Method for determining the fluidization state of a fluidized bed

Abstract

This invention discloses a method for determining the fluidization state of a fluidized bed. The steps include: S1, spectral acquisition, using a diffuse reflectance near-infrared spectral acquisition system to acquire the diffuse reflectance near-infrared spectrum of the material inside the monitoring window of the fluidized bed; S2, moving window update, the moving window includes n continuously acquired spectra, with a moving step size of m spectra; S3, spectral classification, classifying the diffuse reflectance spectra within the moving window into three categories: high baseline, mid baseline, and low baseline spectra; S4, fluidization state determination, comparing the proportions of high baseline and low baseline spectra within x consecutive moving windows with preset thresholds to determine the fluidization state; S5, repeating steps S1-4 to achieve online fluidization state determination. This method is convenient to apply, has a fast response speed, strong anti-interference ability, and good stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fluidized state judgment in fluidized bed production process, and particularly relates to a method for judging fluidized state of fluidized bed based on near-infrared spectrum. BACKGROUND

[0002] Fluidized bed spray granulation is to make material keep in a state of suspension and fluidization in fluidized bed by using air flow, and spray the binder liquid which has been atomized, so as to make the material powder agglomerate into granules. It can complete the traditional mixing, drying, granulation and coating in the same closed instrument at one time, and realize one-step granulation. Fluidized bed one-step granulation has many advantages, and compared with other methods, it has the characteristics of simple process, short operation time and low labor intensity, and has been widely used in pharmaceutical, chemical, agricultural and other industries.

[0003] In a device, the granular material is placed on the distribution plate, and when the gas is introduced into the bed from the lower part of the device, with the increase of the air flow rate to a certain degree, the solid particles in the bed produce a boiling state, which is called fluidized state. The fluidized bed granulation process is complex, and the material can only keep the process normal under the condition of maintaining normal fluidization state. Excessive fluidization state and loss of fluidization state are abnormal process conditions in the process, and the process needs to be adjusted to reduce or avoid unnecessary economic losses.

[0004] In the granulation process, the monitoring of the fluidized state is of great significance to maintain the stability of the granulation process. The current monitoring methods of the fluidized state include differential pressure method, image method and the like, however, these methods have their own shortcomings, such as the differential pressure method is easily affected by the filter bag blockage, and the image acquisition and analysis are also more difficult. In recent years, with the development of process analysis technology.

[0005] The near-infrared spectral region is discovered by Herschel in 1800 when he measured the energy in the infrared part of the visible region of the solar spectrum. In order to commemorate the historical discovery of Herschel, the waveband between 780-1100nm in the near-infrared spectral region is called Herschel spectral region.

[0006] Infrared spectroscopy as an effective analytical method has been recognized in the 1930s, when infrared instruments were mainly used for the study of molecular structure theory. The spectral absorption band in the near-infrared region is the superposition of the frequency doubling, frequency mixing and frequency difference absorption bands of the high-energy chemical bonds (mainly CH, OH, NH) in organic matter in the mid-infrared spectral region. Due to the serious overlapping and discontinuity of the near-infrared spectrum, it is difficult to directly extract the information related to the content of the substance in the near-infrared spectrum and give reasonable spectral analysis. However, the absorption band of organic matter in the mid-infrared spectral region is more, the spectral band is narrow, the absorption intensity is large, and there is significant characteristic absorption. Therefore, the near-infrared spectral region has been ignored and forgotten for a long time. SUMMARY

[0007] In order to overcome the above technical problems, the present application provides a drop pill drop process fault detection method. The method is convenient to use, fast in response, strong in anti-interference ability and good in stability.

[0008] In order to achieve the above purpose, the technical scheme provided by the present application is as follows:

[0009] A method for judging the fluidized state of a fluidized bed, comprising the steps of:

[0010] S1, spectrum collection, using a diffuse reflection near-infrared spectrum collection system to collect the diffuse reflection near-infrared spectrum of the material inside the monitoring window of the fluidized bed;

[0011] S2, moving window update, the moving window includes n continuous spectra, and the moving step is m spectra;

[0012] S3, spectrum classification, the diffuse reflection spectrum in the moving window is divided into three categories: high baseline, medium baseline and low baseline spectrum;

[0013] S4, fluid state judgment, the proportion of high baseline spectrum and the proportion of low baseline spectrum in the continuous x moving windows are compared with the preset threshold value to judge the fluid state;

[0014] S5, repeating steps S1-4 to realize online judgment of the fluid state.

