Medical compressed air on-demand regeneration energy-saving method based on dew point feedback

Abstract

The invention provides an energy-saving method for on-demand regeneration of medical compressed air based on dew point feedback, relates to the technical field of program control systems, and successfully solves a comprehensive technical problem of how to ensure the safety of medical compressed air on the premise of ensuring absolute safety of medical compressed air by constructing and applying a core quantitative index of tower body fatigue accumulation degree. And the short-term operation energy efficiency and long-term equipment health of the compressed air dryer are synergistically optimized. According to the invention, the method achieves the jump from passive response to predictive protection and active repair, remarkably improves the reliability and toughness of a system under a dynamic load, and further reduces the operation and maintenance cost of a whole life cycle through prolonging the effective service life of equipment and reducing non-planned shutdown.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of program control system, and particularly to a medical compressed air on-demand regeneration energy-saving method based on dew point feedback. BACKGROUND

[0002] In the modern industry and medical field, the intelligent operation and maintenance of key infrastructure has become the core development direction to improve system reliability, energy efficiency and safety. Especially in the fluid supply system with high requirements for continuity and stability, the control strategy is evolving from the traditional open-loop control based on fixed parameters to the closed-loop, adaptive and even predictive control based on multi-dimensional real-time data. The goal of this evolution is to enable the system not only to respond to the current state, but also to foresee future risks and actively optimize its long-term health status, thereby maximizing the efficiency throughout the life cycle. However, in the specific technical field of medical compressed air dryers, the existing technology still faces the following challenges in the process of advancing towards the above-mentioned intelligent goal:

[0003] Passivity and hysteresis of control logic: Although the existing improvement scheme introduces a dew point feedback mechanism, its essence is still a "post-response" mode. That is, the control system only starts compensation or switching action after monitoring that the outlet air quality (such as dew point) has begun to deteriorate. In the face of instantaneous and sharply increasing gas load shock, this control hysteresis may cause the outlet air quality to deviate from the medical standard for a short time before the system fully responds, posing potential safety hazards to critical application scenarios such as life support.

[0004] Single and one-sided decision dimension: The decision basis of the existing control strategy is usually limited to a single process parameter, such as outlet dew point or fixed operating time. This method fails to take into account the evolution of the device's own health status. It cannot distinguish between a "healthy" adsorption tower and a "fatigued" adsorption tower. Therefore, it may subject an adsorption tower that is already on the edge of performance decline to a high-intensity load shock, thereby accelerating its irreversible damage, leading to unexpected downtime and high maintenance costs.

[0005] Fragmentation and short-sightedness of device management: The existing technology lacks a unified management model that can run through the entire cycle of "damage assessment - protection - repair" of the device. The energy-saving control, safety protection and long-term maintenance strategies of the system are often independent of each other. For example, the energy-saving strategy may only focus on reducing regeneration energy consumption, while ignoring the long-term fatigue accumulation that some energy-saving operations may cause to the device. This fragmentation of management makes it impossible for the system to achieve a coordinated optimization between short-term operation safety and long-term asset health.

[0006] The above information disclosed in the BACKGROUND section merely to enhance the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0007] The purpose of the present application is to provide a dew point feedback-based medical compressed air on-demand regeneration energy-saving method to solve the problems raised in the background. To achieve the above purpose, the present application provides the following technical solutions:

[0008] A dew point feedback-based medical compressed air on-demand regeneration energy-saving method, the specific steps comprising:

[0009] S1: Based on the real-time pressure difference data of the inlet and outlet of the dryer, the impact characteristics representing the current gas load are extracted through time sequence analysis;

[0010] S2: Based on the historical performance data of the dryer operation, the current health status of each of the at least two adsorption towers is evaluated;

[0011] S3: Fusion of the impact characteristics and the current health status, dynamically calculating a quantitative index representing the tower body fatigue accumulation degree for each adsorption tower;

[0012] S4: Based on the quantitative index of the tower body fatigue accumulation degree, performing predictive work rotation between the at least two adsorption towers;

[0013] S5: Based on the quantitative index of the tower body fatigue accumulation degree, selecting and executing an adaptive regeneration program matched for the adsorption tower to be regenerated.

[0014] Compared with the prior art, the present application has the beneficial effects that: by creatively constructing a quantifiable "tower body fatigue accumulation degree" dynamic model, the model as a technical bridge, the time sequence characteristics representing the instantaneous gas load impact are deeply fused with the health status index representing the long-term performance evolution of the adsorption tower; realizing the fundamental change from single information decision to multi-information collaborative decision. The "tower body fatigue accumulation degree" drives "predictive protection" forward: before perceiving the occurrence of high damage risk load impact, it actively transfers the task to the healthier tower body, thereby solving the safety lag problem existing in the prior art due to passive response. On the other hand, it drives "compensatory repair" backward: after identifying that the tower body has accumulated excessive fatigue, it automatically triggers the repair regeneration program matched therewith, making up for the lack of long-term health status management of the equipment in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The present application is a whole method flowchart;

[0016] Figure 2 Logic diagram for performing step S2 of the present application;

[0017] Figure 3 Logic diagram for performing step S3 of the present application;

[0018] Figure 4 Logic diagram for performing step S4 of the present application;

[0019] Figure 5 Logic diagram for performing step S5 of the present application. DETAILED DESCRIPTION

[0020] In order to make the above objectives, features and advantages of the present application more apparent, specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced in other ways different from those described herein without departing from the scope of the present application, and those skilled in the art can make similar extensions without departing from the scope of the present application, and therefore the present application is not limited to the specific embodiments disclosed below.

[0021] Embodiment one: the execution subject of the following embodiment is a "medical compressed air on-demand regeneration energy-saving method based on dew point feedback", which is executed by a controller (such as PLC, embedded system or computer). Please refer to Figures 1 to 5 The present application provides a technical solution:

[0022] A medical compressed air on-demand regeneration energy-saving method based on dew point feedback, the method is applied to a medical compressed air dryer system including at least two adsorption towers, and the method includes the following steps:

[0023] S1: based on the real-time pressure difference data of the inlet and outlet of the dryer, the impact characteristics representing the current air consumption load are extracted through time sequence analysis;

[0024] S2: based on the historical performance data of the dryer operation, the current health status of each of the at least two adsorption towers is evaluated;

[0025] S3: combining the impact characteristics and the current health status, a quantitative index representing the tower fatigue accumulation degree of each adsorption tower is dynamically calculated;

[0026] S4: based on the quantitative index of the tower fatigue accumulation degree, a predictive work rotation is performed between the at least two adsorption towers;

[0027] S5: based on the quantitative index of the tower fatigue accumulation degree, an adaptive regeneration program matched with the adsorption tower to be regenerated is selected and executed.

[0028] Further illustrate, the step of extracting the impact characteristics in S1 specifically includes: performing statistical analysis on the real-time pressure difference data to obtain a kurtosis of the pressure difference signal for representing the pulse nature of the gas load intensity, and a main peak value of a pressure difference fluctuation frequency for representing the type and stability of the gas load.

[0029] Define the real-time pressure difference data sequence as P seq : This parameter is a one-dimensional numerical array, and its elements are the pressure difference values between the inlet and outlet of the dryer continuously collected by the controller within a preset time window T w with a sampling frequency f s . The specific acquisition approach is as follows: a high-precision and high-response-speed pressure transmitter is installed at the inlet and outlet pipes of the adsorption tower of the dryer; in this embodiment, a diffusion silicon pressure transmitter with a response time less than 1 millisecond is selected, and the analog input module (ADC) of the controller synchronously collects the pressure values of the two transmitters with a frequency f s = 100 Hz, and calculates the difference between them in real time, and a total of N = T w × f s difference points form a real-time pressure difference data sequence P seq . The total number of difference points represents the total number of data points, and in this embodiment, the preset time window T w is set to 2 seconds, so the total number of data points N = 200. Before forming the final real-time pressure difference data sequence P seq , a digital band-pass filter is applied to the original difference sequence to eliminate DC bias and power frequency interference, and a fourth-order Butterworth filter with a cutoff frequency of 0.5 Hz to 45 Hz is specifically selected; the filter model is designed through a standard signal processing library and its coefficients are fixed in the controller program. The signal processing library in this example is the SciPy library in Python; after the obtained original pressure difference sequence [10.1, 10.2, 10.1] is filtered and de-meaned, the real-time pressure difference data sequence P seq for calculation is [-0.05, 0.05, -0.04];

[0030] The kurtosis of the pressure difference signal is denoted as K p : This parameter is a dimensionless floating-point value, which is used to quantitatively describe the sharpness of the probability density distribution function of the real-time pressure difference data sequence P seq relative to the standard normal distribution. Its acquisition method is the kurtosis definition in statistics, specifically the fourth-order standardized moment of the sequence. A function is built-in in the controller program, and the input is the real-time pressure difference data sequence P seq , and the output is the kurtosis K p value of the pressure difference signal.

