CN113317783B

CN113317783B - Multimode personalized longitudinal and transverse calibration method

Info

Publication number: CN113317783B
Application number: CN202110427055.0A
Authority: CN
Inventors: 丁贤根; 丁远彤
Original assignee: Harbour Star Health Biology Shenzhen Co ltd
Current assignee: Harbour Star Health Biology Shenzhen Co ltd
Priority date: 2021-04-20
Filing date: 2021-04-20
Publication date: 2022-02-01
Anticipated expiration: 2041-04-20
Also published as: CN113317783A; WO2022222198A1

Abstract

Due to the diversity of individuals, it is difficult to directly measure and calibrate physicochemical quantities using the same standard in the fields of biomedicine, industrial monitoring, and the like. The multimode personalized longitudinal and transverse calibration method comprises the following steps: 1. obtaining an indication value and a predetermined true value by a low-difficulty method; 2. introducing and acquiring calibration factor values of a function associated with the indication values in a multi-mode manner to calibrate the indication values; 3. and calibrating the correction value of which the indicating value is credible by adopting a longitudinal calibration algorithm of the measuring individual and a transverse calibration algorithm and a calibration function of the group, so that the error of the correction value and the true value is converged within the allowable error. The calibration algorithm comprises the steps of obtaining a calibration parameter set by using Riemann manifold, Kalman filter, deep learning and other methods based on historical and instant indicating values and appointed true values of individuals, and calibrating the individual calibration parameter set based on group data. The invention comprises modes of big data, local area network, single point and the like, and solves the dilemma brought by the traditional method of measuring all individuals by using one standard.

Description

Multimode personalized longitudinal and transverse calibration method

Technical Field

The present invention relates to the field of industrial technology and biomedicine, and in particular to the measurement and calibration of medical and industrial data, and in particular to the calibration calculation of data that cannot be directly measured but is obtained by indirect, multimodal measurements.

Background

The measurement of the physical and chemical quantity comprises direct measurement and indirect measurement, and the direct measurement can be completed by improving the accuracy and precision of the measured physical and chemical quantity; the indirect measurement is limited to the personalized attributes of the person to be measured, and in many cases, it is difficult to directly measure the rational quantity, and sometimes even if the indirect rational quantity data is measured, it is difficult to calibrate and convert the data into the rational quantity data with sufficient accuracy and precision and error meeting the requirements.

For example, it is a very difficult task under the current technical means for the in vitro noninvasive measurement and calibration of the glucose content in human venous blood, how to realize the physicochemical measurement and calibration in the optimized energy-saving control of a cargo fan of an ocean container cargo ship under the condition that the work of an in-cabin refrigerated container is satisfied, and how to realize the physicochemical measurement and calibration in the wind field in the optimal control of a wind driven generator.

For indirect measurement, the inventor simultaneously provides a multi-mode personalized monitoring method, tries to measure the measurement by decomposing the measurement indication value into a plurality of personalized and low-difficulty calibration factor values related to the measurement indication value or a plurality of calibration factor values related to the measurement indication value and measurement indication values with low accuracy and ground difficulty, adopts the multi-mode personalized detection to obtain measurement data, and then calibrates the data, thereby indirectly obtaining the corrected value of the measurement indication value, namely a credible measurement result.

How to obtain a credible measurement result of a correction value of a measurement indication value according to data of indirect measurement, the inventor researches and discovers that the following key steps need to be completed:

1. introducing personalized calibration

1.1, indirect measurement often contains a plurality of interference factors, and must be deducted, otherwise, the error can not be reduced, and a credible measurement result is obtained.

1.2, the individual interference factors existing in each individual are often very different, and need to be calculated and deducted one by one according to individual individuation.

1.3, even for the same individual, the environment-based personalized interference factors are different due to different environments where the individual is located, so that the environment-based personalized interference needs to be deducted.

2. Introducing multi-mode calibration

In the same measuring individual, the rational quantity (indication value) to be measured is often related to other rational quantities (calibration factor values), and there may be a more obvious functional relationship. At the moment, sensors in various modes are introduced to measure corresponding physicochemical quantities, corresponding calculation methods are designed, and measured data are incorporated into calibration calculation, so that obvious effects are brought to indication values.

3. Introducing longitudinal and transverse alignment

The longitudinal calibration of the invention is to calibrate the individual of the measuring object according to the historical data of the individual; the lateral calibration is to perform lateral mutual calibration for several individuals having the same measurement property. This method has significant advantages for the calibration of the indicative values in relation to the environment and in relation to the subdivision classification of the individual.

According to the practical experience of the inventor, the current state of the industry of the current measurement industry mainly uses direct measurement, indirect measurement is less common, and a personalized, multimode and longitudinal and transverse calibration method is adopted and is not found. Especially for the following specific applications, the current situation is:

1. human blood glucose monitoring and calibration current situation

So-called human blood glucose monitoring, defined by the international health organization, is the monitoring of millimole per liter concentration (mmol/L) of glucose content in human venous blood vessels, and currently commonly used methods include:

invasive single-point approach: venous blood extraction assay glucose

Minimally invasive single-point approach: test paper for detecting glucose by pricking finger to take finger capillary blood

Minimally invasive continuous mode: continuous glucose monitoring by inserting indwelling enzyme electrode probe into arm

Non-invasive continuous mode: measuring glucose in interstitial fluid on the skin using electrophoresis, measuring glucose through the skin using infrared, measuring subcutaneous glucose using microwave, measuring glucose in tears using contact lenses with microcircuits, etc. However, until the time of this patent application, no product has been found that has been authorized by medical authorities.

The problem of monitoring blood sugar fluctuation cannot be solved by single-point measurement, and for a type 1 diabetes patient, due to the fact that the short-time rapid drop of blood sugar cannot give an alarm, the life risk of the patient is brought; furthermore, for type 2 diabetic patients, single point measurements do not address optimal management and treatment of diabetes. Whether invasive detection or minimally invasive detection is performed, pain and inconvenience are brought to patients.

2. Cargo bay energy saving monitoring calibration current situation

Until the time of this patent application, no related products have been discovered.

The prior method has the defects of

The inventor considers that the existing calibration algorithm has the following defects:

1. direct measurement cannot be achieved in some application scenarios, and indirect measurement is required.

2. From the measurement principle, the measurement data itself has no information capable of calibrating itself, and cannot calibrate itself.

3. For indirect single-point measurement data which are difficult to measure directly and contain interference factors, the traditional method cannot eliminate the interference factors.

4. The interference factors include common content and personalized content, and the personalized interference can not be eliminated by adopting the same standard.

5. Although some measurements cannot be accurately achieved, the measurement accuracy can be increased by introducing multi-modal calibration variables, but no such method has been found at present.

6. On the premise of large data, there are some commonalities among individuals in a population, and mutual calibration can be achieved through these commonalities, however, no such calibration method has been found so far.

The above disadvantages of the existing calibration method exist in real systems, and a proper indirect measurement and calibration method thereof are needed to solve the problems.

Disclosure of the invention and objects

The inventor proposes a multimode personalized longitudinal and transverse calibration method through long-term observation, experiment and research, and the invention aims and aims to provide a method for the longitudinal and transverse calibration of the multimode personalized longitudinal and transverse calibration, which comprises the following steps:

1. and for the measurement items which cannot be directly and accurately measured or have high direct and accurate measurement difficulty, indirect multi-mode auxiliary calibration measurement data is introduced to realize calibration.

2. For continuous real-time measurement, the invention designs a longitudinal calibration algorithm, and carries out personalized calibration according to historical records and appointed truth values.

3. For a plurality of measuring individuals, the invention provides a transverse calibration algorithm, and the transverse calibration algorithm calibrates the transverse calibration algorithm according to the data of other individuals.

4. Through an artificial intelligence deep learning algorithm, classification calibration is realized for individuals added into a group later, acquisition of an appointed truth value is reduced, and simplified calibration is realized.

Specifically, it states that:

1. the indications referred to in the present invention include, but are not limited to, measured indications.

2. The appointed true value of the invention is not limited to the value obtained by a high-precision measuring device, and can be obtained by artificial intelligence and deep learning.

3. The calibration functions referred to in the present invention are not limited to the functions and formulas listed in the present application, and include other functions and formulas designed by those skilled in the art according to this concept.

4. The numbering of the steps of the present invention, unless otherwise indicated, is not in the order of the numbers.

The scope of application of the invention may include calibration of measurement data and calibration of other data.

It is emphasized that the present invention contemplates that either the type of indication, the type of calibration factor, or the division of individuals and populations may be derived from this concept. The invention application cannot list the kinds and the association of the data information one by one, and the information kind proposed in the invention application is not meant to be a limitation to the idea of the invention.

Advantageous effects of the invention

1. The invention provides an innovative method based on multimode personalized longitudinal and transverse calibration, realizes the invention category and the invention purpose thereof, and solves the defects of the prior calibration technology.

2. The method is based on mathematics, statistics, artificial intelligence deep learning and the like, the calibration of the measured data is realized rapidly and efficiently, and personalized big data resource sharing is realized.

3. Aiming at the in-vitro continuous monitoring similar to human glucose, the method can effectively calibrate and improve the processing accuracy.

4. Aiming at the energy conservation of the cargo hold fan similar to an ocean vessel, the energy conservation can be realized through multi-mode measurement, longitudinal calibration and transverse calibration.

Drawings

FIG. 1 is a schematic diagram of a calibration relationship;

FIG. 2 is a state space diagram;

FIG. 3 Mixed species Raman Spectroscopy;

FIG. 4 glucose Raman shift spectra;

FIG. 5 is a method for energy efficient calibration of a cargo space fan;

figure 6 cargo compartment refrigeration arrangement.