[0015] Further, the diffuse reflection near-infrared spectrum collection system in step S1 comprises a near-infrared sensor, a near-infrared detector and a computer terminal connected in sequence.

[0016] Further, the near-infrared sensor is fixed outside the monitoring window of the fluidized bed, and the fixed height needs to ensure that the irradiation area is in the dense phase zone of the fluidized bed. The two parallel near-infrared light beams penetrate the glass of the monitoring window, and the two light beams are focused on the second end surface of the monitoring window.

[0017] Further, n in the step S2 is 5-30, and m is 5-30.

[0018] n in the step S2 is 10-20, preferably 15; and m is 1-20, preferably 15.

[0019] Preferably, in the method for judging the fluidization state of the fluidized bed, the spectral classification in the step S3 comprises the following steps.

[0020] S31, spectral clustering, summing up the absorbance at each wavelength of the spectrum to obtain the total absorbance to represent each spectrum, and using the k-means clustering algorithm to divide the spectrum into three clusters according to the total absorbance;

[0021] S32, cluster labeling, using the spectral similarity between clusters and the standard deviation of the total absorbance of the spectrum within the cluster as the judgment basis, and attributing the spectrum within each cluster to low baseline, medium baseline or high baseline spectrum; the comparison object of the spectral similarity between clusters is the average spectrum of the spectrum within each cluster, and the spectral similarity is calculated by using the angle cosine algorithm.

[0022] Preferably, in the step S31, the k-means clustering algorithm is used to divide the spectrum into three clusters according to the total absorbance; and in the step S32, the angle cosine algorithm is used to calculate the spectral similarity.

[0023] Further preferably, x in the step S4 is 5, and the threshold value comprises a flow loss threshold T1 and an excessive fluidization threshold T2.

[0024] Further preferably, the proportion of the low baseline spectrum being higher than the threshold T1 represents that the fluidized bed is in a flow loss state, the proportion of the high baseline spectrum being higher than the threshold T2 represents that the fluidized bed is in an excessive fluidization state, and the proportions of the low baseline spectrum and the high baseline spectrum are both less than the threshold value, which represents that the fluidized bed is in a normal fluidization state.

[0025] Further, the threshold T1 is 7.3-25.8%, preferably 16.5%; and the threshold T2 is 2.5-5.6%, preferably 4.2%.

[0026] Compared with the prior art, the present application has the following beneficial effects:

[0027] 1. In the present application, a non-invasive probe is used for monitoring, a double light source light beam is focused on the inside of a monitoring window, and the risk of contaminating the material is avoided while obtaining the fluidization state information of the fluidized bed.

[0028] 2. In the present application, an unsupervised clustering algorithm and a labeling judgment basis which is not sensitive to the spectral peak type are used to complete the spectral baseline level detection, and the fluidization state description without the guidance of prior knowledge related to the chemical information of the material is realized.

[0029] The present application uses near-infrared (NIR) spectrum innovatively, which can reflect the physicochemical information of liquid or solid material in the production process in real time, and through analyzing the obtained information, the process key parameters or process failure can be measured or judged, the real-time monitoring of the process is realized, and the final quality of the prepared product is ensured.

[0030] The physicochemical properties of the measured object are contained in the NIR spectrum information. For monitoring the fluidization state of the fluidized bed using NIR, the information of the particle flow characteristics can be extracted from the NIR spectrum of the material in the fluidized bed reactor through data processing method. The baseline level of the diffuse reflectance NIR spectrum is affected by the physical state of the material at the monitoring window of the fluidized bed. Light scattering and transmission effect is the main reason for the upward drift of the spectral baseline. In the fluidized bed, the increase of the bed voidage leads to the enhancement of the light scattering effect, the large bubbles passing through the glass window lead to the light transmission and the proportion of the bubbles causes the light transmission. Therefore, the diffuse reflectance NIR spectrum can reflect the movement state of the material at the monitoring window of the fluidized bed.

[0031] The movement state of the material in the fluidized bed directly reflects the fluidization state of the fluidized bed. When the fluidized bed is in the defluidization state, the material is in the accumulation state, the bed voidage is low, and the spectral baseline is at a low level. When the fluidized bed is in the normal fluidization state, the material is in a relatively stable movement mode, the bed voidage fluctuates within a certain range, and the spectral baseline is at a medium level. When the fluidized bed is in the overfluidization state, the emulsified phase gas content is high, the material is in a complex movement state, and large bubbles are prone to appear at the monitoring window of the fluidized bed, and high baseline spectrum occurs frequently. Therefore, according to the baseline drift degree of the diffuse reflectance NIR spectrum within a period of time, such as the proportion of low baseline, medium baseline and high baseline spectrum, the fluidization state of the fluidized bed can be detected.