[0031] The main peak value of the pressure difference fluctuation frequency is denoted as F peakThe unit is Hertz (Hz), representing the time window T. w Internal, real-time differential pressure data sequence P seq The frequency of the most concentrated energy fluctuation is obtained by analyzing the real-time differential pressure data sequence P. seq Spectrum analysis was performed. The specific calculation model was the Fast Fourier Transform (FFT) algorithm, a standard method for calculating Discrete Fourier Transform. The controller called the FFT function to process the real-time differential pressure data sequence P containing 200 points. seq This yields an energy spectrum array containing 100 effective frequency points. The index corresponding to the maximum value is found by traversing this array, and then the index is used in conjunction with the sampling frequency f. s The relationship can be used to calculate the main peak value F of the differential pressure fluctuation frequency. peak If the FFT results show that the energy is highest at the 10th frequency point, then the main peak value of the differential pressure fluctuation frequency is F. peak The calculation is 10×(f) s / N)=10×(100 / 200)=5Hz. In this embodiment, the complete calculation process for extracting impact features is executed periodically by the controller, and the specific steps are as follows:

[0032] 1.1) Data Input: The initial input to the process is the real-time differential pressure data sequence P obtained after acquisition and preprocessing within the past 2 seconds. seq The sequence contains 200 floating-point numbers;

[0033] 1.2) Calculate the kurtosis of the differential pressure signal: First, calculate the real-time differential pressure data sequence P. seq The arithmetic mean of all data points is calculated. Then, the real-time differential pressure data sequence P is calculated. seq The standard deviation of all data points in the dataset. Then iterate through the real-time differential pressure data series P. seq Each difference data point p in i Let i represent the index of the difference data point. Calculate the sum of the fourth power of the differences between the difference data points and the mean, then divide by the total number of data points N to obtain the fourth-order central moment m4. The final kurtosis K of the pressure difference signal... p It is obtained by dividing the fourth central moment m4 by the fourth power of the standard deviation.

[0034] 1.3) Calculate the main peak value of the differential pressure fluctuation frequency: The first step is to process the real-time differential pressure data sequence P seq As input, the controller's built-in Fast Fourier Transform (FFT) function is invoked. The FFT function outputs a complex array representing the real-time differential pressure data sequence P. seq The distribution in the frequency domain is then determined. Next, the square of the modulus of each element in the complex array is calculated to obtain the energy spectrum array. Since the signal is real and the energy spectrum is symmetrical, only the difference points from the first half of the total number of data points N need to be taken. Then, the energy spectrum array is traversed to find the maximum value and its corresponding array index idx.max The final differential pressure fluctuation frequency main peak value F peak By multiplying the maximum energy index idx max By the frequency resolution f s / N.

[0035] The final output of this example process is a tuple or structure containing two elements: (kurtosis of differential pressure signal, differential pressure fluctuation frequency main peak value). The output of one calculation is (8.5, 5.0), which indicates that in the past 2 seconds, the system has monitored a load impact with a significant pulse with a main frequency of 5 Hz. The kurtosis of the differential pressure signal K p The differential pressure fluctuation frequency main peak value F peak Are calculated in parallel as two independent, orthogonal features, which together constitute the description of the "impact feature" and serve as input for subsequent steps. The kurtosis of the differential pressure signal K p Is sensitive to the amplitude jump of the signal, while the differential pressure fluctuation frequency main peak value F peak Is sensitive to the periodic behavior of the signal.

[0036] Further, the step of evaluating the current health state in S2 specifically includes: based on the temperature and humidity, flow data at the inlet of the adsorption tower and its working condition correction theoretical maximum adsorption capacity, calculating a single-tower adsorption efficiency index representing the degree of performance degradation. Let the actual water vapor adsorption total mass be M actual : unit: kilogram (kg), represents the total mass of water vapor actually adsorbed by the target adsorption tower from compressed air in a complete adsorption cycle. Its acquisition method is to integrate the intermediate parameter - instantaneous water vapor mass flow rate m w (t) in the entire adsorption cycle time T cycle ;

[0037] The instantaneous water vapor mass flow rate m w (t) is a floating point value that changes with time, with a unit of kilogram per second (kg / s), representing the mass flow rate of water vapor carried by the compressed air flowing through the inlet of the adsorption tower at time point t. Its acquisition method is to multiply the instantaneous compressed air volume flow rate q v (t) at the same time point t with the instantaneous water vapor density p v (t).

[0038] The instantaneous compressed air volume flow rate q v(t) is a time-varying floating-point value with the unit of cubic meters per second (m3 / s). Its specific acquisition approach is: a high-precision flow meter is installed at the total inlet pipeline of the dryer, preferably a vortex flow meter or a thermal mass flow meter with fast response speed and wide range ratio, and the output signal of the flow meter is continuously collected by the analog input module of the controller at a frequency synchronized with the system master clock.

[0039] Instantaneous water vapor density p v (t) is a time-varying floating-point value with the unit of kilograms per cubic meter (kg / m 3 ). Its acquisition method is to calculate it through a standard physical calculation model built-in the controller, i.e. the saturated water vapor density calculation formula. The input of the standard physical calculation model is the instantaneous inlet temperature T in (t) and the instantaneous inlet relative humidity RH in (t) at the same time point t, and the output is the water vapor density under the temperature and humidity. For example, at 25℃ and 100% RH, the water vapor density is about 0.023 kg / m 3 . The saturated water vapor density calculation formula in this example is calculated accurately by the Arden-Buck equation or the Goff-Gratch equation;

[0040] Instantaneous inlet temperature T in (t) is a time-varying floating-point value with the unit of Celsius (℃) or Kelvin (K). Its specific acquisition approach is: a PT100 platinum resistance temperature sensor is installed in the pipeline downstream of the flow meter and before the inlet of the adsorption tower, and its measurement value is continuously collected by the controller.

[0041] Instantaneous inlet relative humidity RH in (t) is a time-varying floating-point value expressed as a percentage (%). Its specific acquisition approach is: a pressure-resistant and oil-resistant capacitive film humidity sensor is installed side by side with the temperature sensor, and its measurement value is continuously collected by the controller.

[0042] Adsorption cycle length T cycle is a floating-point value with the unit of seconds (s), representing the total time experienced by the target adsorption tower from the start of adsorption to the switching to the regeneration state. The value is determined and recorded by the controller according to its running logic.

[0043] The working condition corrected theoretical maximum adsorption capacity is denoted as M max,c: This parameter is a floating-point value, with the unit of kilogram (kg), which is the core innovation point of the embodiment; it represents the maximum water vapor mass that a brand-new, unaged adsorption tower can theoretically adsorb under the average working condition of the current adsorption cycle. Its acquisition method is to calculate through an adsorption isotherm model for the used desiccant (such as activated alumina) pre-stored in the controller. The inputs of the model are the cycle average inlet temperature T avg and the cycle average inlet pressure P avg . The adsorption isotherm model includes but is not limited to the Freundlich or Langmuir model, and the model parameters are provided by the desiccant supplier or calibrated through experiments; the example of the embodiment is that M max,c is 5.0 kg at 20℃ and 7 bar pressure; and at 35℃ and 7 bar pressure, the value of M max,c will decrease to 3.8 kg due to high-temperature inhibition of adsorption.

[0044] The steps of calculating the working condition corrected theoretical maximum adsorption capacity include: calling the pre-stored adsorption isotherm model for the desiccant used in the adsorption tower in the controller, taking the cycle average inlet temperature and the cycle average inlet pressure as the inputs of the adsorption isotherm model; the adsorption isotherm model is used to represent the equilibrium adsorption capacity of the desiccant under different temperature and pressure conditions; the adsorption isotherm model is a Langmuir type model or a Freundlich type model; the theoretical maximum equilibrium adsorption capacity per unit mass of desiccant under the current working condition is determined through the adsorption isotherm model, and multiplied by the total mass of the desiccant in the adsorption tower to obtain the working condition corrected theoretical maximum adsorption capacity.

[0045] In a preferred embodiment, the adsorption isotherm model is used to describe the maximum equilibrium adsorption capacity of the desiccant for water vapor under different temperatures and pressures; specifically, the construction method of the model includes: performing equilibrium adsorption experiments on the used desiccant (such as activated alumina) under multiple temperature and pressure combination conditions, recording the equilibrium moisture content under different water vapor partial pressures, and fitting the experimental data into a Langmuir type or Freundlich type isothermal adsorption curve to obtain the model parameters of the desiccant. The Langmuir type model assumes that there are limited and uniform adsorption sites on the surface of the adsorbent, each site can only adsorb one water molecule at a time, and the adsorption energy of each site is the same, and an increase in temperature will cause the residence time of water molecules on the surface to be shortened, thereby reducing the maximum adsorption capacity; the Freundlich type model assumes that the adsorption site energy is not uniform, which can reflect the nonlinear relationship of the gradual stabilization of the adsorption capacity of porous materials under different temperatures and different partial pressures, and also reflects the adverse effect of temperature rise on adsorption capacity through the temperature-related parameter.

[0046] In actual operation, at the end of each adsorption cycle, the controller obtains the cycle average inlet temperature and cycle average inlet pressure of the current cycle and inputs them as input parameters into the adsorption isotherm model; the adsorption isotherm model first adjusts the saturated adsorption capacity curve of the desiccant based on the temperature, and then determines the theoretical maximum equilibrium adsorption amount per unit mass of the desiccant according to the water vapor partial pressure corresponding to the pressure; finally, the controller multiplies the unit mass maximum adsorption amount by the known total mass of the desiccant in the adsorption tower to obtain the current working condition corrected theoretical maximum adsorption capacity value.