Detailed Description

The purpose and intention of the invention are realized by adopting the technical scheme of the following embodiment:

embodiment I, calibration method for human glucose monitoring data

One of the application embodiments of the invention is a method for diabetes-oriented artificial intelligence personalized tube therapy treatment, which is a typical application example of the invention. In the present embodiment, the method of the present invention is described only, and is not intended as a complete design of an actual system or as a limitation of the present invention.

1. Description of the drawings

FIG. 1 is a schematic diagram of the calibration relationship

Monitoring blood glucose levels from outside the body is an extremely difficult task. According to the World Health Organization (English name: World Health Organization, WHO for short, Chinese is the World Health Organization for short), the World diabetic patients account for about 11.4% of the general population, and the glucose content (blood sugar for short) in the veins of the human body is the golden standard for determining diabetes. At present, conventional blood sugar monitoring is divided into invasive, minimally invasive and non-invasive methods, wherein the invasive method comprises venous blood extraction in a hospital and blood sugar value detection by adopting a biochemical enzyme method; most of minimally invasive methods adopt test paper for finger pricking and finger blood squeezing to detect and prick a wireless sensor on an arm to detect; the noninvasive method is a method which does not penetrate the skin at all, is limited by the development of detection technology, and a relatively popular case is not found at present. In addition, the above methods are mostly single-point measurements, and continuous monitoring of blood glucose levels cannot be achieved.

In this embodiment, the method for detecting glucose in human body based on Raman scattering spectrum, i.e. the indication value of the present invention, is adopted. Since all glucose levels will be detected in the raman laser loop, including at least the levels of all glucose at the epidermal site, intradermal site, interstitial fluid site, capillary site, venous site, etc., and even at other unknown sites, the detected glucose values are the sum of these glucose levels. According to the definition of the world health organization, only the glucose in the vein is needed. Therefore, the blood pressure variation and the blood fat variation are used as calibration factor components, and the glucose component in the venous blood vessel is separated from the total glucose value according to the fluctuation of the blood pressure and the blood fat, so that the longitudinal multimode calibration is provided by the invention, namely the indication value is calibrated according to the historical data and the instant data of an individual to form a correction value. In addition, according to the statistics of big data of a group, for specific situations of different skin colors, race, working properties and the like, the invention also provides transverse multi-mode calibration, namely calibrating the indication value of an individual to be a correction value according to other historical statistical data in the group.

Accordingly, as shown in fig. 1, the present invention proposes to use a riemann manifold coordinate system, the deployment system is a differential geometric function coordinate system, and the blood pressure function is at least calculated as an abscissa axis and a Y axis. In fact, the increase in glucose value, not a vertical increase in the ideal state, is also a non-linear curve according to the large inertia and non-linear analysis of the physiological changes of the human body, the vertical axis X. Only the T-axis time direction is a linear line.

FIG. 2 state equation diagram

The present embodiment is shown in fig. 2 according to a signal system analysis method of a state equation. Including an indication component, calibration factor values, state variables, and longitudinal correction components. Wherein the decomposed DTM (t) is the epidermal blood glucose level DTM₁(t) (Raman shift spectrum line 4001 in FIG. 4), intracutaneous blood glucose level DTM₂(t) (Raman shift spectrum line 4002 in FIG. 4), blood glucose level in interstitial fluid DTM₃(t) (Raman shift spectrum line 4003 in FIG. 4), capillary blood glucose level DTM₄(t) (Raman shift spectrum line 4004 in FIG. 4), and venous blood glucose level DTM₅(t) (Raman shift spectrum line 4005 in FIG. 4), and other interference values DTM₆(t) venous blood glucose level alone DTM, according to the definition of blood glucose by the International health organization (WHO)₅The value of (t) is actually needed, so the DTM needs to be solved here₅(t), i.e. DTM₀(t)＝DTM₅(t)。

It should be reminded that the epidermal blood glucose level DTM₁(t) the degree of dispersion of the fluctuation is large in the case of contamination with glucose-like substances due to the influence of the degree of skin cleanliness at the site irradiated with the Raman laser spot.

Taking DTC (t) as a cardiogram measuring function, carrying out in-vitro measurement according to the period as TC, obtaining PPG, ECG and blood oxygen values, and respectively recording as DTC₁(t)、DTC₂(t)、DTC₃(t)。

In the algorithm of this embodiment, a kalman filter, including fourier transform, laplace transform and Z transform, is further included.

FIG. 3 Mixed species Raman Spectroscopy

Fig. 3 is a spectrum diagram obtained by monitoring raman scattering of a mixed substance, that is, a composition of raman waves of all the detected substances at a raman laser detection point. Wherein 3010 is a horizontal axis showing a shift wave of Raman scattering, 3020 is a vertical axis showing the intensity of Raman wave, and a curve is a composition of Raman waves of all the detected substances at a Raman laser detection point, wherein 3001 is a shift amount of 1003cm ^-13002 is a displacement of 1125cm ^-13003 is the displacement of 1450cm^-1The 3 peaks constitute the characteristic "fingerprint" of glucose. The other waveform is the characteristic wave of the mixed non-glucose substance.

The method of the present invention requires separating the glucose semaphore from the raman scattered wave of the mixture of materials in fig. 3.

FIG. 4 Raman shift Spectroscopy of glucose

FIG. 4 shows a Raman shift spectrum of pure glucose, which is also the amount of glucose signal that needs to be isolated from FIG. 3 according to the present invention. Wherein 4010 is the horizontal axis and represents the Raman scattered shifted wave, 4020 is the vertical axis and represents the intensity of the Raman wave, the graph is the component of the glucose Raman shifted wave, 4001 is the epidermal blood glucose level DTM₁(t), 4002 is the intradermal blood glucose level DTM₂(t), 4003 is the interstitial fluid blood glucose level DTM₃(t), 4004 is the capillary blood glucose level DTM₄(t), 4005 is the venous blood glucose level DTM₅(t) of (d). According to this classification, the final longitudinal correction value DTM₀(t)＝DTM₅(t)。

2. Basic protocol steps

2.1：

A method of multimodal personalized longitudinal and lateral calibration comprising the steps of, and combinations of:

s2000, step: and acquiring the indication value of the individual and the multimode calibration factor value of the correlation function with the indication value according to the time sequence.

S3000, a step: based on a set longitudinal calibration algorithm, the calibration factor values and the indicating values which comprise history and current are adopted, the indicating values are calibrated into longitudinal correction values, and the error of the longitudinal correction values and the agreed true value is ensured to be converged within an allowable error.

For a hospital or institution user, the following S4000 steps may be added in addition to the above-described S2000 and S3000 steps. It should be noted that the numbering of the steps is not restricted to the order of the numerals unless otherwise stated.

And S4000: and acquiring the calibration factor value and the indicating value of a group consisting of more than one individual based on a set transverse calibration algorithm, and calibrating the individual data of the individual according to other individual data to form a transverse correction value.

2.2：

On the basis of the foregoing technical solutions, the present invention may specifically adopt the following or various measures for local improvement in the steps including but not limited to S2000:

step S2010: the measurement module is established according to the formula (2.1), and the calibration module is established according to the formula (2.2).

DTM＝{DTM_αt|DTM_αt，1≤α≤m，1≤t≤n} (2.1)

DTC＝{DTC_βt|DTC_βt，1≤β≤p，1≤t≤n} (2.2)

DTM_a(t)＝DTM_α(t+gT_M) (2.3)

DTC_β(t)＝DTC_β(t+hT_C) (2.4)

Wherein:

the measurement module comprises more than one indicating value component, and the indicating value component comprises more than one indicating value acquired according to the time sequence.

The calibration module comprises more than one calibration factor component, and the calibration factor component comprises more than one calibration factor value acquired according to the time series.

DTM is the set of said indications, also the measurement module, DTM_αtThe time sequence with the indicating value component number alpha is the indicating value of t, m is the maximum number value of the indicating value component, and n is the current value of the time sequence.

DTC is the set of calibration factor values, also the calibration module, DTC_βtThe time series with the number of the calibration factor component is beta is the value of the calibration factor with t, and p is the maximum number value of the calibration factor component.

Formula (2.3) is a time function of the indicating value component, alpha is the number of the indicating value component, g is any integer, T_MIs the period of the indication component.

Equation (2.4) is a time function of the calibration factor component, β is the number of the calibration factor component, h is any integer, T_CIs the period of the calibration factor component.

As shown in FIG. 1, in this embodiment, the measurement module includes an indication value of glucose (abbreviated as "blood glucose") of a human body, and a human bodyThe body blood fat (abbreviated as blood fat) indicates the value component. The calibration module comprises a pulse calibration factor component and a blood pressure calibration factor component. The indication value is the set of blood sugar and blood fat indication values and is the DTM, the calibration factor value is the set of pulse and blood pressure and is the DTC, and the correction value is the finally needed blood sugar value. The system establishes a Riemann manifold (geometric) coordinate system comprising X, Y, T, wherein the abscissa X takes the value of the calibration factor DTC_β(t) curve; the vertical coordinate Y is the amplitude of fluctuation, and in a human body, the Y axis comprises a nonlinear curve and a linear line; the depth axis takes the time series T, the axis being a straight line.

S2020, step: the correlation function includes a constraint relationship according to equation (2.5) and equation (2.6):

when TC_β1＜TM_α1Taking TC_β2≤TM_α2＜TC_β1 (2.5)

When TC_β1≥TM_α1Taking TC_β2＜TM_α1And TM_α2＜TM_α1 (2.6)

Wherein:

TC_β1a period of change, TC, of said calibration factor value of said calibration factor component number β_β2A sampling period, TM, of the calibration factor value numbered beta for the calibration factor component_α1For the variation period of the indication value with the indication value component number alpha, TM_α2The sample period of the indication value with the indication value component number alpha.