[0032] The present application uses diffuse reflectance NIR spectrum technology to detect the fluidization state of the fluidized bed. The movement state of the material in the irradiation area is reflected according to the baseline drift degree of the online spectrum, and the fluidization state of the fluidized bed is characterized. The detection range of the NIR spectrum technology is expanded, and the detection method of the fluidization state of the fluidized bed is further improved. BRIEF DESCRIPTION OF DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed in the following embodiment or prior art description will be briefly introduced. Obviously, the drawings in the following description are only exemplary, and other drawings can be derived from the provided drawings without creative labor for those skilled in the art.

[0034] Figure 1 The workflow diagram of the method for judging the fluidization state of the fluidized bed of the present application;

[0035] Figure 2A spectral classification flow chart for the method for judging the fluidization state of a fluidized bed of the present application;

[0036] Figure 3 An embodiment spectral classification result for the method for judging the fluidization state of a fluidized bed of the present application;

[0037] Figure 4 An embodiment fluidization state judgment result for the method for judging the fluidization state of a fluidized bed of the present application. DETAILED DESCRIPTION

[0038] The following experimental examples and embodiments are used to further illustrate but not limit the present application.

[0039] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0040] Embodiment 1: The detection object is a certain Chinese patent medicine fluidized bed granulation process, the work flow chart for judging the fluidization state of the fluidized bed is as shown in Figure 1 , and the specific steps are as follows:

[0041] S1, spectrum collection, the near-infrared sensor is fixed outside the monitoring window of the fluidized bed, the irradiation area is located in the dense phase zone of the fluidized bed, and the double light sources are focused inside the monitoring window of the fluidized bed to monitor the material fluidization state in the material fluidization process. The parallel near-infrared light beams penetrate through the glass of the monitoring window, and the two light beams are focused on the second end face of the monitoring window. The diffuse reflection near-infrared spectrum is different under different material fluidization states. The diffuse reflection near-infrared spectrum detection range is 900-1700 nm, and one spectrum is collected every 2 s.

[0042] S2, moving window update, the moving window includes 15 continuously collected spectra, and the step length is 15 spectra.

[0043] Since the moving window duration is short, the chemical properties of the particles change little, the spectrum absorbance is mainly affected by light scattering and transmission, and the baseline drift is the main source of variation of the spectrum.

[0044] S3, spectrum classification, the diffuse reflection spectrum in the moving window is classified into three categories of high baseline, medium baseline and low baseline spectrum, and the flow chart of the spectrum classification is as shown in Figure 2 .

[0045] The spectrum classification includes the following steps:

[0046] S31. Spectral clustering: The total absorbance is obtained by summing the absorbance at each wavelength of the spectrum to represent each spectrum. The k-means clustering algorithm is used to divide the spectrum into 3 clusters based on the total absorbance.

[0047] Since baseline drift is positively correlated with total absorbance, to ensure optimal clustering results, the maximum, median, and minimum values ​​of total absorbance were set as initial cluster centers, dividing the spectra into three clusters: C1, C2, and C3. The iteration count was set to one. The actual assignment of spectra within each cluster can be categorized into seven classes, as shown in Table 1.

[0048] Table 1. Spectral Attribution Types within Each Cluster

[0049]

[0050] Note: L, M, and H represent low baseline, mid baseline, and high baseline spectra, respectively.

[0051] S32. Cluster labeling: Using inter-cluster spectral similarity and the standard deviation of total absorbance within a cluster as the criteria, the actual classification of the spectra within each cluster is determined, assigning them to low-baseline, mid-baseline, or high-baseline spectra. The inter-cluster spectral similarity is compared using the average spectrum within each cluster, and the cosine similarity algorithm is used to calculate the spectral similarity.

[0052] Because high-baseline spectra are dominated by light transmission effects and exhibit peak distortion, they can be distinguished from low- and mid-baseline spectra by comparing inter-cluster spectral similarity. Low-baseline spectra originate from stationary particles, resulting in the lowest fluctuation in total absorbance. Differences in mid-baseline spectra arise from random fluctuations in bed porosity under normal fluidization conditions, with total absorbance fluctuating within a small range. High-baseline spectra, however, originate from bubbles occupying the irradiated area; this phenomenon has poor repeatability, leading to a larger fluctuation in total absorbance. Low-baseline spectra can be distinguished from mid- and high-baseline spectra by comparing the standard deviation of total absorbance. Specific discrimination criteria are as follows:

[0053] Condition ①: The standard deviation of the total absorbance of the spectrum within cluster C2 is less than the threshold.