[0047] cycle average inlet temperature T avg and cycle average inlet pressure P avg : These two parameters are the arithmetic means of the instantaneous inlet temperature T cycle (t) and the instantaneous inlet pressure P in (t) (the pressure sensor is installed in the same place) within the entire adsorption cycle time T in ; the instantaneous inlet pressure P in (t) of the example is obtained by installing the pressure sensor in the same installation position as the PT100 platinum resistance temperature sensor;

[0048] The single-tower adsorption efficiency index is denoted as η eff : This parameter is a dimensionless floating-point value between 0 and 1, which is the final output for evaluating the current health status of the adsorption tower. It is obtained by dividing the actual water vapor adsorption total mass M actual by the working condition corrected theoretical maximum adsorption capacity M max,c .

[0049] In this embodiment, the calculation process for evaluating the health status of a single adsorption tower is automatically executed once by the controller after the adsorption tower completes a complete adsorption cycle. Taking the evaluation of the adsorption tower as an example, the calculation process is as follows:

[0050] 2.1) The starting input is three time series arrays stored in the controller memory, which are recorded at a frequency of 1 Hz within the adsorption cycle time T cycle of the adsorption tower: the instantaneous inlet temperature array, the instantaneous inlet relative humidity array, and the instantaneous compressed air volume flow array;

[0051] The step of calculating the actual water vapor adsorption total mass M actual is to initialize an accumulator variable for storing the actual water vapor adsorption total mass M actual , which is initially 0. The controller iterates through each time sampling point of the adsorption cycle time T cycle , specifically from t=0 to t=T cycle . At each time point t, the controller first calls the saturated water vapor density calculation model, inputs the temperature and relative humidity of the point, and calculates the instantaneous water vapor density ρv (t). Then, the resulting p v (t) is multiplied by the instantaneous compressed air volume flow rate q v (t) at that point, resulting in the instantaneous water vapor mass flow rate m w (t). m w (t) is multiplied by the sampling time interval (1 second in this case) to obtain the water vapor mass that flowed in during that infinitesimal time period, which is accumulated into the actual water vapor adsorption total mass M actual . This process is numerically equivalent to numerically integrating the instantaneous water vapor mass flow rate m w (t). In this case, the "sampling time interval" is 1 second; after the traversal is complete, the value of the accumulator variable is the actual water vapor adsorption total mass M actual for this cycle.

[0052] 2.2) Calculate the working condition corrected theoretical maximum adsorption capacity M max,c , specifically: the controller calculates the arithmetic mean of the temperature time series array within the length of the current adsorption cycle T cycle , resulting in the cycle average inlet temperature T avg . Similarly, the cycle average inlet pressure P avg is calculated. The controller calls the pre-installed adsorption isotherm model function, taking T avg and P avg as inputs, calculates and returns the working condition corrected theoretical maximum adsorption capacity M max,c under the current working condition.

[0053] 2.3) Calculate the single-tower adsorption efficiency index η eff : divide the actual water vapor adsorption total mass M actual by the working condition corrected theoretical maximum adsorption capacity M max,c to obtain the preliminary efficiency index. If the preliminary efficiency index is greater than 1.0, the value of the single-tower adsorption efficiency index η eff is forcibly corrected to 1.0. This step is used to handle abnormal calculation results caused by sensor instantaneous errors or model boundary problems, ensuring that the index has the correct physical meaning. Otherwise, the value of the single-tower adsorption efficiency index η eff remains unchanged. This embodiment chooses physical calculation based on the adsorption isotherm model, which does not require a large amount of training data and the calculation results are physically credible. In this way, this method avoids the inaccuracy brought by simply weighting multiple parameters with different dimensions and different influence laws.

[0054] Further explanation: the core algorithm output value of this embodiment is the single-tower adsorption efficiency index η eff . The theoretical value range of this index is limited to the [0, 1] interval. The single-tower adsorption efficiency index η effis calculated by dividing the actual total water vapor adsorption mass M actual calculated based on the average working condition dynamic of the cycle by the working condition correction theoretical maximum adsorption capacity M max,c . If the calculation result accidentally exceeds 1 due to sensor noise, it will be forcibly corrected to 1 to ensure the effectiveness of its physical meaning.

[0055] When the single-tower adsorption efficiency index η eff tends to 1, the performance state of the adsorption tower tends to be brand new and ideal. This means that its adsorption capacity is less attenuated under the current working condition.

[0056] When the single-tower adsorption efficiency index η eff tends to 0, the performance attenuation of the adsorption tower is more serious, the desiccant (including but not limited to activated alumina) inside the tower tends to be close to the end of its adsorption life, or the degree of structural damage (including but not limited to pulverization and hardening) is greater, resulting in a lower effective adsorption capacity of water vapor under the current working condition. The key input parameters affecting the final output single-tower adsorption efficiency index η eff mainly affect its numerator M actual and denominator M max,c , and their influence relationship is analyzed as follows:

[0057] The actual total water vapor adsorption mass M actual is positively correlated with the single-tower adsorption efficiency index η eff . The calculation formula of the single-tower adsorption efficiency index η eff is M actual / M max,c . Under the condition that the theoretical adsorption capacity represented by the denominator M max,c remains unchanged, the greater the actual adsorption amount represented by the numerator M actual , the higher the efficiency of the tower.

[0058] The cycle average inlet temperature T avg and the single-tower adsorption efficiency index η eff have an indirect, working condition-eliminating compensation relationship, rather than a direct positive / negative correlation. Temperature rise will cause two effects: the first effect: in the real world, high temperature will inhibit the physical adsorption capacity of the adsorbent, resulting in a decrease in M actual ; the second effect: in this embodiment, high temperature will cause the denominator M max,c to decrease through the adsorption isotherm model; this design is the core innovation point of the present application. It makes the calculation of the single-tower adsorption efficiency index η eff a relative value. If a healthy adsorption tower causes M actual to decrease due to a rise in working temperature, its M max,cThe result η, obtained by dividing the two, will also be adjusted downwards accordingly. eff It will remain stable, that is, approach 1. Only when the adsorption tower itself ages and its actual capacity declines more severely than the influence of operating conditions will the single-tower adsorption efficiency index η increase. eff Only then will it decrease significantly. Therefore, the periodic average inlet temperature T avg The changes were inherently and reasonably compensated for by the algorithm, resulting in a lower single-tower adsorption efficiency index η. eff It can truly reflect the health decline of the tower itself, rather than the fluctuations in operating conditions.

[0059] Periodic average inlet pressure P avg With the single-tower adsorption efficiency index η eff There also exists an indirect compensatory relationship used to eliminate the effects of operating conditions; increased pressure promotes adsorption, leading to M actual Increase. Simultaneously, the algorithm will use the adsorption isotherm model to increase the denominator M. max,c This synchronous adjustment increases the single-tower adsorption efficiency index η. eff It can focus on the performance degradation of the adsorption tower itself based on the apparent changes caused by pressure fluctuations.

[0060] The following table shows a comparison between the method of this invention and the traditional fixed-benchmark method under different operating conditions and different aging degrees. The traditional method uses a fixed factory-calibrated theoretical maximum adsorption capacity of 5.0 kg as a benchmark.

[0061]

[0062]

[0063] In this embodiment, if the calculated value of the traditional efficiency index of the conventional method exceeds 1, it will be forcibly corrected to 1.0.

[0064] Comparing Experiment 1 and Experiment 3: These two sets of data demonstrate the ability of this invention to evaluate the novel tower structure under different temperature conditions. Under standard conditions (Group 1), both this invention and the conventional method provide accurate evaluations close to 1. However, when the conditions change to high temperatures (Group 3), the conventional method, still using 5.0 kg as a baseline, incorrectly evaluates the efficiency index as 0.76, which could falsely report a healthy tower as underperforming. In contrast, the single-tower adsorption efficiency index η of this invention... eff Because the denominator M max,c The pressure was dynamically adjusted from 5.00 kg to 3.80 kg, and the final output remained at 0.99, accurately identifying that the tower was still healthy, and its performance decline was solely due to the operating conditions. This strongly demonstrates that the present invention can effectively decouple the interference of the equipment's own state from external operating conditions, avoiding misjudgments.

[0065] Comparing group two and group four: these two groups of data show the evaluation of the tower body which has aged 50%. In the standard working condition (group two), the evaluation results of the two methods are consistent. But in the high temperature working condition (group four), the efficiency index given by the traditional method is 0.38, which has a significant decrease compared with 0.50 of group two, which will exaggerate the aging degree of the tower body. While the single-tower adsorption efficiency index η eff of the present application is 0.50 in both of the two working conditions, which accurately and consistently reflects the real 50% health level of the tower body. This proves the consistency and robustness of the evaluation results of the present application, which can give stable health measurement regardless of the change of external environment.

[0066] Analyzing group five: in the high pressure working condition, the actual total mass M actual of water vapor adsorbed by the new tower (5.35kg) even exceeds the fixed benchmark (5.0kg) of the traditional method, resulting in abnormal calculation. While the present application raises the theoretical capacity M max,c to 5.40kg, which not only avoids the abnormal calculation, but also gives an accurate evaluation of 0.99. This again highlights the core advantage of the dynamic benchmark of the present application.