In this embodiment:

taking the variation period of the pulse in the calibration module as TC₁₁The sampling period of the pulse is TC₁₂The period of change of blood pressure is TC₂₁The blood pressure sampling period is TC₂₂。

Taking the variation period of blood sugar in the measurement module as TM₁₁The sampling period of blood glucose is TM₁₂The period of change of blood lipid is TM₂₁Blood lipid sampling period is TM₂₂。

TC is obtained according to the measurement of the change cycle of the actual pulse, blood pressure, blood sugar and blood fat of human body₁₁、TC₂₁、TM₁₁，TM₂₁The sampling period is based on the ability of the sensor and the sampling period TC of the pulse, the blood pressure, the blood sugar and the blood fat₁₂、TC₂₂、TM₁₂，TM₂₂The respective selection ranges are: 1ms to 100ms, 20ms to 120 s. The unit ms is milliseconds and s is seconds.

In the present embodiment, the specific functional relationship between the components is calculated according to the medical functional relationship between the pulse, blood pressure, blood sugar and blood fat of the human body.

And S2030 step: when TC_β1＜TM_α1Simultaneous TM_a2＜TC_β1When it is, the TM_α2Decomposition into short sampling periods TM_α2SAnd a long sampling period TM_α2LOf, with TM_α2LObtaining the indication value component DTM with slow response speed but higher measurement accuracy_αLBy TM_α2SObtaining the indication value component DTM with high corresponding speed but low measurement precision_αSUsing DTM_αSWith DTC_βThe correlation function is established, and the DTM is calibrated_αLThe step (2).

In this embodiment: taking TC_β1The pulse period is usually about 1 second, and TM is taken_α1Setting a pulse sampling period TM for a period of blood glucose variation, at least of the order of minutes_α250 milliseconds (mS), it is impossible to take the order of milliseconds, that is, TM, according to the current state of the art for sampling blood glucose values_α2＜TC_β1. To further optimize the calibration of the blood glucose level, the invention is designed to combine the length of the sampling period with the set of the blood glucose sampling period, for example, taking TM_α2SAt 200 ms, take TM_α2LFor 30 seconds, in blood glucose data obtained by sampling 200 milliseconds, the inventor knows that the measurement precision is insufficient and the signal-to-noise ratio is large, but enough influence information is provided for the influence trend of the blood glucose due to the fluctuation of the pulse. In blood glucose data obtained by sampling for 30 seconds, measurement accuracy and signal-to-noise ratio are sufficient, but due to the fact that the sampling period is far longer than 200 milliseconds, due to the influence of integration, the calculation cannot be carried out according to the Fourier transform principleThe influence of pulse fluctuation on blood glucose level. Therefore, a mode of combining the sampling period length is introduced, the fluctuation function of the pulse fluctuation on the blood sugar is found out by using the short-period sampling, and the blood sugar value sampled in the long period is calibrated, so that the influence of the pulse fluctuation on the blood sugar can be found out, the blood sugar measurement precision is ensured, and the method is an optimized choice.

S2040: the correlation function includes establishing a constraint relationship from equation (2.7) to equation (2.12), specifying the DTM₀(t) the longitudinal correction value as the calibration factor value:

establishing an input state space with formula (2.7) as a system and a state space with formula (2.8) as the system, establishing a state equation system with formula (2.9) and formula (2.11), calculating the longitudinal correction value according to formula (2.13) or formula (2.14), reviewing the calculation result according to formula (2.12), and establishing a kalman filter to include the indication variable and the calibration factor value variable as input variables, establishing a state observer to predict and calculate the longitudinal correction value:

DTM(t)＝[DTM₁(t)，DTM₂(t)…DTM_m(t)]^T (2.7)

DTC(t)＝[DTC₁(t)，DTC₂(t)…DTC_p(t)]^T (2.8)

Z(t)＝[Z₁(t)，Z₂(t)…Z_q(t)]^T (2.9)

DTM(t)＝Ψ(Z(t)，DTC(t)，t) (2.11)

DTM₀(t)∈{DTM_α(t)|DTM_α(t)，1≤α≤m} (2.13)

DTM₀(t)＝DTM(t)-R(t) (2.14)

wherein:

equation (2.7) is derived from DTM₁(t) axis, DTM₂(t) Axis … DTM_m(t) a function of the state space of the indicative value component of the m-dimensional state space formed by the dimensions (t). Equation (2.8) is represented by DTC₁(t) axis, DTC₂(t) Axis … DTC_p(T) axis, T being the transpose of the matrix, and formula (2.9) being a function of the state space of the calibration factor component of the p-dimensional state space formed by the Z axis₁(t) axis, Z₂(t) axis, … Z_qQ formed by the (t) axis is a function of the state space of said state variable of the state space, equation (2.10) is a differential equation of the state variable, equation (2.11) is an output equation of said system, equation (2.12) is said longitudinal correction value function DTM₀Error e of (t)₀(t) is less than or equal to the allowable error e₀Formula (2.13) is that the longitudinal correction value belongs to one element of the indicating value set, and formula (2.14) is that the longitudinal correction value is the calculated value of the output equation of the system minus the value of the error correction function r (t).

As shown in fig. 2, this is a system deployed according to the principles of state space equations. Fig. 3 is a graph showing a raman scattering principle, in which infrared laser is projected to the skin of a human body, and the content of glucose is measured by detecting raman scattered waves scattered by glucose molecules in a raman scattering effect due to the fact that the infrared laser can be transmitted to subcutaneous tissues and the glucose molecules are encountered in a transmission path. Since the glucose measured in this process includes the total glucose content at the epidermal site, the intradermal site, the interstitial fluid site, the capillary site, the venous site, and the like. In this embodiment, the values of the indicating value and the calibration factor are selected differently from the steps S2010 and S2020, in order to provide another partial improvement of the method of the present invention. The method comprises the following steps:

1. and obtaining indication values according to a period TM by using DTM (t) as a function for measuring blood sugar (glucose in vivo) in vitro of a human body, for example, by adopting Raman laser-based in vitro monitoring, infrared light in vitro monitoring, electrophoresis in vitro enzyme electrode monitoring and the like.

2. In FIG. 4Resolution of DTM (t) is the epidermal blood glucose level DTM₁(t) (Raman shift spectrum line 4001 in FIG. 4), intracutaneous blood glucose level DTM₂(t) (Raman shift spectrum line 4002 in FIG. 4), blood glucose level in interstitial fluid DTM₃(t) (Raman shift spectrum line 4003 in FIG. 4), capillary blood glucose level DTM₄(t) (Raman shift spectrum line 4004 in FIG. 4), and venous blood glucose level DTM₅(t) (Raman shift spectrum line 4005 in FIG. 4), and other interference values DTM₆(t) venous blood glucose level alone DTM, according to the definition of blood glucose by the International health organization (WHO)₅The value of (t) is actually needed, so the DTM needs to be solved here₅(t), i.e. DTM₀(t)＝DTM₅(t)。

3. Taking DTC (t) as a cardiogram measuring function, carrying out in-vitro measurement according to the period as TC, obtaining PPG, ECG and blood oxygen values, and respectively recording as DTC₁(t)、DTC₂(t)、DTC₃(t)。

4. From DTM, according to equations (2.7) to (2.13)₁(t) to DTM₆(t) selecting the closest value as DTM₀(t), that is, the longitudinal correction value, or, the longitudinal correction value is calculated according to the formula (2.7) to the formula (2.12) and the formula (2.14).

Step S2050: the correlation function further comprises a step of recording the one-way change and/or the two-way change, specifically:

a change in one or more of the said indicators results in a unidirectional change in one or more other of the said calibration factor values, or a change in one or more of the said calibration factor values results in a unidirectional change in one or more other of the said indicators, or a change in either one of the two results in a bidirectional change in the other, the said changes comprising a functional relationship as determined by equation (2.7) or equation (2.8) or equation (2.9).

Wherein:

f_2.7is a bidirectional mapping function, operation sign

The indicator value is affected by a change in the value of the calibration factor, which is a two-way mapping operator.

f_2.8Is a left-hand mapping function, operator

The left-hand mapping operator, i.e. a change in the value of the calibration factor will affect the value of the calibration factor, whereas a change in the value of the calibration factor will not affect the value of the indication.

f_2.9Is a right-mapping function, operator

The right-hand mapping operator, i.e. a change in the value of the calibration factor will affect the indication, whereas a change in the indication will not affect the value of the calibration factor.

In the embodiment, according to the pulse, blood pressure and the properties of blood sugar and blood fat of the human body, the change of the pulse and blood pressure will affect the change of the blood sugar and blood fat based on the above sampling period setting, and the change of the blood sugar and blood fat will not affect the change of the pulse and blood pressure, that is, the mapping relationship of the formula (2.9) is satisfied. In other embodiments, there will be a mapping that conforms to equation (2.7) or equation (2.8), depending on the particular situation.

Step S2060: the correlation function further includes a relationship of the indication function being non-periodic and the calibration factor value function being periodic, including the T in the formula (2.3)_MTaking infinity and said T in said equation (2.4)_CAnd taking the period of the calibration factor value function.

In this case, the present implementation conforms to this definition, T_MThe period of change of blood sugar and blood fat, T_CThe blood pressure and pulse change period is actually a non-periodic function of blood sugar and blood fat.