[0054] Condition ②: The standard deviation of the total absorbance of the spectrum within cluster C3 is less than the threshold.

[0055] Condition ③: The spectral similarity between clusters C1 and C2 is less than the threshold.

[0056] Condition ④: The spectral similarity between clusters C2 and C3 is less than the threshold.

[0057] Condition ⑤: The standard deviation of the total spectral absorbance within cluster C1 is less than the threshold.

[0058] The standard deviation of total absorbance within a cluster and the threshold for spectral similarity between clusters were 0.2 and 0.9950, respectively.

[0059] S4. Flow regime determination: The proportions of high baseline and low baseline spectra within five consecutive moving windows are compared with preset thresholds, and the fluidization regime is determined every 2.5 minutes. The loss-of-flow threshold T1 and the over-fluidization threshold T2 are 16.5% and 4.2%, respectively. A low baseline spectra proportion higher than threshold T1 indicates that the fluidized bed is in a loss-of-flow state, a high baseline spectra proportion higher than threshold T2 indicates that the fluidized bed is in an over-fluidized state, and both the low and high baseline spectra proportions are lower than the thresholds, indicating that the fluidized bed is in a normal fluidization state.

[0060] S5. Repeat steps S1-4 to achieve online flow regime determination until the fluidized bed granulation process ends.

[0061] In this embodiment, a total of 1905 diffuse reflectance near-infrared spectra were collected, with total absorbance representing the spectrum. The spectral classification results are as follows: Figure 3 As shown, although the chemical properties of the material change during the granulation process, low-baseline, medium-baseline, and high-baseline spectra can be effectively identified at each stage of granulation, verifying that the spectral classification method proposed in this invention does not require prior knowledge related to the chemical properties of the material.

[0062] Because the material adheres to the fluidized bed monitoring window during the early stages of granulation, the spectral differences are small, making it impossible to determine the internal fluidization state of the fluidized bed. This state lasts for approximately 3 minutes. Therefore, fluidization state determination begins at the 5th minute, and the fluidization state determination results are as follows: Figure 4 As shown, during the early and late stages of the spraying phase and the drying phase, the proportion of high baseline spectra within five consecutive moving windows was higher than the over-fluidization threshold T2, indicating that large bubbles frequently occupied the irradiated area, suggesting that the fluidized bed was in an over-fluidized state. In the middle stage of the spraying phase, the proportions of both high and low baseline spectra were lower than the threshold, indicating that the bed porosity in the irradiated area fluctuated randomly within a certain range, suggesting that the fluidized bed was in a normal fluidized state, which is consistent with the observation results.

[0063] The test results of Example 1 show that the fluidized bed fluidization state determination method proposed in this invention has good determination performance, expands the application range of diffuse reflectance near-infrared spectroscopy, and provides a new means for the detection of fluidized bed fluidization state.

[0064] Example 2:

[0065] A method for determining the fluidization state of a fluidized bed, comprising the following steps:

[0066] S1. Spectral acquisition: The diffuse reflectance near-infrared spectral acquisition system is used to acquire the diffuse reflectance near-infrared spectrum of the material inside the fluidized bed monitoring window;

[0067] S2, Moving window update, the moving window includes n continuously acquired spectra, and the moving step size is m spectra;

[0068] S3. Spectral classification: Diffused reflectance spectra within the moving window are classified into three categories: high baseline, medium baseline, and low baseline spectra.

[0069] S4. Flow regime determination: The proportions of high baseline spectrum and low baseline spectrum within x consecutive moving windows are compared with preset thresholds to determine the fluidization regime.

[0070] S5. Repeat steps S1-4 to achieve online flow regime determination.

[0071] Example 3:

[0072] A method for determining the fluidization state of a fluidized bed, comprising the following steps:

[0073] S1. Spectral acquisition: The diffuse reflectance near-infrared spectral acquisition system is used to acquire the diffuse reflectance near-infrared spectrum of the material inside the fluidized bed monitoring window;

[0074] S2, Moving window update, the moving window includes n continuously acquired spectra, and the moving step size is m spectra;

[0075] S3. Spectral classification: Diffused reflectance spectra within the moving window are classified into three categories: high baseline, medium baseline, and low baseline spectra.