[0067] The present embodiment can avoid misjudgment and exaggerated evaluation of the health status caused by the change of working conditions (such as temperature, pressure), thereby providing a more accurate, stable and reliable equipment health measurement index.

[0068] Based on the output value of the single-tower adsorption efficiency index η eff , combined with the expert experience and long-term running data analysis of the dryer maintenance, when the calculated single-tower adsorption efficiency index η eff is less than the preset health threshold η th , it is determined that the adsorption tower performance has declined; otherwise, it is determined that the adsorption tower is healthy. The health threshold is determined by one of the following methods:

[0069] Based on the factory calibration method: the equipment manufacturer selects a brand new adsorption tower when it leaves the factory, runs under standard working conditions (specified inlet temperature, pressure and flow rate), and measures the single-tower adsorption efficiency index as the health benchmark value η0; in order to avoid measurement fluctuation, the health threshold is set as η th =k1·η0, wherein k1 is a correction coefficient, and the value range is 0.85-0.95.

[0070] Based on the historical running data method: the single-tower adsorption efficiency index data of the past 300 cycles of the adsorption tower are called, the average value of the highest 10% is taken as η0, and the health threshold η th is obtained by multiplying the above correction coefficient k1.

[0071] To further explain, the specific steps in S3 for integrating and calculating the quantitative index of tower fatigue accumulation include:

[0072] Establish a damage weighting function based on the kurtosis K of the differential pressure signal. p Pressure difference fluctuation frequency, main peak value F peak and the single-tower adsorption efficiency index η eff The instantaneous fatigue damage value is calculated, and the instantaneous fatigue damage value is accumulated over time to obtain a quantitative index of the fatigue accumulation degree of the tower body.

[0073] To further clarify, the basic kurtosis damage is defined as D. B,k Characterized by the kurtosis K of the pressure difference signal p The independently contributed basic damage component, which ignores the current health status of the tower; basic kurtosis damage D B,k The damage mapping model was calculated using a "basic kurtosis damage mapping model." This model is a nonlinear function calibrated based on experimental data, designed to map kurtosis values ​​to the damage domain. The establishment and calculation method of the "basic kurtosis damage mapping model" are as follows:

[0074] 1. Prepare multiple sets of brand-new, identical desiccant samples. Using an experimental setup capable of precisely controlling the airflow pulse, apply a series of known pressure differential signals with different kurtosis K to these samples. p The airflow impact was tested. After each impact test, the degree of microstructural damage to the desiccant sample was quantitatively measured using precision instruments such as scanning electron microscopy (SEM) or surface area and pore size analyzer (BET). Specifically, the degree of microstructural damage was characterized by microcrack density or surface area attenuation rate.

[0075] 2. The kurtosis K of the pressure difference signal measured in the experiment. p The value is used as input, and the corresponding degree of microstructural damage is used as output for nonlinear regression analysis. The analysis reveals a correlation between damage and the kurtosis K of the pressure difference signal. p There exists a threshold and exponential growth relationship between them. Therefore, the following linguistic computational model is established: First, from the input pressure difference signal kurtosis K... p Subtract a preset kurtosis damage threshold K from the middle. th Then, the difference obtained (or zero if negative) is subjected to a kurtosis damage index E. k Exponentiation.

[0076] 3. Kurtosis damage threshold K th The value is set to 3.5, and its value is obtained by analyzing the kurtosis K of differential pressure signals under a large number of normal operating conditions. p The statistical distribution is determined, and in this embodiment, the 95th or 99th percentile is used, representing fluctuations below this value that are not considered valid damage. Kurtosis Damage Index Ek Set to 1.5, whose value is determined by the best fitting curve of the aforementioned nonlinear regression analysis, reflecting the nonlinear growth rate of damage with the part of the impact intensity exceeding the threshold. If the input pressure difference signal kurtosis K p is 8.5, the kurtosis damage threshold K th is 3.5, the kurtosis damage exponent E k is 1.5, then first calculate the difference value as 5.0, and then perform the 1.5 power operation on this difference value to obtain the value of the basic kurtosis damage D B,k is about 11.18.

[0077] Define the basic frequency damage as D B,f : representing the basic damage part contributed by the main peak value F peak of the pressure difference fluctuation frequency, also ignoring the current health status of the tower body. Its value is calculated through a "basic frequency damage mapping model". This model is a piecewise function or lookup table, whose data is derived from vibration response experiments on the desiccant bed. The determination method of the basic frequency damage D B,f is as follows:

[0078] 1. Place a transparent tower body filled with desiccant on a vibration table, and arrange multiple acceleration sensors on the tower body. By controlling the vibration table to generate vibrations of different frequencies (corresponding to the main peak value F peak of the pressure difference fluctuation frequency), monitor the behavior of the bed layer, such as whether fluidization, channeling or compaction occurs, using a high-speed camera and sensors. Record the frequency interval that triggers significant adverse phenomena and the corresponding vibration intensity.

[0079] 2. Based on the experimental results, establish a dangerous frequency damage lookup table. This lookup table divides the entire frequency range into multiple intervals, and assigns a basic damage value to each interval. This lookup table can be defined as follows: if the main peak value F peak of the pressure difference fluctuation frequency is in the 0 to 2 Hz interval of the normal working condition fluctuation, the corresponding basic frequency damage D B,f is 0; if the main peak value F peak of the pressure difference fluctuation frequency is in the 5 to 10 Hz interval (representing the resonance danger zone of the corresponding equipment), the corresponding basic frequency damage D B,f is a higher constant value of 5.0, which is proportional to the degree of bed layer disturbance observed during resonance; for other non-dangerous frequency intervals, the corresponding basic frequency damage D B,f is a lower background value of 0.1. Thus, if the current input pressure difference fluctuation frequency main peak value F peak is 6.2 Hz, then the basic frequency damage D B,f obtained through the lookup table is 5.0.

[0080] Define the health status damage coefficient as C Hη as a multiplier factor, quantifying the amplification of the basic damage by the current health state of the adsorption tower body eff The core innovation of the embodiment is to characterize the amplification effect of the basic damage on the current health state of the adsorption tower body, that is, the more unhealthy the adsorption tower body is, the more vulnerable it is to the impact; the health state damage coefficient C H is calculated through a “health state damage modulation model”. The model is an exponential function based on the aging mechanism of the tower body, and the parameters are determined by analyzing the impact experimental data of tower bodies with different aging degrees. The specific determination method is as follows: use multiple groups of desiccant samples with different known single-tower adsorption efficiency indexes η eff to represent different aging degrees, and apply the same known airflow impact to them. Measure and compare the performance attenuation increments of different aging samples after suffering the same impact. Analysis shows that the performance attenuation increment increases exponentially with the decrease of the single-tower adsorption efficiency index η eff . Based on this, subtract the input single-tower adsorption efficiency index η eff from 1 to obtain an intermediate value representing “unhealthiness”; then, multiply the intermediate value by a health attenuation sensitivity factor λ H ; finally, calculate the power of the product obtained in the previous step with the natural constant e as the base. The health attenuation sensitivity factor λ H is set to 2.0, and its value is obtained by fitting the aforementioned experimental data, which quantifies the degree of severity of the damage amplification effect with the decline of health. If the input single-tower adsorption efficiency index η eff is 0.5 and the health attenuation sensitivity factor λ H is 2.0, then the “unhealthiness” is calculated to be 0.5, and then multiplied by the health attenuation sensitivity factor λ H to obtain 1.0, and finally calculate the 1.0 power of e to obtain the value of the health state damage coefficient C H , which is 2.718.

[0081] The instantaneous fatigue damage value is defined as ΔD fatigue : represents the total damage suffered by the tower body in a calculation period, which is modulated by the health state. The calculation period of the embodiment is set to 2 seconds; the tower body fatigue accumulation degree D total : is a state variable representing the total fatigue damage accumulated by the tower body since the last “restorative regeneration” or system initialization. In the embodiment, the process of calculating the tower body fatigue accumulation degree is executed periodically by the controller, and is synchronized with the calculation period of S1, which is 2 seconds; the specific steps are as follows: the starting input of the process of the embodiment is the kurtosis K p and the main peak value F peak of the pressure difference fluctuation frequency calculated by S1 in the current calculation period, and the single-tower adsorption efficiency index η effThe tower fatigue accumulation degree D at the moment is stored in the controller memory total .

[0082] First, calculate the basic damage: call the "basic kurtosis damage mapping model", input the differential pressure signal kurtosis K p , calculate the basic kurtosis damage D B,k ; call the "basic frequency damage mapping model", input the differential pressure fluctuation frequency main peak value F peak , calculate the basic frequency damage D B,f ; second, calculate the health state damage coefficient: call the "health state damage modulation model", input the single tower adsorption efficiency index η eff , calculate the health state damage coefficient C H ; third, calculate the instantaneous fatigue damage value: add the basic kurtosis damage D B,k and the basic frequency damage D B,f obtained in the first step to obtain a total basic damage value. Then, multiply the sum by the health state damage coefficient C H obtained in the second step to obtain the final instantaneous fatigue damage value ΔD fatigue ; the fourth step is to update the tower fatigue accumulation degree: add the instantaneous fatigue damage value ΔD fatigue calculated in the third step to the tower fatigue accumulation degree at the last moment to obtain the updated tower fatigue accumulation degree D total at the current moment. The final output is the updated tower fatigue accumulation degree D total which is stored back into the controller memory and serves as the core decision basis for subsequent S4 and S5 modules; this embodiment adopts a "series type, two-stage nonlinear fusion model".