Step S2070: the correlation function also comprises the relationship between the indicating function with linear time variation and the calibration factor value function with linear time invariance, including the relationship between the output and the input of the function in the formula (2.3) is linear, the relationship between the response of the function and the input is related to the time, the relationship between the output and the input of the function in the formula (2.4) is linear, and the relationship between the response of the function and the input is not related to the time.

As mentioned above, in this case, this definition is not met by the present embodiment, but for other embodiments, such as industrial measurements combined with long and short periods, this definition is met.

Step S2080: the correlation function also includes the relationship of the indication function of causality-stability and the calibration factor value function of causality-stability, including that the output of the functions in equations (2.3) and (2.4) is independent of the input after the output instant.

This embodiment is partly in accordance with this definition, and the change of blood sugar and blood fat partly conforms to the causal-stability function relationship of blood pressure and pulse, but blood pressure and pulse are not the only dependent variables of the change of blood sugar and blood fat, diet and exercise are the other dependent variables of blood sugar and blood fat, and are dependent variables with larger weight.

S2090: the correlation function also includes a mapping between the set of calibration factor values and the set of indicative values that is difficult to determine, but obtained from analysis by big data and artificial intelligence algorithms.

In fact, in this embodiment, we can design more calibration factor components in the calibration module, such as carbohydrate in diet, sugar content, exercise amount, etc. Through the statistics of big data, more accurate calibration will be brought about for each individual calibration.

And S2100 step: the correlation function further comprises establishing a Riemann popular coordinate system by using the calibration factor value, the indication value and the time sequence, and establishing a relation function among the calibration factor value, the indication value and the time sequence by using Riemann geometry so as to avoid the problem that the mapping relation among the calibration factor value set and the indication value set generates non-one-to-one mapping in the step S2070.

In this embodiment, the algorithm can be designed completely by using the blood pressure as the standard for calibration, that is, by using the method of riemann manifold as shown in fig. 1, and by using the blood pressure value as the horizontal axis of the riemann coordinate, that is, the X axis.

Step S2110: and storing historical and current time points of all the individuals in the group, the calibration module, the calibration factor value, the measurement module and the indication value into a database, wherein the historical time points comprise partial time points or all time points before the current time point.

In the embodiment, the database is stored in the cloud center, and the user is connected with the cloud mode to operate through the wireless mobile phone terminal. 2.3:

on the basis of the foregoing technical solutions, the present invention may specifically adopt one or more of the following measures for local improvement in the steps including but not limited to S3000:

and S3010: and calculating the current indication value into the longitudinal correction value by adopting a calibration function according to the current calibration factor value and the indication value, so that the error between the longitudinal correction value and the default true value is smaller than the allowable error.

For example, the current blood sugar and blood fat are calibrated by adopting the current pulse and blood pressure, the current blood sugar and blood fat are related according to the current value and fluctuation of the pulse and blood pressure by adopting a partial differential equation, and the fluctuation of the pulse and blood pressure on the blood sugar and blood fat is obtained by differentiation so as to reduce interference factors in the blood sugar and blood fat data and obtain higher calibration precision.

And S3020: and setting weights according to more than one historical calibration factor values and indication values, the front and back sequence of the time sequence and waveform attributes, and calculating the current indication value into the longitudinal correction value by adopting the calibration function so that the error between the longitudinal correction value and the appointed true value is smaller than the allowable error.

Step S3030: the longitudinal calibration algorithm further comprises steps executed in sequence from S3031 to S3033, and specifically comprises the following steps:

step S3031: and the individual performs synchronous calculation on the indicating value according to the same time period by adopting the calibration function on the time sequence of the same time period according to the fluctuation of the calibration factor value, so as to obtain a longitudinal synchronous correction value of which the error between the individual and the agreed true value is smaller than an allowable error.

Step S3032: and the individual uses the longitudinal synchronization correction value as an independent variable in the time sequence of different time periods, enlarges the time period to include a day, a week, a month, a year or a user-specified period, adopts the calibration function to calculate and obtain a longitudinal history correction value, and calculates the error between the longitudinal history correction value and the longitudinal synchronization correction value.

Step S3033: and reducing or eliminating the error between the longitudinal historical corrected value and the longitudinal synchronous corrected value by adopting a statistical algorithm and an artificial intelligence classification algorithm.

In this example, errors are analyzed using a T-test or Z-test to eliminate outliers. And classifying the historical values by adopting a Support Vector Machine (SVM) algorithm and a Convolutional Neural Network (CNN) algorithm.

The time period includes more than one consecutive time points or all the time points within one cycle of the calibration factor value in the calibration module.

Step S3040: the longitudinal calibration algorithm further comprises: and establishing a calibration parameter set between the calibration factor value and the indicating value by adopting a statistical algorithm, an artificial intelligence classification algorithm and the historical calibration factor value and the indicating value, and calibrating the current indicating value into the longitudinal correction value according to the calibration parameter set and the current calibration factor value and the indicating value.

S3050: the longitudinal calibration algorithm further comprises a step of obtaining an agreed true value, wherein the individual is measured by using monitoring equipment with precision, accuracy or precision level higher by more than one level and measurement error smaller than the allowable error to obtain a measurement value as the agreed true value, and the ratio of the measurement error smaller than the standard monitoring equipment to the monitoring error of the method is smaller than 0.5 or a value specified by a user. The step of obtaining the agreed truth value specifically comprises the step of obtaining through calculation by adopting historical data of other individuals.

In this embodiment, a medical-grade blood sugar and blood fat blood drawing test measuring device is used to obtain individual blood sugar and blood fat measurement values and record measurement times, which are used as an agreed true value. Meanwhile, in the same time interval, the numerical values and the time sequence of blood sugar and blood fat measured by the equipment to be calibrated are adopted to calculate the longitudinal synchronous correction value, the error between the longitudinal synchronous correction value and the defined value is calculated and verified, and if the error is greater than the allowable error, the parameter set is modified and the calculation is iterated until the error is less than the allowable error.

The appointed true value can also be obtained by adopting methods such as statistics and artificial intelligence algorithms for other individuals, and at the moment, the appointed true value is obtained without adopting higher-level medical equipment for measurement.

In this embodiment, the calibration function may use algorithms such as a mathematical statistics algorithm, a support vector machine SVM, a convolutional neural network CNN, and the like, to train, for example, a convolution kernel or related parameters, which generalize and calibrate the parameter set, with an error smaller than an allowable error as a target.

2.4：

On the basis of the foregoing technical solution, the method of the present invention further includes, but is not limited to, the step S4000, and may specifically adopt one or more of the following measures for local improvement:

s4010 step: establishing a group consisting of more than one individual with the same measurement attribute, acquiring the calibration factor value and the indication value of the individual in the group, and respectively incorporating the standard module and the measurement module, wherein the group comprises other individuals and self-individual.

In this embodiment, the candy friend groups are established for cities, districts and hospitals, and can also be established according to gender, skin color, occupation and the like. From the point of view of statistics and artificial intelligence big data, the more candy friends in the population, the better the effect of the lateral calibration.

S4020: the transverse calibration algorithm comprises the step of step-by-step implementation, and specifically comprises the following steps:

the first step is as follows: executing the longitudinal calibration algorithm for all of the individuals in the population, and obtaining the longitudinal correction values and the errors of all of the individuals, wherein the longitudinal correction values comprise the longitudinal synchronous correction values and the longitudinal historical correction values.

The second step is that: and calculating a calibration parameter set by adopting the calibration functions including a weighting algorithm, a filtering algorithm, a partial differential algorithm, a fuzzy algorithm, a differential geometry algorithm and an artificial intelligence deep learning algorithm according to the longitudinal correction values and the errors of the other individuals so as to calibrate the longitudinal correction values of the individuals, so that the errors of the longitudinal correction values are smaller than the allowable errors, and acquiring the transverse correction values of the individuals.

Step S4030: the transverse calibration algorithm further comprises a single step implementation step, specifically:

and calibrating the indicating value of the self-individual by adopting the calibration function according to the calibration factor value and the indicating value of all the other individuals in the group and the calibration factor value and the indicating value of the self-individual, and acquiring the transverse correction value and the error of the self-individual, wherein the error of the self-individual is smaller than the allowable error.

In the present embodiment, in software design, single-step implementation is adopted, and in practice, single-step implementation may be understood as that the distribution is implemented in the same piece of software code.

S4040: the transverse calibration algorithm further comprises the step of circularly implementing, and specifically comprises the following steps:

and circularly executing the step S4020 or the step S4030 according to a historical time period to acquire the lateral correction value and the error of the self individual, wherein the period of the circular body is selected to be between a time period smaller than the expansion time period and larger than the time period according to the measurement attribute.

In this embodiment, in the cloud center, on the server, the calculation result is stored in the database by adopting the cycle implementation, and when the calculation result is required on the mobile phone of the user terminal, the horizontal calibration result is directly obtained in the cloud center database.

S4050: the transverse calibration algorithm further comprises a classification step, specifically:

and classifying the individuals in the group according to the calibration parameter set, selecting the other individuals of the same class or similar classes to perform calibration calculation when the individuals perform transverse calibration, and acquiring the transverse correction value and the error of which the error of the individuals is smaller than the allowable error.

S4060: the transverse calibration algorithm further comprises a calibration step, specifically:

s4061 step: calculating individual errors of the individuals in the population, and establishing individual error categories and error correction parameters with the errors smaller than the allowable errors according to the individual errors by adopting the error categories and the error severity.

S4062: and circularly executing the step S4061 according to the historical time period to obtain the current individual error category and the current error correction parameter of the individual.

S4063: for the newly added individual in the population, performing the step S4061, or performing the steps S4061 and S4062.

S4064: and eliminating the individual error according to the individual error category and the error correction parameter of the individual or the current individual error category and the current error correction parameter.