[0076] S4. Flow regime determination: The proportions of high baseline spectrum and low baseline spectrum within x consecutive moving windows are compared with preset thresholds to determine the fluidization regime.

[0077] S5. Repeat steps S1-4 to achieve online flow regime determination;

[0078] Furthermore, the diffuse reflectance near-infrared spectral acquisition system in step S1 includes a near-infrared sensor (disclosed in CN212658619U), a near-infrared detector, and a computer terminal connected in sequence. The near-infrared sensor is fixed outside the fluidized bed monitoring window, and the fixed height must ensure that the irradiation area is in the dense phase region of the fluidized bed. Parallel near-infrared beams pass through the glass of the monitoring window, and the two beams are focused on the second end face of the monitoring window.

[0079] Furthermore, in step S2, n is 18 and m is 18.

[0080] Example 4:

[0081] A method for determining the fluidization state of a fluidized bed, comprising the following steps:

[0082] S1. Spectral acquisition: The diffuse reflectance near-infrared spectral acquisition system is used to acquire the diffuse reflectance near-infrared spectrum of the material inside the fluidized bed monitoring window;

[0083] S2, Moving window update, the moving window includes n continuously acquired spectra, and the moving step size is m spectra;

[0084] S3. Spectral classification: Diffused reflectance spectra within the moving window are classified into three categories: high baseline, medium baseline, and low baseline spectra.

[0085] S4. Flow regime determination: The proportions of high baseline spectrum and low baseline spectrum within x consecutive moving windows are compared with preset thresholds to determine the fluidization regime.

[0086] S5. Repeat steps S1-4 to achieve online flow regime determination.

[0087] Furthermore, the diffuse reflectance near-infrared spectral acquisition system in step S1 includes a near-infrared sensor, a near-infrared detector, and a computer terminal connected in sequence.

[0088] Furthermore, the near-infrared sensor is fixed to the outside of the fluidized bed monitoring window. The fixed height must ensure that the irradiation area is in the dense phase region of the fluidized bed, so that parallel near-infrared beams pass through the glass of the monitoring window and the two beams are focused on the second end face of the monitoring window.

[0089] Furthermore, in step S2, n is 15 and m is 15.

[0090] The spectral classification in step S3 includes the following steps:

[0091] S31. Spectral clustering: The total absorbance is obtained by summing the absorbance at each wavelength of the spectrum to represent each spectrum. The k-means clustering algorithm is used to divide the spectrum into three clusters based on the total absorbance.

[0092] S32. Cluster labeling: The inter-cluster spectral similarity and the standard deviation of the total absorbance of the intra-cluster spectrum are used as the criteria to classify the spectra within each cluster into low-baseline, medium-baseline, or high-baseline spectra. The comparison object for inter-cluster spectral similarity is the average spectrum of the spectra within each cluster, and the spectral similarity is calculated using the cosine similarity algorithm.

[0093] In step S31, the k-means clustering algorithm is used to divide the spectrum into three clusters based on the total absorbance; in step S32, the cosine similarity algorithm is used to calculate the spectral similarity.

[0094] In step S4, x is 5, and the thresholds include the loss-of-flow threshold T1 and the excessive fluidization threshold T2. A low baseline spectrum percentage higher than threshold T1 indicates that the fluidized bed is in a loss-of-flow state, a high baseline spectrum percentage higher than threshold T2 indicates that the fluidized bed is in an excessive fluidization state, and both the low baseline and high baseline spectrum percentages are lower than the thresholds, indicating that the fluidized bed is in a normal fluidization state. The thresholds T1 and T2 are 16.5% and 4.2%, respectively.

[0095] Example 5:

[0096] A method for determining the fluidization state of a fluidized bed, comprising the following steps:

[0097] S1. Spectral acquisition: The diffuse reflectance near-infrared spectral acquisition system is used to acquire the diffuse reflectance near-infrared spectrum of the material inside the fluidized bed monitoring window;

[0098] S2, Moving window update, the moving window includes n continuously acquired spectra, and the moving step size is m spectra;

[0099] S3. Spectral classification: Diffused reflectance spectra within the moving window are classified into three categories: high baseline, medium baseline, and low baseline spectra.

[0100] S4. Flow regime determination: The proportions of high baseline spectrum and low baseline spectrum within x consecutive moving windows are compared with preset thresholds to determine the fluidization regime.

[0101] S5. Repeat steps S1-4 to achieve online flow regime determination.