[0083] First stage: fuse the differential pressure signal kurtosis K p and the differential pressure fluctuation frequency main peak value F peak by addition to obtain the total basic damage. Second stage: multiply the total basic damage in the first stage with the single tower adsorption efficiency index η eff representing the health state. A healthy tower can basically withstand the basic damage, while an unhealthy tower will amplify the same basic damage by several times. This multiplicative relationship can more accurately simulate the physical reality of "fragility amplification" than the additive relationship. The single tower adsorption efficiency index η eff and the final instantaneous fatigue damage value ΔD fatigue are monotonically decreasing negative correlation. That is, the healthier the tower, the higher the single tower adsorption efficiency index η eff , the smaller the corresponding health state damage coefficient C H , and the smaller the final calculated instantaneous fatigue damage value ΔD fatigue .

[0084] For further explanation of step S3: the core algorithm of this embodiment outputs the value of tower fatigue accumulation degree D total . The theoretical value range of this index is [0, +∞). It is a non-negative, monotonically non-decreasing cumulative value.

[0085] When the value of tower fatigue accumulation degree D total tends to 0, it indicates that the adsorption tower is in a "fatigue clean" state, and the structural cumulative damage caused by dynamic load impact is low, close to its initial or repaired optimal physical state.

[0086] The value of tower fatigue accumulation degree D total continuously increases over time, indicating that the tower is continuously accumulating irreversible physical fatigue. The size of its value directly quantifies the total effect of all impact loads the tower has suffered since the last "zeroing". A higher tower fatigue accumulation degree D total value indicates that the tower's desiccant bed has a higher risk of pulverization, channeling or caking, and its mechanical stability and service life have been significantly consumed.

[0087] When the kurtosis K p of the pressure difference signal exceeds the preset kurtosis damage threshold K th , the increase in its value will cause the basic kurtosis damage D B,k to increase sharply through a power function, thereby directly increasing the instantaneous fatigue damage value ΔD fatigue , thereby accelerating the accumulation of tower fatigue accumulation degree D total . The impact and wear effect of sharp pressure pulses (high kurtosis) on desiccant particles is much greater than that of gentle pressure changes, and its destructive effect increases exponentially with the increase in impact strength. The main peak value F peak of the pressure difference fluctuation frequency has a segmented positive correlation with the instantaneous fatigue damage value ΔD fatigue .

[0088] When the main peak value F peak of the pressure difference fluctuation frequency falls within the dangerous frequency interval calibrated by experiments, the basic frequency damage D B,f will jump from 0 or a very small value to a significant set value, thereby increasing the instantaneous fatigue damage value ΔD fatigue , accelerating the accumulation of tower fatigue accumulation degree D total . It indicates that only a specific frequency excitation can cause the resonance or fluidization of the desiccant bed, causing the most serious structural disturbance and particle wear, while other frequency excitations have little effect.

[0089] The single-tower adsorption efficiency index η eff has a segmented positive correlation with the instantaneous fatigue damage value ΔD fatiguea non-linear negative correlation. The single-tower adsorption efficiency index η eff The decrease of the health state impairment coefficient C H exponentially. Since the health state impairment coefficient C H acts as a multiplier factor on the total basic impairment determined by the kurtosis K p and the main peak value F peak of the pressure difference fluctuation frequency, a lower single-tower adsorption efficiency index η eff will amplify the instantaneous fatigue impairment value ΔD fatigue caused by the same external impact. This design profoundly simulates the fatigue damage accumulation theory in material mechanics: a material that already has micro-cracks or structural degradation (low single-tower adsorption efficiency index η eff ), when subjected to the same external load, its damage propagation speed is much faster than that of a perfect material. The following data table shows the instantaneous fatigue impairment values calculated by the method of the present invention under the combination of different impact loads and different tower health states.

[0090]

[0091]

[0092] The parameters used in the calculation are K th = 3.5, E k = 1.5, λ H = 2.0. In the sixth experimental group, K p < K th , so D B,k is 0; comparison of the first and second experimental groups: these two groups of data show the core effect of the present invention. When subjected to the same strong pulse impact (K p = 8.5), the healthy tower (group one, η eff = 0.99) produces an instantaneous impairment value of 11.40. However, when the tower is in a sub-healthy state (group two, η eff = 0.50), its instantaneous impairment value increases to 30.41, which is 2.67 times that of the former. This strongly proves that the present invention is not simply recording the impact, but successfully quantifies the non-linear amplification effect of "health state" on "impact damage". Compared with the prior art that only records the number or intensity of impacts, the present invention can more accurately assess the real damage to the equipment, because it reveals that the harm of impact on a sub-healthy device is much greater than that on a healthy device.

[0093] Comparison of experimental groups three, four, and five: comparing group three and group four, it can be seen that the same amplification effect also applies to the resonance frequency impact. Further comparison of group two (only strong pulse), group four (only resonance frequency), and group five (double impact), it can be found that, under the same sub-healthy state (ηeff The damage (30.41 or 13.60) caused by a single impact source (strong pulse or resonance frequency) is significant on the tower body with a fatigue damage value of 0.50, while when the two impacts occur simultaneously (Group Five), the instantaneous damage value (44.01) is close to the sum of the two independent damages, accurately reflecting the superimposed effect of multiple risks. This proves the rationality and comprehensiveness of the fusion mechanism of the present application.

[0094] Further, the step of performing predictive work rotation in S4 specifically comprises: when it is monitored that a gas use load with a high instantaneous fatigue damage value will occur in a future time window, if the tower body fatigue accumulation degree of the current working tower is higher than the preset tower body protection threshold, a protective switching is performed to transfer the working task to another adsorption tower with a lower tower body fatigue accumulation degree.

[0095] In another embodiment, the present embodiment is intended to be used in a dryer system with at least two adsorption towers, to actively protect the equipment from high-intensity load impact. Through a forward-looking load prediction model, an upcoming gas use event that can produce a high instantaneous fatigue damage value is identified in real time. Once such an event is predicted, the system will immediately evaluate the tower body fatigue accumulation degree of the current working tower. If the accumulation degree has exceeded a preset protection threshold, and there is another standby tower with a lower fatigue accumulation degree, the system will trigger a protective switching to dynamically transfer the working task to the healthier adsorption tower, thereby achieving intelligent avoidance of the fatigued equipment and optimized management of the entire system life. In the present embodiment, the tower body fatigue accumulation degree D total,A of the current working tower and the tower body fatigue accumulation degree D total,B of the standby tower are respectively calculated according to the following formula:

[0096] The predicted instantaneous fatigue damage value is denoted as ΔD predict : a predicted estimate of the instantaneous fatigue damage value in a future time window, used to identify an upcoming high-intensity impact in advance. The present embodiment sets the future time window to be 5-10 seconds; the value is calculated through a "forward-looking load prediction model based on time series analysis". The forward-looking load prediction model is preferably an autoregressive model (AR model); the specific determination method is as follows:

[0097] The historical time series data of the instantaneous fatigue damage value ΔD fatigue of the dryer under various typical working conditions for a long time is collected. Then, the least squares method or the Yule-Walker equation of the system identification method is used to analyze the time series to determine the optimal model order p1 and the corresponding autoregressive coefficients (φ1, φ2,..., φ p1 ). These coefficients together constitute the parameter set of the AR model.

[0098] In real-time operation, the model is calculated as follows: the instantaneous fatigue damage values (AD fatigue (t-1),...,AD fatigue (t-p1)) of the last p1 historical time points are obtained, each historical value is multiplied by its corresponding autoregressive coefficient (φ1, φ2,..., φ p1 ), and then all the products are summed up to obtain the predicted instantaneous fatigue damage value AD predict .

[0099] In this embodiment, a second-order AR model (p1=2) is determined offline, and its coefficients are φ1=0.7 and φ2=0.2. If the instantaneous fatigue damage values AD fatigue of the two time points before the current time point are 10.0 and 5.0 respectively, the calculation of the predicted instantaneous fatigue damage value AD predict is as follows: 10.0 is multiplied by 0.7 to obtain 7.0, 5.0 is multiplied by 0.2 to obtain 1.0, and the sum of the two is 8.0, which is the value of the predicted instantaneous fatigue damage value AD predict . H,th A high-damage load threshold D H,th is defined: as the first judgment threshold for triggering the protective switching logic; this threshold is determined based on statistical analysis of historical instantaneous fatigue damage value data, aiming to identify statistically significant "small probability high intensity" impact events. Its determination method is as follows: a large amount of historical instantaneous fatigue damage value data is analyzed for probability density distribution, and a certain high percentile of the distribution is selected as the threshold. The 98th percentile is selected. Any predicted damage value exceeding this threshold is considered to be a rare event that may cause significant damage to the equipment. In this example, the 98th percentile value of the historical instantaneous fatigue damage value data is 25.0, so the high-damage load threshold D P,th is set to 25.0.