In this embodiment, the efficiency of calibration can be further improved and the error can be reduced by using the inter-population lateral calibration, and the calibration is performed specifically for the situations due to the biodiversity of the human body, such as different skin colors, different ages, different professions, and the like, so as to obtain better effect.

Further, a group classification model is established, a classified calibration parameter set is established, calculation is carried out on the individuals newly entering the group to be included in the classification, and the calibration parameter set corresponding to the classification is used for carrying out rapid calibration.

2.5：

On the basis of the foregoing technical solutions, the present invention may specifically adopt one or more of the following measures for local improvement in steps including but not limited to S5000:

step S5010: according to the formula (2.13), the calibration function comprises t-test and Z-test methods, and particularly under the condition that the error of the indication value is smaller than the allowable error, the method comprises the following steps or the combination of the following steps:

and acquiring one appointed true value as an average value of the time period, acquiring more than one indication value, calculating the error according to a formula (5.1) or a formula (5.2), calculating the standard deviation according to a formula (5.3) or a formula (5.4), calculating the calibrated error according to a formula (5.5), checking that the calibrated error is smaller than the allowable error according to a formula (5.6), and constraining the corrected value of the calibrated indication value according to a formula (5.7) as a constraint condition.

Wherein:

and

agreed truth values, e.g., for the indication component and calibration factor value component, numbered alpha and beta, respectively_αtFor the error of the indication, e_βtAs an error in the value of said calibration factor, δ_αtIs the standard deviation, δ, of the indicative component_βtIs the standard deviation of the calibration factor component, m is the total number of said indications over said time period,

is a correction value after the calibration is performed,

is the corrected value error of the calibrated indication, e₀Is the allowable error.

If m is less than 30, then pair delta_αtAnd delta_βtA statistical t-test was performed to calibrate outliers.

If m is greater than 30, then pair δ_αtAnd delta_βtA statistical Z-test was performed to calibrate outliers.

And deleting abnormal values or carrying out interpolation calculation to supplement the indicating values, and taking all the indicating values and the indicating values subjected to interpolation calculation as the longitudinal correction values.

In the present embodiment, after an abnormal value is deleted, interpolation calculation is performed in the time series to supplement a point, and the interpolation algorithm includes an arithmetic median, an arithmetic mean, a curved median, a curved mean, a least square method, a gaussian method, a neural network method, and the like.

S5020: according to the formula (2.14), the calibration function further includes steps of calculating an error correction function r (t) from steps S5021 to S5023:

s5021: according to the step of obtaining the agreed truth value, a plurality of time periods are selected for the individual to obtain the agreed truth value, R (t) is taken as 0, and the longitudinal correction value is calculated by adopting a function fitting method.

S5022: executing the step S5021 for a plurality of individuals in the group, and obtaining the corresponding table calculated by the formula (2.14).

S5023: and calculating R (t) by adopting a statistical method or a deep learning or neural network method aiming at the corresponding table.

In the embodiment, in the initial stage of big data, a large amount of data with different time periods of individuals collecting the engagement truth value are needed, so that after the stage, deep learning is used for reasoning and classifying, and therefore the action of collecting the engagement truth value is reduced, and the actual cost expenditure and workload are reduced. The function fitting herein includes an arithmetic median, an arithmetic mean, a curve median, a curve mean, a least square method, a gaussian method, a neural network method, and the like.

Step S5030: the calibration function also comprises a set mapping method, and under the condition that the error of the indicating value is greater than the allowable error, the method comprises the following steps:

and selecting the indicating value component and the calibration factor component which are mutually associated according to the association function, and establishing a corresponding mapping relation function of set elements according to the calibration factor value set of the calibration factor component of the time sequence and the indicating value set of the indicating value component in the same time period.

And performing partial differential operation on the time sequence according to the corresponding mapping relation function, selecting a component having a linear relation with the fluctuation in the calibration factor value set according to the fluctuation of elements in the indicating value component set, taking the component as a main indicating value component, and taking the rest as an auxiliary indicating value component.

And calibrating the abnormal value of the principal component according to the t-test or the Z-test.

In this embodiment, the algorithm is a simplified calculation method, and data for obtaining a plurality of agreed true and indication values can be used, and if they satisfy a certain linear relationship or a fittable curve relationship, the calculation workload can be greatly reduced by using a geometric mapping method. The precondition for the investigation is that statistical methods or convolutional neural networks are used for verification.

And S5040, step: the calibration function further includes a state equation method, specifically, solving the solutions in the output equations of the system of the formula (2.11) and the formula (2.12) in the step S2030, and decomposing all the output components of the system to one of the output components DTM₀(t) as said longitudinal correction value, the other output components are regarded as disturbance values.

As shown in fig. 2, in the present embodiment, the DTM is calculated by the step including S2030₀(t) blood glucose level.

S5050, the calibration function further includes a neural network method, specifically:

and aiming at the relational data records in the database, taking the data records as neurons, and establishing a connection function between the neurons according to a calculation result comprising the mathematical model to form more than one layer of neural network.

And dividing and establishing exciting type, inhibiting type, explosion type and plateau type connection subfunctions according to the effect of the indicating value and the calibration factor value on the error in the connection functions, wherein the connection subfunctions comprise constant type weight coefficients and functional type weight coefficients.

And optimizing the connector functions by adopting a deep learning algorithm, including supervised learning, unsupervised learning and reinforcement learning algorithms.

And classifying and screening the error and the output component by adopting a support vector machine algorithm, and screening the longitudinal correction value and the interference value.

And carrying out convolution, activation, pooling, full connection and training of the connection subfunction by adopting a convolution neural network algorithm so as to screen out the longitudinal correction value, the interference value and the corresponding indication value.

And establishing an in-layer association function by adopting a recurrent neural network algorithm, and training the connection subfunction to screen out the longitudinal correction value and the interference value.

And establishing an interlayer correlation function for the indicating value, the calibration factor value and the error among the layers of the neural networks by adopting a deep neural network algorithm, and training the connection subfunction to screen out the longitudinal correction value and the interference value.

And training the connector functions by adopting a feed-forward neural network algorithm under the condition that each neuron is only connected with the neuron of the previous layer so as to screen out the longitudinal correction value and the interference value.

And training the connector functions by adopting a feedback neural network algorithm under the condition that each neuron is only connected with the neuron of the next layer so as to screen out the longitudinal correction value and the interference value.

In this embodiment, the specific steps are:

1. and performing single-substance classification and single-substance Raman spectrum model database establishment on substances existing in the skin irradiated by the Raman laser spots.

2. The selection includes a Convolutional neural network model (english name: Convolutional neural Networks, abbreviated as CNN), deep learning is performed on the model database, and a convolution kernel and a calibration parameter set are calculated.

3. Establishing an initial learning group consisting of a plurality of people (100 people in the example), collecting the appointed true value and the corresponding Raman spectrum indication value of the initial learning group and the calibration factor value database of blood pressure and blood oxygen, executing deep learning, and calculating and optimizing a convolution kernel and a calibration parameter set.

4. And expanding the population number, performing spot check of an appointed true value for newly added individuals according to probability, executing deep learning, and calculating and optimizing a verification convolution kernel and a calibration parameter set.

5. When the error of the spot check is lower than the index of the State food and drug administration for the in vitro noninvasive glucometer, the system is formally released, deep learning is continuously executed, and the optimized verification convolution kernel and the calibration parameter set are calculated.

And S5060, establishing a state observer by using a Kalman filtering method, calculating the output of the state observer by using the indicating variable and the calibration factor value variable as input variables, comparing the output with the longitudinal calibration value, and analyzing a comparison result by deep learning to correct the calibration parameter set so that the error converges with the allowable error.

And S5070: the calibration function further comprises calculating the value of the indication value minus the auxiliary indication value component in the subsequent time period as the longitudinal correction value by performing statistical calculation on the auxiliary indication value component if the auxiliary indication value component is found to be consistent in other time periods.

And S5080: the calibration function further comprises simplified calculation by adopting an optimized calibration parameter set to obtain the longitudinal correction value.

In this embodiment, according to the population for a specific classification, for example, the classification conditions are: huang breeder + retired woman + Shenzhen regionLive people and people with the age between 55 years and 75 years are matched with the statistical worker to sequentially establish the classified sub-databases, and because the Raman spectrum data of the live people are stable, the blood glucose value DTM of the epidermis is relatively stable₁(t) is relatively stable and fluctuates to a lesser degree, and thus, DTM is defined₁(t) to DTM₄And (t) is an auxiliary indication value, and simplified calculation is carried out.

2.6：

On the basis of the foregoing technical solutions, the present invention may specifically adopt one or more of the following measures for local improvement in steps including but not limited to step S6000:

s6010: and sorting the calculation parameters in the calibration function according to the time sequence, establishing a fixed data format, and recording the time sequence and the algorithm number to form a calibration parameter set record.

Step S6020: the calibration parameter set includes all of the calibration parameter set records accumulated according to the time series.

And S6030: the calibration parameter set also comprises data of algorithm itself, constants, coefficients, weight coefficients, process parameters and time sequence based on the artificial intelligence algorithm.

And S6040: and performing iterative optimization upgrading by adopting a statistical method or an artificial intelligence method according to the calibration parameter set and taking the minimum error, the minimum standard deviation and the minimum calculated amount as optimization indexes to form the optimized calibration parameter set with the time sequence label.

And S6050: a step of storing the calibration parameter set in the database.

In this embodiment, the calibration parameter set is incorporated into the synchronization calculation along with the convolution kernel, as previously described.

