unit is configured to: when the posture sensing unit indicates that the raw material barrel is in a physical steady state, the edge processing unit is configured to: accumulate the temperature gradient over time based on a nonlinear weight function pre-set for the peroxide raw material, and calculate a degradation time series integral value that represents the degree of cumulative degradation of the peroxide in the raw material barrel; when the degradation time series integral value exceeds the safety threshold configured for the peroxide raw material, an early warning signal is sent to the external management system; The edge processing unit is further configured to: pause time accumulation when the posture sensing unit indicates that the raw material barrel is in a physical motion state, or reduce the weight of the temperature difference gradient in the cumulative calculation when the posture sensing unit indicates that the raw material barrel is in a physical motion state; perform a digital low-pass filtering algorithm on the time series data of the temperature difference gradient to obtain a low-frequency trend signal generated by the heat generated by the decomposition of the peroxide raw material itself; and only use the low-frequency trend signal to calculate the degradation time series integral value; when the posture sensing unit indicates that the raw material barrel is in a physical steady state, analyze the output data of the posture sensing unit to determine the static tilt angle and direction of the raw material barrel; and according to the static tilt angle and direction, query and obtain a compensation coefficient from a compensation model established based on the raw material barrel heat exchange model to correct the value of the temperature difference gradient.

2. The peroxide raw material quality intelligent control system according to claim 1, characterized in that: The differential heat flow sensing unit includes a first thermistor close to the wall of the raw material barrel and a second thermistor exposed to the ambient air around the raw material barrel. The temperature gradient is the difference between the measured temperature of the first thermistor and the measured temperature of the second thermistor.

3. The peroxide raw material quality intelligent control system according to claim 1, characterized in that: The posture sensing unit is a micro-electromechanical system accelerometer, and the basis for judging the physical motion state is the frequency spectrum characteristics or variance changes of the output data of the micro-electromechanical system accelerometer.

4. The peroxide raw material quality intelligent control system according to claim 1, characterized in that: The edge processing unit is also configured to periodically enter a self-diagnosis mode, in which it generates a standardized internal heat pulse by executing a preset high-computational load program. The internal heat pulse is transmitted to the wall of the raw material barrel through the structure of the micro wireless node; and the transient response curve of the differential heat flow sensing unit to the internal heat pulse is captured, and the thermal response parameters of the transient response curve are converted into With a stored reference thermal response parameter A comparison is performed to determine the health status of the differential heat flow sensing unit. When the comparison result indicates an abnormality, a sensor health alarm signal is sent. The thermal response parameter R is calculated as follows: in, is the peak temperature difference of the transient response curve, The equivalent instantaneous power consumption of the edge processing unit when a high computational load program is running.

5. The peroxide raw material quality intelligent control system according to claim 1, characterized in that: The edge processing unit is further configured to: periodically execute a preset high-computational load program to generate a standardized internal thermal pulse as an active detection signal; capture the transient response curve of the differential heat flow sensing unit to the internal thermal pulse, and analyze the peak height, time to peak, and decay half-cycle of the transient response curve to infer a thermal fingerprint that characterizes the equivalent thermal resistance of the peroxide raw material and its packaging; According to the difference between the thermal fingerprint and a reference fingerprint measured when the micro wireless node leaves the factory, a dynamic calibration factor is generated to correct the calculation of the degradation time series integral value.

6. The peroxide raw material quality intelligent control system according to claim 1, characterized in that: The micro wireless node uses low-power wide area network technology, including LoRaWAN technology or NB-IoT technology for data communication, and is powered by a button battery.

Images