[0100] A tower protection threshold D P,th is defined: as the basis for judging whether the current working tower is in a "fatigue" or "fragile" state. The determination of this threshold is related to the overall life management strategy of the equipment and the fatigue characteristics of the material. It is set to a safe proportion of the theoretical fatigue life limit of the equipment. Its determination method is as follows: through offline accelerated aging experiments or finite element simulation, the limit tower fatigue accumulation D limit corresponding to irreversible structural damage of the desiccant bed is estimated. Then, the limit value is multiplied by a pre-set safety protection factor γ P . The safety protection factor γ P is a constant less than 1, and in this example it is initially set to 0.4, which is based on engineering experience and maintenance strategy, aiming to leave enough safety margin before the equipment reaches a truly dangerous state.

[0101] Define the switching cooldown time T cool This parameter is a time value in seconds, used to prevent excessively frequent protective switching due to system fluctuations. Its determination is primarily based on engineering requirements for protecting the mechanical lifespan of system valves and other actuators, as well as avoiding high-frequency oscillations in the control logic. This value is set directly by the engineer as an empirical constant based on the minimum switching interval recommended by the dryer valve manufacturer and the stability requirements of the system control. In this embodiment, considering valve wear during switching and the need for stable airflow, the switching cooling time T is... cool The interval is set to 300 seconds. In this embodiment, the predictive job rotation logic is executed cyclically by the controller at a high frequency, with the current working tower designated as A3 and the standby tower as B3; the high frequency is set to once per second; the cumulative fatigue degree D of the current working tower A3 is obtained. total,A The cumulative fatigue degree D of the standby tower B3 total,B , and the most recent p1 historical instantaneous fatigue damage values ​​used for prediction; and the system clock variable recording the last protective switching time;

[0102] The predictive job rotation logic includes a first-level judgment to determine whether a "high-intensity impact is anticipated", a second-level judgment to determine whether the "current working tower is in a vulnerable state", a third-level judgment to determine whether a "better backup tower exists", and a fourth-level judgment to determine whether the "timeout exceeds the switching cooldown time".

[0103] When the predictive job rotation logic outputs "Execute", it indicates that the system has simultaneously met the four core conditions: "foreseeing a high-intensity impact", "the current working tower is in a vulnerable state", "there is a better backup tower", and "greater than the switching cooldown time". This output directly triggers a physical valve switching action. At the same time, the "last protective switching time" variable inside the system is updated to the current time. The above is the final embodiment of the core technical idea of ​​actively avoiding damage in this invention.

[0104] When the predictive job rotation logic outputs "Maintain," it indicates that at least one of the four core conditions mentioned above is not met; the system determines that a protective switch is not necessary or cannot be performed at present, and will maintain the current operating state. This output ensures the accuracy of the switchover action and avoids unnecessary system disturbances. The "Proactive Load Forecasting Model" is invoked, and a sequence of historical instantaneous fatigue damage values ​​is input to calculate the predicted instantaneous fatigue damage value ΔD. predict .

[0105] The first-level judgment is to use the obtained predicted instantaneous fatigue damage value ΔD predict With high damage load threshold D H,th The comparison is performed, and the result is used to characterize whether a "high-intensity impact was anticipated"; if the "predicted instantaneous fatigue damage value ΔD" is true, then the comparison is performed.predict " Greater than "High Damage Load Threshold D H,th ": indicates that a high intensity impact is imminent, proceed to the next step; if not: indicates that the future load is normal, no need to start the protection logic, end of this calculation cycle;

[0106] The second level judgment is to obtain the tower fatigue accumulation D total,A of the current working tower A3, and compare it with the tower protection threshold D P,th . The comparison result is used to determine whether the current working tower is in a fragile state or not; if the tower fatigue accumulation D total,A of the working tower A3 is greater than the tower protection threshold D P,th : it indicates that the current working tower A3 is fragile and needs protection, proceed to the next step; if not: it indicates that the tower A3 is healthy enough to withstand this impact, no need to switch, end of this calculation cycle;

[0107] The third level judgment is to obtain the tower fatigue accumulation D total,B of the standby tower B3, and compare it with the tower fatigue accumulation D total,A of the tower A3; the comparison result is used to determine whether there is a better standby tower or not; if the tower fatigue accumulation D total,B of the standby tower B3 is less than the tower fatigue accumulation D total,A of the working tower A3: it indicates that there is a tower in a better state to take over the work, proceed to the next step; if not: it indicates that the standby tower is worse or comparable, switching is meaningless, no need to switch, end of this calculation cycle.

[0108] The fourth level judgment is to check whether the time difference between the current system time and the last time when the protective switching is performed is greater than the switching cooling time T cool ; the comparison result is used to determine whether it is greater than the switching cooling time or not; if yes: it indicates that the switching can be performed, proceed to the final action; if not: it indicates that the switching is too frequent, in order to ensure the stability of the system, the switching is prohibited this time, end of the calculation cycle.

[0109] If all the above logical judgments are "yes", the controller outputs a switching instruction to control the related electromagnetic valve to switch the airflow path from the working tower A3 to the standby tower B3, and at the same time, updates the "last protective switching time" variable in the system to the current time.

[0110] The decision mechanism of the embodiment adopts a rule-based, multi-level serial condition decision tree model. The model organizes multiple judgment conditions in a serial manner with clear logical priority. In this decision model, there is a monotonic relationship between all input parameters and the probability of triggering the "protective switching" action, and there is no contradiction or cancellation. The predicted instantaneous fatigue damage value ΔDpredict The increase of D will increase the probability of passing the first level of judgment, thus positively promoting the triggering of switching. The increase of D total,A The increase of D will increase the probability of passing the second level of judgment, thus positively promoting the triggering of switching. The increase of D total,B The increase of D will decrease the probability of passing the third level of judgment, thus negatively inhibiting the triggering of switching.

[0111] Further explanation for step S4: the value set of the decision in this embodiment is {execute, maintain}. The decision is obtained in the following way: in each calculation period, the system judges multiple input parameters through a multi-level serial condition decision tree model, and outputs “execute” if all the preceding conditions are met, otherwise outputs “maintain”. The prediction of instantaneous fatigue damage value ΔD predict has a positive relationship with the triggering of switching decision in a threshold activation manner. Only when its value exceeds the high damage load threshold D H,th , the subsequent judgment process will be activated. A lower ΔD predict value will directly lead to the decision result of “maintain”. This design fully complies with the “predictive” nature of the present application: only when a truly threatening impact is foreseen, the system needs to consider starting the protection mechanism, thus avoiding overreaction to regular load fluctuations.

[0112] The tower fatigue accumulation degree D total,A of the current working tower has a positive relationship with the triggering of switching decision in a threshold activation manner. This parameter is the key to judging the necessity of protection. Only when its value exceeds the tower protection threshold D P,th , indicating that the current working tower has entered a “fragile” state, the system needs to avoid damage to it. If the working tower A3 itself D total,A has a low value representing health, even if a high-intensity impact is coming, it is fully capable of bearing it, and there is no meaning to execute switching at this time. This design ensures the accurate deployment of protection resources, protecting only the equipment that needs protection the most.

[0113] The tower fatigue accumulation degree D total,B of the standby tower has a negative relationship with the triggering of switching decision in a relative comparison manner; only when the value of D total,B is strictly less than D total,A , switching has a positive significance. If the standby tower B3 is in a worse or comparable state than the working tower A3, switching does not need to be executed. This design embodies the global optimization idea of the system, ensuring that each switching is a clear, positive operation that maximizes the overall system life.

[0114] The following table shows the decision results under different system states and load predictions, and is compared with the conventional system without using the present application.

[0115]

[0116]

[0117] The decision accuracy verification of the present application (comparison scenarios 1, 2, 3, and 4): This group of comparisons strongly proves the accuracy and robustness of the decision logic of the present application. In scenario 1, all conditions (high predicted impact, tower A3 fatigue, tower B3 healthier) are met, and the system correctly makes the decision to "execute" the switch. In scenario 2, although a high impact is predicted, because tower A3 is healthy enough (D total,A <D P,th ), the system decides to "maintain", avoiding unnecessary switching. In scenario 3, although tower A3 is fatigued, because the standby tower B3 is in a worse state, switching is not beneficial, and the system also decides to "maintain". In scenario 4, because the predicted load is normal (ΔD predict <D H,th ), the protection logic is not triggered at all. This clearly shows that the multi-level serial decision mechanism of the present application can accurately identify the only opportunity that must and deserves protection.

[0118] Quantitative verification of the beneficial effects of the present application (comparison of scenario 1 and scenario 5): This is the core evidence to prove that the present application has outstanding substantive features and significant progress. Under the same initial conditions, the conventional system (scenario 5) lacks prediction and decision-making capabilities and can only passively allow the already fatigued tower A3 to withstand a damage of up to 30.0 units, causing its total fatigue degree to rise to 500,030, further approaching its life limit. The system using the present application (scenario 1) decisively transfers the 30.0 units of damage to the healthier tower B3 to bear at the critical moment, while the fatigue degree of tower A3 increases by 0. Through this operation, the present application directly prevents a high-intensity damage to the vulnerable equipment, greatly delays the failure process of tower A3, and significantly extends the average trouble-free operation time of the entire system.