2.7：

On the basis of the foregoing technical solutions, the present invention specifically includes, but is not limited to, a single step or a combination of multiple steps of an S7100 cloud big data mode, an S7200 local area network mode, and an S7300 single point mode in the S7000 step:

s7100, cloud big data mode:

and S7110, establishing a cloud center based on an Internet mode, transmitting all the acquired data including all the groups and the individuals, the intermediate calculation result and the database obtained by the method through a wide area network, and storing all the data on more than one cloud server based on the Internet, so that the data can be used as more than one cloud center to manage, calculate and support in the cloud calculation mode.

S7120, more than one cloud center is established by adopting a block chain mode to store, manage and support the database and the steps, the individuals adopt anonymous records, information in the database adopts a chain structure with a timestamp, users access the database and adopt encryption and decryption communication, the information supports tamper resistance, and a repudiation-proof, multi-center and centerless mode is supported.

S7130, establishing, managing and supporting more than one mechanism in a safe multi-party computing mode, performing appointed computing according to the database content of each mechanism on the premise of not exchanging the database core data of the cloud center of each mechanism, and sharing the obtained computing result by the participating mechanisms; the organization comprises more than one cloud center and manages more than one individual; the secure multi-party computation includes: public key mechanisms, hybrid circuits, careless transmission, secret sharing, privacy protection set intersection protocol, homomorphic encryption, zero knowledge proof and a method without a trust center so as to enhance the safety of information and protect the privacy of objects.

And S7140, establishing and training a model for the non-emphasized object privacy protection by adopting a centralized learning mode, wherein the database is stored in a cloud center.

And S7150, establishing and training a model when the object privacy protection needs to be emphasized by adopting a federal learning mode, wherein the model training is carried out among more than one stored cloud centers, and data of the cloud centers are not exchanged.

Step S7200, local area network mode:

the server based on the local area network is established for storing and managing the support center, and the monitored data including all the groups and all the data of the individuals, the intermediate calculation results and the database are transmitted through the local area network and are all stored on the server based on the local area network, so as to manage, calculate and support.

Step S7300, single point mode:

the single point is a step of monitoring detection, storage, management, calculation and support of one individual, and all data, intermediate calculation results and the database of the individual acquired by the invention are all stored in the storage of the single point.

In this embodiment, a cloud big data mode is adopted, and the personal terminal device is connected with the smart phone through bluetooth and then is accessed to a cloud big data center to work.

Embodiment II, calibration method for energy conservation of container ship cargo hold fan

1. Brief introduction to the basic scheme

This example is given as an example of the application of the present invention in the industrial field. The energy-saving control system is a multi-input multi-output control system for the high-power fan of the cargo hold in an ocean container ship, wherein input quantity comprises ambient temperature, wind speed, set temperature in the cargo hold, an integral wind field of the cargo hold and the like of all refrigerated containers stacked in the cargo hold, output control comprises frequency control of all the cargo hold fans, a wind channel air outlet air door, a cargo hold air inlet, a cargo hold air outlet, control signals of the refrigerated containers and the like, and overall control indexes are that under the condition that quality guarantee of goods in the refrigerated container is ensured and the temperature in the cargo hold set by a user is not changed, energy consumption of the cargo hold fans is reduced as much as possible. According to actual system operation data, the method for multi-mode longitudinal and transverse calibration can achieve reasonable and stable fan control and energy-saving effect as high as 58%.

2. Description of the drawings

FIG. 5: as shown in 5010, is a cargo space in which a refrigerated container (cooler) is stacked, and in the refrigerated container, the cooler has a closed-loop temperature control system, for example, a temperature control system using a PID controller. In this temperature control system, signals including an internal set temperature, a supply air temperature, a return air temperature, an external ambient temperature, etc. of the refrigerated container are acquired from an external port of the refrigerated container through an external communication device, or the signals are acquired through an internal communication circuit, the signals are output as a number 5036, input as a negative feedback signal as a number 5032 to an input of the system, and form a complete feedback signal as a number 5043 together with a system given signal as a number 5031 and an AI given signal as a number 5033, a fan control subsystem as a number 5020 is driven to control the fan, and finally a method of control theoretical closed-loop control as a number 5035 is formed. That is, according to the collected data, the control function is generated as the negative feedback data of the closed loop, and the fan or the fan and the gate are directly controlled. It is noted, however, that the closed loop is referred to herein as a relatively short time difference, e.g., less than 20 minutes, for the process time from acquisition to control. It should be noted that 20 minutes is only an example, and this number may be longer or shorter depending on the size of the compartments.

FIG. 2: the control method associated with fig. 5 is a multiple-input multiple-output state space control method as in fig. 2.

FIG. 6: the number 6010 and the number 6020 are vertical planes on which the box heads of the refrigerated containers stacked in one cargo compartment of the ship are located, respectively, and the box head of the refrigerated container is the head at which the refrigerated container is located on the operation panel of the refrigerating machine. In the ship, the containers are stacked with a three-dimensional coordinate number, namely, a line number Bay, a column number Row and a layer number Tier, which are called BRT coordinates for short. In fig. 6, white squares as indicated by

numbers

6012, 6022 are positions of the bottom ballast water tanks, and gray squares as indicated by

numbers

6011, 6021 are heads of the refrigerated container stacks, in which the air temperature at the heads varies with the operation of the refrigerated container. For convenience of description, the data of the air temperature or the air discharge amount and the air discharge speed of the cooling fan at the BRT coordinate of the position point where the refrigerated container is located are indicated by C1 to C9, and it should be noted that the data are continuous quantities, and the differences of the continuous quantities from small to large are indicated by C1 to C9 in fig. 5, but the continuous quantities are not limited to 9 levels of C1 to C9. The data of one frame picture is collected once for a container in a cargo hold and stored in an environment data database according to BRT coordinates. For other active, passive, or non-containerized cargo, and so on.

3. Description of differentiation

The same points as the first embodiment will not be repeated here, but the following points are different:

3.1、

setting indication values at least comprising position coordinates of refrigerated containers in the cargo hold, ambient temperature, wind speed, set temperature in the cargo hold, air outlet temperature, air supply outlet temperature, cargo hold fan running state and total energy consumption of the cargo hold.

And selecting an agreed true value as a historical record value, and setting a calibration factor value as a cargo compartment fan control function and a cargo compartment total energy consumption function. And setting the correlation function as a refrigerated container control function, a cargo compartment fan control function, a cargo compartment temperature control function and a cargo compartment energy consumption function.

Setting the individual as the brand model of the refrigerated container, and setting as the breeding group, wherein each cargo hold is used as a group, each ship is used as a group in grandma, and each airline is used as a group.

3.2、

The control of each fan or the control of the dampers on each fan is selected as a longitudinal correction value, and the system of spatial state equations shown in fig. 2 is used as a correlation function and a control function. For each ship, each cargo hold is adopted as a group to carry out transverse calibration.

3.3、

The artificial intelligence algorithm is adopted for deep learning and marking, and training is carried out by adopting a Convolutional Neural Network (CNN) algorithm, a Bayesian Bayes algorithm, an antagonistic neural network (GAN) algorithm, a firefly algorithm, an ant colony algorithm and the like. The optimized indexes comprise strategies of minimum energy consumption of the fan, minimum equipment action (such as minimum number of start-stop times and minimum number of start-stop fans), minimum disturbance to an internal temperature field, minimum disturbance to air outlet data and the like, actual control room results are stored in a database according to time, and simultaneously, relevant data such as ship route data, weather data, container loading and unloading data and berthing wharf data which can be obtained are all input into the database. These are used as the class and experience content of the deep learning of the artificial intelligence algorithm, so that the result of the artificial intelligence algorithm becomes more and more smart along with the operation of the system, and the service life of the equipment is prolonged due to the reduction of the action of the equipment.

3.4、

Each ship is networked with a cloud big data center in real time by adopting a data communication satellite, and the information of each ship and each cargo owner is protected by adopting a block chain and safe multi-party calculation.

Claims

1. The multimode personalized longitudinal and transverse calibration method is characterized by comprising the following steps:

s2000, step: obtaining an indication value of an individual and a multi-mode calibration factor value of a correlation function existing with the indication value according to a time sequence;

the association function has a constraint relation: selecting a sampling period of the calibration factor value and a sampling period of the indication value according to the relation between the variation period of the calibration factor value and the variation period of the indication value;

s3000, a step: based on a set longitudinal calibration algorithm, calibrating the indication value into a longitudinal correction value by adopting the historical and current calibration factor values and the indication value, and ensuring that the error of the longitudinal correction value and the agreed true value is converged within an allowable error; and/or the presence of a gas in the gas,

and S4000: based on a set transverse calibration algorithm, acquiring the calibration factor value and the indicating value of a group consisting of more than one individual, and calibrating the individual data of the individual according to other individual data to form a transverse correction value;

the correlation function includes establishing a constraint relationship from equation (2.7) to equation (2.12), specifying the DTM₀(t) the longitudinal correction value as the calibration factor value:

DTM(t)＝[DTM₁(t)，DTM₂(t)…DTM_m(t)]^T (2.7)

DTC(t)＝[DTC₁(t)，DTC₂(t)…DTC_p(t)]^T (2.8)

Z(t)＝[Z₁(t)，Z₂(t)…Z_q(t)]^T (2.9)

DTM(t)＝Ψ(Z(t)，DTC(t)，t) (2.11)

DTM₀(t)∈{DTM_α(t)|DTM_α(t)，1≤α≤m} (2.13)

DTM₀(t)＝DTM(t)-R(t) (2.14)

wherein:

equation (2.7) is derived from DTM₁(t) axis, DTM₂(t) Axis … DTM_m(t) a function of the state space of the indicative value component of an m-dimensional state space formed by Dimensions (DTM)_m(t) is a time function of the indicative component, m is the number of said indicative component, said indicative component comprises more than one said indicative value acquired according to said time series, and formula (2.8) is by DTC₁(t) axis, DTC₂(t) Axis … DTC_p(t) axis of the p-dimensional state space, the DTC_p(T) is a time function of calibration factor components, p is the number of the calibration factor components, the calibration factor components include more than one calibration factor value acquired according to the time series, T is the transpose of the matrix, and formula (2.9) is represented by Z₁(t) axis, Z₂(t) axis, … Z_q(t) axis, formula (2.10) is a differential equation of the state variable, formula (2.11) is an output equation of the system, and formula (2.12) is the longitudinal correction value function DTM₀Error e of (t)₀(t) is less than or equal to the allowable error e₀Formula (2.13) is that the longitudinal correction value belongs to one element of the indicating value set, and formula (2.14) is that the longitudinal correction value is the calculated value of the output equation of the system minus the value of the error correction function r (t).

2. The method according to claim 1, wherein the step S2000 specifically comprises:

step S2010: establishing a measuring module according to a formula (2.1), and establishing a calibration module according to a formula (2.2);

DTM＝{DTM_αt|DTM_αt，1≤α≤m，1≤t≤n} (2.1)

DTC＝{DTC_βt|DTC_βt，1≤β≤p，1≤t≤n} (2.2)

DTM_α(t)＝DTM_α(t+gT_M) (2.3)

DTC_β(t)＝DTC_β(t+hT_C) (2.4)

wherein:

the measurement module comprises more than one indicating value component, and the indicating value component comprises more than one indicating value acquired according to the time sequence;

the calibration module comprises more than one calibration factor component, and the calibration factor component comprises more than one calibration factor value acquired according to the time series;

DTM is the set of said indications, also the measurement module, DTM_αtThe time sequence with the indicating value component number alpha is an indicating value of t, m is the maximum number value of the indicating value component, and n is the current value of the time sequence;

DTC is the set of calibration factor values, also the calibration module, DTC_βtThe time sequence with the number of the calibration factor component is beta is a calibration factor value with t, and p is the maximum number value of the calibration factor component;

formula (2.3) is a time function of the indicating value component, alpha is the number of the indicating value component, g is any integer, T_MIs the period of the indicative value component;

equation (2.4) is a time function of the calibration factor component, β is the number of the calibration factor component, h is any integer, T_CA period of the calibration factor component;

when TC_β1＜TM_α1Taking TC_β2≤TM_α2＜TC_β1 (2.5)

When TC_β1≥TM_α1Taking TC_β2＜TM_α1And TM_α2＜TM_α1 (2.6)

Wherein:

TC_β1a period of change, TC, of said calibration factor value of said calibration factor component number β_β2A sampling period, TM, of the calibration factor value numbered beta for the calibration factor component_α1For the variation period of the indication value with the indication value component number alpha, TM_α2The sampling period of the indication value with the indication value component number alpha; and/or the presence of a gas in the gas,

and S2030 step: when TC_β1＜TM_α1Simultaneous TM_α2＜TC_β1When it is, the TM_α2Decomposition into short sampling periods TM_α2SAnd a long sampling period TM_α2LOf (1) with TM_α2LObtaining the indication value component DTM with slow response speed but higher measurement accuracy_αLBy TM_α2SObtaining the indication value component DTM with high corresponding speed but low measurement precision_αSUsing DTM_αSWith DTC_βThe correlation function is established, and the DTM is calibrated_αLA step (2);

a change in one or more of the said indicators results in a unidirectional change in one or more other of the said calibration factor values, or a change in one or more of the said calibration factor values results in a unidirectional change in one or more other of the said indicators, or a change in either one or both of them results in a bidirectional change in the other, the said change comprising a functional relationship as determined by equation (2.15) or equation (2.16) or equation (2.17);

wherein:

f_2.15is a bidirectional mapping function, operation sign

Is a two-way mapping operator, i.e., a change in the value of the calibration factor will affect the indication value, and a change in the indication value will also affect the value of the calibration factor;

f_2.16is a left-hand mapping function, operator

A left-hand mapping operator, i.e. a change in the value of the calibration factor will affect the value of the calibration factor, whereas a change in the value of the calibration factor will not affect the value of the indication;

f_2.17is a right-mapping function, operator

A right-mapping operator, i.e., a change in the value of the calibration factor will affect the indication, while a change in the indication will not affect the value of the calibration factor; and/or the presence of a gas in the gas,

step S2060: the correlation function further includes a relationship of the indication function being non-periodic and the calibration factor value function being periodic, including the T in the formula (2.3)_MTaking infinity and said T in said equation (2.4)_CTaking the period of the calibration factor value function; and/or the presence of a gas in the gas,

step S2070: the correlation function also comprises the relationship between the indicating function with linear time variability and the calibration factor value function with linear time invariance, including the relationship between the output and the input of the function in the formula (2.3) is linear, the relationship between the response and the input of the function is related to time, the relationship between the output and the input of the function in the formula (2.4) is linear, and the relationship between the response and the input of the function is not related to time; and/or the presence of a gas in the gas,

step S2080: said correlation function further comprises a relationship of said indication function of causality-stability and said calibration factor value function of causality-stability, including that the output of the functions in said equations (2.3) and (2.4) is independent of the input after the output instant; and/or the presence of a gas in the gas,

s2090: the correlation function further includes a mapping between the set of calibration factor values and the set of indicative values that is difficult to determine, but obtained from analysis by big data and artificial intelligence algorithms; and/or the presence of a gas in the gas,

and S2100 step: the correlation function further comprises the steps of establishing a Riemann manifold coordinate system by adopting the calibration factor value, the indication value and the time sequence, and establishing a relation function among the calibration factor value, the indication value and the time sequence by adopting Riemann geometry so as to avoid the problem that the mapping relation among the calibration factor value set and the indication value set generates non-one-to-one mapping in the step S2080;

step S2110: and storing the historical time point and the current time point of all the individuals in the group, the calibration module, the calibration factor value, the measurement module and the indication value into a database, wherein the historical time point is a part of or all of the time points before the current time point.

3. The method according to claim 1, wherein the step S3000 specifically comprises:

and S3010: a longitudinal calibration algorithm step of calculating the current indication value as the longitudinal correction value by using a calibration function according to the current calibration factor value and the indication value, so that an error between the longitudinal correction value and the agreed true value is smaller than an allowable error; and/or the presence of a gas in the gas,

and S3020: a longitudinal calibration algorithm step of calculating the current indication value as the longitudinal correction value according to more than one historical calibration factor value and indication value, according to the front-back order of the time series and the waveform setting weight, and adopting the calibration function, so that the error between the longitudinal correction value and the appointed true value is smaller than the allowable error;

step S3031: a step of synchronously calculating the indication value according to the same time period by the individual on the time sequence of the same time period according to the fluctuation of the calibration factor value by adopting the calibration function, thereby obtaining a longitudinal synchronization correction value of which the error between the individual and the agreed true value is smaller than an allowable error;

step S3032: the individual uses the longitudinal synchronization correction value as an independent variable in the time sequence of different time periods, enlarges the time period to include a day, a week, a month, a year or a user-specified period, adopts the calibration function to calculate and obtain a longitudinal history correction value, and calculates an error between the longitudinal history correction value and the longitudinal synchronization correction value;

step S3033: reducing or eliminating the error between the longitudinal historical corrected value and the longitudinal synchronous corrected value by adopting a statistical algorithm and/or an artificial intelligence classification algorithm;

the time period comprises more than one continuous time point or all the time points in one period of the calibration factor value in the calibration module; and/or the presence of a gas in the gas,

step S3040: the longitudinal calibration algorithm further comprises: establishing a calibration parameter set between the calibration factor value and the indicating value by adopting a statistical algorithm and/or an artificial intelligence classification algorithm and the historical calibration factor value and the indicating value, and calibrating the current indicating value into the longitudinal correction value according to the calibration parameter set and the current calibration factor value and the indicating value;

s3050: the longitudinal calibration algorithm also comprises a step of obtaining an agreed true value, wherein the individual is measured by monitoring equipment with precision, accuracy or precision grade higher by more than one grade and measurement error smaller than the allowable error to obtain a measurement value as the agreed true value, and the ratio of the measurement error smaller than the ratio of the standard monitoring equipment to the monitoring error of the method is smaller than 0.5 or a value specified by a user; and/or, the step of obtaining the engagement truth value specifically comprises the step of obtaining through calculation by adopting historical data of other individuals.