[0119]

[0120] Further explanation: the step of selecting and executing the adaptive regeneration program in S5 specifically includes: when the tower body fatigue accumulation degree of the adsorption tower to be regenerated exceeds a repair threshold, an additional rest period is added to the conventional regeneration program to perform a repair regeneration.

[0121] Further explanation: the repair threshold is defined as D repair: which is the decision threshold to distinguish between regular regeneration and restorative regeneration. It represents a critical point of tower fatigue accumulation, beyond which regular regeneration is insufficient to maintain its long-term performance, and more in-depth repair measures need to be taken. The repair threshold is determined through the "multi-cycle fatigue-regeneration performance correlation calibration experiment". Offline experimental calibration method: in the laboratory environment, a dry tower sample is subjected to cyclic high-intensity impact load, so that its tower fatigue accumulation D total is gradually increased. At D total a series of preset checkpoints (the preset checkpoints in this embodiment are set to 500,000, 600,000, and 700,000 units in turn), a regular regeneration program is performed once, and then its full-load adsorption performance (the time to reach the dew point breakthrough) is tested. At the same time, another group of the same samples is subjected to a regeneration program with an added fixed-length (the initial fixed length in this embodiment is set to 30 minutes) rest period when reaching the same checkpoints, and its performance is tested. The repair threshold D repair is defined as: at the repair threshold D repair , the performance improvement percentage brought by the regeneration program with the added rest period first significantly exceeds a preset performance gain determination value; the initial performance gain determination value in this embodiment is set to 5%;

[0122] If the experiment finds that when D total is 700,000 units, the adsorption performance after adding the rest period is improved by 6% compared with regular regeneration, which first exceeds the performance gain determination value of 5%, then the repair threshold D repair can be set to 700,000.

[0123] The rest period length T rest : this parameter is a time value in minutes, which is the core innovation point of this embodiment. It is a variable that is self-adaptively adjusted according to the tower fatigue degree, and is used to control the duration of the additional rest period in restorative regeneration; its value is determined by the "self-adaptive length calculation model" to determine the rest period length; the model is preferably a piecewise linear model to balance the repair effect and time cost. The calculation method of the "self-adaptive length calculation model" is to add a fatigue gain length to a basic rest period length T base . The basic rest period length T base is a fixed minimum rest time; the calculation method of the fatigue gain length is to first obtain the tower fatigue accumulation D total of the current tower and the difference between the repair threshold D repair , and then multiply the difference by a rest period length gain coefficient β. The basic rest period length T baseand the rest length gain coefficient β are obtained by the above-mentioned "multi-cycle fatigue-regeneration performance correlation calibration experiment", and are determined by fitting the rest length data points of the best performance recovery with the highest time efficiency under different fatigue degrees.

[0124] If calibrated by experiment, T base is 20 minutes, and β is 0.0001 minute / fatigue unit. When the D total of a tower to be regenerated is 800,000, the difference between it and the D repair of 700,000 is 100,000. The difference multiplied by β gives a fatigue gain length of 10 minutes. The final rest length T rest is the sum of the base rest length of 20 minutes and the fatigue gain length of 10 minutes, i.e. 30 minutes.

[0125] The fatigue degree repair coefficient is defined as R factor : This parameter is a dimensionless floating point number between 0 and 1, which is used to reduce the tower fatigue accumulation degree after performing a repair regeneration, so as to quantitatively reflect the effect of "repair". This coefficient is also determined by the "multi-cycle fatigue-regeneration performance correlation calibration experiment". In the calibration experiment, the performance recovery degree of a tower that has performed a repair regeneration is measured. For example, if the adsorption efficiency of a tower decreases from 100% to 80% due to fatigue, and the efficiency recovers to 95% after repair regeneration, the performance recovery degree is (95-80) / (100-80)=75%. The fatigue degree repair coefficient R factor is set to this average performance recovery percentage. In this example, the average performance loss that can be recovered by repair regeneration is initially set to 75%, so R factor is set to 0.75. The adaptive regeneration program of this embodiment is triggered when any adsorption tower completes its adsorption work cycle and is about to enter the regeneration phase; the adsorption tower to be regenerated is set as A4, which is the representation of the current working tower A3 or the standby tower B3.

[0126] The initial input is set as the tower fatigue accumulation degree D total,A4 of the tower to be regenerated A4; the input tower fatigue accumulation degree D total,A4 of the tower to be regenerated A4 is compared with the preset repair threshold D repair ; if D total,A4 >D repair : it indicates that the adsorption tower to be regenerated has been seriously fatigued and needs to be deeply repaired. The system decision execution path one: repair regeneration program. If not: it indicates that the fatigue degree of the adsorption tower to be regenerated is still within the controllable range, and the regular regeneration can be performed. The system decision execution path two: regular regeneration program.

[0127] For path one, the repair regeneration program:

[0128] a. Perform regular regeneration: The controller first executes a standard regeneration procedure, e.g. purging the adsorbent bed with dry air for a specified time to remove the adsorbed moisture. b. Calculate the rest duration: Call the "adaptive duration calculation model" with input D total,A4 , and calculate the rest duration T rest required for this repair. c. Perform rest: After the regular regeneration is finished, the controller closes all the inlet and outlet valves of the tower, making it completely sealed and resting, and starts a timer with duration T rest d. Update the fatigue: After the rest period, the controller updates the fatigue of the tower. The new tower fatigue accumulation D total,A4,new is calculated as: D total,A4 current times (1 minus the fatigue repair coefficient R factor ).

[0129] For path two, the regular regeneration procedure:

[0130] a. Perform regular regeneration: The controller only executes a standard regeneration purge procedure. b. Update the fatigue: After the regeneration is finished, the tower fatigue accumulation D total,A4 remains unchanged, or is reset to a very small initial value according to system settings, but does not undergo reduction based on repair effect.

[0131] The final output of this embodiment, whether path one or path two, is an adsorption tower that has completed regeneration and is in standby state, and a tower fatigue accumulation value that records its current health status, which has been updated accordingly. In this embodiment, the rest duration T rest is calculated using an adaptive linear scaling model based on threshold triggering. This model combines a fixed reference value and a dynamic increment value proportional to the super-threshold fatigue.

[0132] Further explanation: The core algorithm output value of this embodiment is the rest duration T rest calculated dynamically according to the system state to guide physical operations. The parameter value range is [0, +∞), with the unit being minutes. Determine whether the tower fatigue accumulation D total of the tower to be regenerated exceeds the repair threshold D repair . If not, T rest is 0; if it does, the rest duration T rest is obtained by adding a basic rest duration T base and a fatigue gain duration scaled linearly according to the super-threshold fatigue. When the output value rest duration T restWhen equal to 0, it represents that the system judges that the fatigue degree of the current tower body has not reached the critical point needing deep repair, and the normal regeneration program can be executed. This represents the normal operation and maintenance state of the equipment. When the output value is static, the length T rest is greater than 0 and tends to increase, it represents that the cumulative fatigue of the tower body has entered a stage needing to be actively intervened and repaired. When the output value is static, the length T rest is greater than 0 and tends to increase, it represents that the cumulative fatigue of the tower body has entered a stage needing to be actively intervened and repaired. The numerical value of the length T

[0133] The only key input parameter affecting the final output value of the length T rest is the tower body fatigue accumulation degree D total . The relationship between D total and the length T rest is a segmented linear positive correlation based on a threshold activation.

[0134] Setting the repair threshold D repair as the activation condition is based on the dual considerations of cost benefit and physical mechanism. For slight fatigue accumulation, its influence on the performance of the equipment is not significant, and the normal regeneration can cope with it. At this time, if the repair regeneration is executed, the availability of the equipment will be reduced. Therefore, only when D total exceeds D repair , it indicates that the damage has accumulated to a certain degree, and the normal regeneration is insufficient to inhibit the performance degradation trend, and the repair program is started. This ensures the accurate allocation of repair resources. Once the repair program is activated, the linear model is used to determine T rest , which is based on the deep understanding of the adsorbent repair mechanism. The higher the tower body fatigue accumulation degree is, the more the internal microscopic stress concentration points are, and the more uneven the residual moisture distribution is. According to the physical diffusion and stress relaxation theory, more serious internal disorder needs a longer time to reach a relatively balanced and stable state. Linearly increasing T rest is an efficient and effective engineering approximation of the physical law that the greater the damage is, the longer the repair time is. The following table shows the performance changes of the adsorption tower in different fatigue states after adopting the adaptive regeneration scheme and the normal regeneration scheme of the present application respectively.

[0135]

[0136]

[0137] The accuracy verification of the adaptive mechanism of the present application, comparison scenarios A6, B6, C6: this group of comparisons clearly shows the adaptive decision and execution ability of the algorithm of the present application. In scenario A6, since Dtotal Below D repair , the system correctly selects regular regeneration, T rest = 0. Entering scenario B6, when D total just exceeds D repair , the system immediately switches to remedial regeneration mode and calculates T rest = 30 minutes. In scenario C6, facing more severe fatigue, the system not only similarly performs remedial regeneration, but also adaptively extends T rest to 70 minutes. This strongly proves that the present application can accurately identify the remediation opportunity and dynamically adjust the remediation strength according to the damage degree, fully consistent with the algorithm behavior analysis in the first step.