4. The method according to one of claims 2 or 3, wherein the step S4000 specifically includes:

s4010 step: establishing a group consisting of more than one individual with the same measurement attribute, acquiring the calibration factor value and the indication value of the individual in the group, and respectively incorporating the calibration factor value and the indication value into a standard module and a measurement module, wherein the group comprises other individuals and self individuals;

the first step is as follows: executing the longitudinal calibration algorithm for all the individuals in the population, and acquiring the longitudinal correction values and the errors of all the individuals, wherein the longitudinal correction values comprise the longitudinal synchronous correction values and the longitudinal historical correction values;

the second step is that: calculating the calibration parameter set by adopting the calibration functions including a weighting algorithm, a filtering algorithm, a partial differential algorithm, a fuzzy algorithm, a differential geometry algorithm and an artificial intelligence deep learning algorithm according to the longitudinal correction values and the errors of the other individuals so as to calibrate the longitudinal correction values of the individuals, so that the errors of the longitudinal correction values are smaller than the allowable errors, and acquiring the transverse correction values of the individuals; and/or the presence of a gas in the gas,

calibrating the indicating value of the self-individual by adopting the calibration function according to the calibration factor value and the indicating value of all the other individuals in the group and the calibration factor value and the indicating value of the self-individual, and acquiring the transverse correction value and the error of the self-individual, wherein the error of the self-individual is smaller than the allowable error; and/or the presence of a gas in the gas,

circularly executing the step S4020 or the step S4030 according to a historical time period to acquire the lateral correction value and the error of the individual, and selecting a time period which is less than the expanded time period and greater than the time period according to the measurement attribute; and/or the presence of a gas in the gas,

s4050: the transverse calibration algorithm further comprises a step of classified implementation, specifically:

classifying the individuals in the group according to the calibration parameter set, when the individuals perform transverse calibration, selecting the other individuals of the same class or similar classes to perform calibration calculation, and acquiring the transverse correction value and the error of which the error of the individuals is smaller than the allowable error; and/or the presence of a gas in the gas,

s4061 step: calculating individual errors of the individuals in the population, and establishing individual error categories and error correction parameters of which the errors are smaller than the allowable errors according to the individual errors by adopting the error sizes and the error severity; and/or the presence of a gas in the gas,

s4062: circularly executing the step S4061 according to the historical time period to obtain the current individual error category and the current error correction parameter of the individual;

s4063: performing the step S4061, or performing the steps S4061 and S4062 for the newly added individual in the population;

5. The method according to claim 4, characterized in that it further comprises a step of calibrating the function, in particular:

acquiring one appointed true value as an average value of the time period and acquiring more than one indication value, calculating the error according to a formula (5.1) or a formula (5.2), calculating the standard deviation according to a formula (5.3) or a formula (5.4), calculating the calibrated error according to a formula (5.5), checking that the calibrated error is smaller than the allowable error according to a formula (5.6), and constraining the corrected value of the calibrated indication value according to a formula (5.7) as a constraint condition;

wherein:

and

is a correction value after the calibration is performed,

is the corrected value error of the calibrated indication, e₀Is the allowable error;

if m is less than 30, then pair delta_αtAnd delta_βtPerforming t-test in statistics to calibrate abnormal values;

if m is greater than 30, then pair δ_αtAnd delta_βtPerforming a statistical Z-test to calibrate outliers;

deleting abnormal values or carrying out interpolation calculation to supplement the indicating values, and taking all the indicating values and the indicating values subjected to interpolation calculation as the longitudinal correction values;

s5021: according to the step of obtaining the agreed truth value, selecting a plurality of time periods for obtaining the agreed truth value aiming at the individual, taking R (t) as 0, and calculating the longitudinal correction value by adopting a function fitting method;

s5022: executing the step S5021 for a plurality of the individuals in the population, and obtaining a corresponding table calculated by the formula (2.14);

s5023: calculating R (t) by adopting a statistical method or a deep learning or neural network method aiming at the corresponding table;

selecting the indicating value component and the calibration factor component which are mutually associated according to the association function, and establishing a corresponding mapping relation function of set elements according to the calibration factor value set of the calibration factor component of the time sequence and the indicating value set of the indicating value component in the same time period;

partial differential operation is carried out on the time sequence according to the corresponding mapping relation function, a component which has a linear relation with the fluctuation is selected from the calibration factor value set according to the fluctuation of elements in the indicating value component set, the component is taken as a main indicating value component, and the rest is taken as an auxiliary indicating value component;

calibrating the abnormal value of the principal component according to the t-test or the Z-test;

deleting abnormal values or carrying out interpolation calculation to supplement the indicating values, and taking all the indicating values and the indicating values subjected to interpolation calculation as the longitudinal correction values; and/or the presence of a gas in the gas,

and S5040, step: the calibration function further includes a state equation method, specifically, solving the solutions in the output equations of the system of the formula (2.11) and the formula (2.12) in the step S2030, and decomposing all the output components of the system to obtain one of the output components DTM₀(t) as said longitudinal correction value, the other output components being considered as interference values; and/or the presence of a gas in the gas,

aiming at relational data records in the database, taking the data records as neurons, and establishing a connection function between the neurons according to a calculation result comprising a mathematical model to form more than one layer of neural network;

dividing and establishing exciting type, inhibiting type, explosion type and plateau type connection subfunctions according to the effect of the indicating value and the calibration factor value on the error in the connection functions, wherein the connection subfunctions comprise constant type weight coefficients and functional type weight coefficients;

optimizing the connector function by adopting a deep learning algorithm comprising supervised learning, unsupervised learning and reinforcement learning algorithms; and/or the presence of a gas in the gas,

classifying and screening the error and the output component by adopting a support vector machine algorithm, and screening the longitudinal correction value and the interference value; and/or the presence of a gas in the gas,

performing convolution, activation, pooling, full connection and training of the connection subfunction by adopting a convolutional neural network algorithm to screen out the longitudinal correction value, the interference value and the corresponding indication value; and/or the presence of a gas in the gas,

establishing an in-layer association function by adopting a recurrent neural network algorithm, and training the connection subfunction to screen out the longitudinal correction value and the interference value; and/or the presence of a gas in the gas,

establishing an interlayer association function for the indicating values, the calibration factor values and the errors among the layers of each neural network by adopting a deep neural network algorithm, and training the connection subfunction to screen out the longitudinal correction value and the interference value; and/or the presence of a gas in the gas,

training the connector functions by adopting a feedforward neural network algorithm under the condition that each neuron is only connected with the neuron of the previous layer so as to screen out the longitudinal correction value and the interference value; and/or the presence of a gas in the gas,

training the connector functions by adopting a feedback neural network algorithm under the condition that each neuron is only connected with a neuron of the next layer so as to screen out the longitudinal correction value and the interference value; and/or the presence of a gas in the gas,

s5060, the calibration function further includes establishing a state observer by using a kalman filtering method, calculating an output of the state observer by using the indication variable and the calibration factor value variable as input variables, comparing the output with the longitudinal calibration value, and analyzing a comparison result by deep learning to correct the calibration parameter set so that the error converges with the allowable error;

and S5070: the calibration function further comprises the step of calculating the value obtained by subtracting the auxiliary indication value component from the indication value in the subsequent time period as the longitudinal correction value if the auxiliary indication value component is found to be consistent in other time periods through statistical calculation on the auxiliary indication value component; and/or the presence of a gas in the gas,

and S5080: and the calibration function also comprises simplified calculation by adopting the optimized calibration parameter set to obtain the longitudinal correction value.

6. The method according to claim 5, comprising the step of calibrating the set of parameters, in particular:

s6010: sorting the calculation parameters in the calibration function according to the time sequence, establishing a fixed data format, and recording the time sequence and the algorithm number to form a calibration parameter set record;

step S6020: the calibration parameter set comprises all the calibration parameter set records accumulated according to the time sequence;

and S6030: the calibration parameter set also comprises data of an algorithm body, constants, coefficients, weight coefficients, process parameters and time sequences based on an artificial intelligence algorithm;

and S6040: according to the calibration parameter set, adopting a statistical method or an artificial intelligence method, and taking the minimum error, the minimum standard deviation and the minimum calculated amount as optimization indexes to perform iterative optimization upgrading to form an optimized calibration parameter set with a time sequence label;

and S6050: a step of storing the calibration parameter set in the database.

7. The method according to claim 6, comprising a step S7000, specifically comprising a single step or a combination of multiple steps of a S7100 cloud big data mode, a S7200 local area network mode, a S7300 single point mode:

s7100, cloud big data mode:

s7110, establishing a cloud center based on an internet mode, transmitting all data including all groups and individuals, intermediate calculation results and the database acquired by the method through a wide area network in a cloud terminal form, storing all the data on more than one cloud server based on the internet, taking the data as more than one cloud center, and managing, calculating and supporting by adopting the cloud calculation mode; and/or the presence of a gas in the gas,

s7120, establishing more than one cloud center by adopting a block chain mode to store, manage and support the database and the steps, wherein the individuals adopt anonymous records, information in the database adopts a chain structure with a timestamp, a user accesses the database and adopts encryption and decryption communication, the information supports tamper resistance, and a repudiation-proof, multi-center and centerless mode is supported; and/or the presence of a gas in the gas,

s7130, establishing, managing and supporting more than one mechanism in a safe multi-party computing mode, performing appointed computing according to the database content of each mechanism on the premise of not exchanging the database core data of the cloud center of each mechanism, and sharing the obtained computing result by the participating mechanisms; the organization comprises more than one cloud center and manages more than one individual; the secure multi-party computation includes: public key mechanism, hybrid circuit, careless transmission, secret sharing, privacy protection set intersection protocol, homomorphic encryption, zero knowledge proof and method without trusted center to enhance information security and protect object privacy; and/or the presence of a gas in the gas,

s7140, establishing and training a model for the privacy protection of the de-emphasized object by adopting a centralized learning mode, wherein the database is stored in a cloud center; and/or the presence of a gas in the gas,

s7150, establishing and training a model training for the object privacy protection needing to be emphasized by adopting a federal learning mode, wherein the model training is carried out among more than one stored cloud center, and data of the cloud centers are not exchanged; and/or the presence of a gas in the gas,

step S7200, local area network mode:

establishing a server based on a local area network for storing and managing a support center, transmitting all the data including all the groups and the individuals, the intermediate calculation result and the database monitored by the method in a network terminal form through the local area network, and storing all the data on the server based on the local area network so as to manage, calculate and support; and/or the presence of a gas in the gas,

step S7300, single point mode:

the single point is a step of monitoring detection, storage, management, calculation and support of one object, all information, intermediate calculation results and the information base of the object monitored by the invention are stored in the storage of the single point, and all steps are executed.