[0138] The quantitative verification of the beneficial effects of the present application is the comparison between scenario C6 and scenario D6: under the same initial state of severe fatigue (D total = 1,200,000), the system using regular regeneration (scenario D6) has no effect on the accumulated physical fatigue during the regeneration process, and the "tower body fatigue accumulation degree" after regeneration is still as high as 1,200,000, and the adsorption efficiency can only recover to 85%, and the equipment is still in a "sick work" state. While the system using the present application (scenario C6), by performing a remedial regeneration including a 70-minute rest, successfully reduces the tower body fatigue accumulation degree by 75% to 300,000, and restores its adsorption efficiency to nearly 98% of the new level. This reflects the fundamental change of the present application from "passive cleaning" to "active repair", which gives the system a "self-healing" ability to reverse the performance degradation caused by long-term operation, and its technical effect has a qualitative leap compared with the prior art. Based on the different intervals of the core input parameter tower body fatigue accumulation degree D total , the regeneration strategy level of the system is defined.

[0139]

[0140]

[0141] The "tower body fatigue accumulation degree" model constructed by the present application as the core realizes multiple and deep linkage control; forward "predictive protection": when a high-risk impact is perceived, the healthier tower body is allowed to "suffer" through protective switching, realizing the guarantee of short-term stability and safety of the system. Backward "compensatory repair": after identifying that the tower body has been excessively fatigued, the accumulated damage is "treated" through remedial regeneration, realizing the maintenance of the long-term health and life of the equipment.

[0142] It should be noted that all the calculation formulas in the application file use regression analysis including but not limited to machine learning algorithms to deeply analyze the collected relevant parameters, identify their natural trend and mutual relationship. Professional software such as Python's Scikit-learn library or R language is used to automatically generate mathematical models matching the data. Then, the model performance is objectively evaluated through methods such as cross-validation, and combined with continuous feedback and optimization to ensure that the created formula truly reflects the inherent law of the data, thereby ensuring its effectiveness and accuracy. In all the calculation formulas in the application, the parameters in each formula are processed by consistent range of dimensionless to ensure that different physical quantities are compared on the same scale; the dimensionless technique includes but is not limited to Min-Max-Normalization, Z-Score standardization;

[0143] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and not to limit it. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, which should be covered in the scope of the claims of the present application.

Claims

1. A method for on-demand regeneration and energy saving of medical compressed air based on dew point feedback, the method being applied to a medical compressed air dryer system comprising at least two adsorption towers, characterized in that, The specific steps include: S1: Based on the real-time differential pressure data at the inlet and outlet of the dryer, the impact characteristics representing the current gas load are extracted through time-series analysis; S2: Based on the historical performance data of the dryer operation, assess the current health status of each of the at least two adsorption towers; S3: Integrate the impact characteristics with the current health status to dynamically calculate a quantitative index characterizing the cumulative fatigue of each adsorption tower. S4: Based on the quantitative index of the cumulative fatigue of the tower body, perform predictive job rotation between the at least two adsorption towers; S5: Based on the quantitative index of the cumulative fatigue of the tower body, select and execute an adaptive regeneration program that matches the adsorption tower to be regenerated.

2. The energy-saving method for on-demand regeneration of medical compressed air based on dew point feedback according to claim 1, characterized in that: The step of extracting the impact feature in S1 specifically includes: Statistical analysis is performed on the real-time differential pressure data to obtain the kurtosis of the differential pressure signal, which characterizes the pulse nature of the gas load intensity, and the main peak value of the differential pressure fluctuation frequency, which characterizes the type and stability of the gas load.

3. The method for on-demand regeneration and energy saving of medical compressed air based on dew point feedback according to claim 2, characterized in that: The step of assessing the current health status in S2 specifically includes: Based on the temperature, humidity, and flow data at the inlet of the adsorption tower, as well as its maximum adsorption capacity according to the operating condition correction theory, the single-tower adsorption efficiency index, which characterizes the degree of performance degradation, is calculated.

4. The energy-saving method for on-demand regeneration of medical compressed air based on dew point feedback according to claim 3, characterized in that: The steps for calculating the maximum adsorption capacity under the corrected operating conditions include: calling a preset adsorption isotherm model for the desiccant used in the adsorption tower in the controller, and using the periodic average inlet temperature and periodic average inlet pressure as inputs to the adsorption isotherm model; the adsorption isotherm model is used to characterize the equilibrium adsorption capacity of the desiccant under different temperature and pressure conditions. The theoretical maximum equilibrium adsorption capacity per unit mass of desiccant under the current operating conditions is determined by the adsorption isotherm model, and then multiplied by the total mass of desiccant in the adsorption tower to obtain the corrected theoretical maximum adsorption capacity under the operating conditions.

5. The energy-saving method for on-demand regeneration of medical compressed air based on dew point feedback according to claim 4, characterized in that: The specific steps in S3 for calculating the quantitative index of the cumulative fatigue degree of the tower body include: A damage weighting function is established, which calculates the instantaneous fatigue damage value based on the kurtosis of the differential pressure signal, the main peak value of the differential pressure fluctuation frequency, and the single-tower adsorption efficiency index; and the instantaneous fatigue damage value is accumulated over time to obtain a quantitative index of the fatigue accumulation degree of the tower body.

6. The method for on-demand energy-saving regeneration of medical compressed air based on dew point feedback according to claim 5, characterized in that: For a detailed explanation of the cumulative fatigue degree of the tower body: The first step is to calculate the basic damage: call the "basic kurtosis damage mapping model", input the pressure difference signal kurtosis, and calculate the basic kurtosis damage; Call the "fundamental frequency damage mapping model", input the main peak value of the differential pressure fluctuation frequency, and calculate the fundamental frequency damage; The second step is to calculate the health state damage coefficient: call the "health state damage modulation model", input the single tower adsorption efficiency index, and calculate the health state damage coefficient. The third step is to calculate the instantaneous fatigue damage value: add the basic kurtosis damage and the basic frequency damage obtained in the first step to obtain a total basic damage value; then, multiply the total basic damage value by the health status damage coefficient obtained in the second step to obtain the final instantaneous fatigue damage value. The fourth step is to update the tower fatigue accumulation: add the instantaneous fatigue damage value calculated in the third step to the tower fatigue accumulation value at the previous moment to obtain the updated tower fatigue accumulation value at the current moment.

7. The method for on-demand energy-saving regeneration of medical compressed air based on dew point feedback according to claim 6, characterized in that: The steps for performing the predictive job rotation in S4 specifically include: When a gas load with a high instantaneous fatigue damage value is detected to occur within a future time window, if the cumulative fatigue degree of the current working tower is higher than the preset tower protection threshold, a protective switch will be performed, and the task will be handed over to another adsorption tower with a lower cumulative fatigue degree.

8. The energy-saving method for on-demand regeneration of medical compressed air based on dew point feedback according to claim 6, characterized in that: Obtain the cumulative fatigue level of the working tower, the cumulative fatigue level of the standby tower, and the most recent p1 historical instantaneous fatigue damage values ​​for prediction; and record the system clock variable of the last protective switch time. The predictive job rotation logic includes a first-level judgment to determine whether a "high-intensity impact is anticipated", a second-level judgment to determine whether the "current working tower is in a vulnerable state", a third-level judgment to determine whether a "better backup tower exists", and a fourth-level judgment to determine whether the "timeout exceeds the switching cooldown time". When the predictive duty cycle logic outputs "Execute", it indicates that the system has simultaneously met the four core conditions: "high-intensity impact was anticipated", "the current working tower is in a vulnerable state", "a better backup tower exists", and "the switching cooldown time is greater than the expected time". This output directly triggers a physical valve switching action. At the same time, it updates the "last protective switching time" variable in the system to the current time. When the predictive job rotation logic outputs "maintain", it indicates that at least one of the four core conditions mentioned above has not been met; the system determines that a protective switch is not necessary or cannot be performed at present, and will maintain the existing working state.

9. A method for on-demand energy-saving regeneration of medical compressed air based on dew point feedback according to claim 7 or 8, characterized in that: The step of selecting and executing the adaptive regeneration program in S5 specifically includes: When the cumulative fatigue of the adsorption tower to be regenerated exceeds the repair threshold, an additional settling period is added to the regular regeneration procedure to perform a restorative regeneration. The adaptive regeneration process is triggered when any adsorption tower completes its adsorption cycle and is about to enter the regeneration phase.

10. A method for on-demand energy-saving regeneration of medical compressed air based on dew point feedback according to claim 9, characterized in that: The resting and recuperation period is defined as the duration of the additional resting and recuperation phase in restorative regeneration. Determine whether the cumulative fatigue of the tower body to be regenerated exceeds the repair threshold; if it does not exceed the threshold, the resting time is 0; if it exceeds the threshold, the resting time is obtained by adding a basic resting time to a fatigue gain time that is linearly scaled according to the fatigue level exceeding the threshold.

Images