[0102] Furthermore, the diffuse reflectance near-infrared spectral acquisition system in step S1 includes a near-infrared sensor, a near-infrared detector, and a computer terminal connected in sequence; the near-infrared sensor is fixed outside the fluidized bed monitoring window, and the fixed height must ensure that the irradiation area is in the dense phase region of the fluidized bed, so that parallel near-infrared beams pass through the glass of the monitoring window, and the two beams of light are focused on the second end face of the monitoring window.

[0103] Furthermore, in step S2, n is 12 and m is 12.

[0104] The method for determining the fluidization state of a fluidized bed, wherein the spectral classification in step S3 includes the following steps:

[0105] S31. Spectral clustering: The total absorbance is obtained by summing the absorbance at each wavelength of the spectrum to represent each spectrum. The k-means clustering algorithm is used to divide the spectrum into three clusters based on the total absorbance.

[0106] S32. Cluster labeling: The inter-cluster spectral similarity and the standard deviation of the total absorbance of the intra-cluster spectrum are used as the criteria to classify the spectra within each cluster into low-baseline, medium-baseline, or high-baseline spectra. The comparison object for inter-cluster spectral similarity is the average spectrum of the spectra within each cluster, and the spectral similarity is calculated using the cosine similarity algorithm.

[0107] In step S31, the k-means clustering algorithm is used to divide the spectrum into three clusters based on the total absorbance; in step S32, the cosine similarity algorithm is used to calculate the spectral similarity.

[0108] In step S4, x is 4, and the thresholds include the loss of flow threshold T1 and the excessive fluidization threshold T2.

[0109] A low baseline spectrum percentage higher than the threshold T1 indicates that the fluidized bed is in a non-fluidized state; a high baseline spectrum percentage higher than the threshold T2 indicates that the fluidized bed is in an over-fluidized state; and both the low baseline and high baseline spectrum percentages are lower than the thresholds, indicating that the fluidized bed is in a normal fluidized state.

[0110] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for determining the fluidization state of a fluidized bed, characterized in that: The steps include, S1. Spectral acquisition: The diffuse reflectance near-infrared spectral acquisition system is used to acquire the diffuse reflectance near-infrared spectrum of the material inside the fluidized bed monitoring window; S2, Moving window update, the moving window includes n continuously acquired spectra, and the moving step size is m spectra; S3. Spectral classification: Diffused reflectance spectra within the moving window are classified into three categories: high baseline, medium baseline, and low baseline spectra. S4. Flow regime determination: The proportions of high baseline spectrum and low baseline spectrum within x consecutive moving windows are compared with preset thresholds to determine the fluidization regime. S5. Repeat steps S1-4 to achieve online flow regime determination; The spectral classification in step S3 includes the following steps: S31. Spectral clustering: The total absorbance is obtained by summing the absorbance at each wavelength of the spectrum to represent each spectrum. The k-means clustering algorithm is used to divide the spectrum into 2-4 clusters based on the total absorbance. S32. Cluster labeling: Using the inter-cluster spectral similarity and the standard deviation of the total spectral absorbance within a cluster as the criteria, the spectra within each cluster are classified as low-baseline, medium-baseline, or high-baseline spectra. The comparison object for inter-cluster spectral similarity is the average spectrum of the spectrum within each cluster, and the spectral similarity is calculated using the cosine similarity algorithm.

2. The method for determining the fluidization state of a fluidized bed as described in claim 1, characterized in that, The diffuse reflectance near-infrared spectral acquisition system in step S1 includes a near-infrared sensor, a near-infrared detector, and a computer terminal connected in sequence.

3. The method for determining the fluidization state of a fluidized bed as described in claim 2, characterized in that: The near-infrared sensor is fixed to the outside of the fluidized bed monitoring window. The fixed height must ensure that the irradiation area is in the dense phase region of the fluidized bed. Parallel near-infrared beams pass through the glass of the monitoring window, and the two beams are focused on the second end face of the monitoring window.

4. The method for determining the fluidization state of a fluidized bed as described in claim 1, characterized in that: In step S2, n is 5-30 and m is 5-30.

5. The method for determining the fluidization state of a fluidized bed as described in claim 1, characterized in that: In step S2, n is 10-20; m is 1-20.

6. The method for determining the fluidization state of a fluidized bed as described in claim 1, characterized in that: In step S31, the k-means clustering algorithm is used to divide the spectrum into three clusters based on the total absorbance; in step S32, the cosine similarity algorithm is used to calculate the spectral similarity.

7. The method for determining the fluidization state of a fluidized bed as described in claim 1, characterized in that: In step S4, x is 2-8, and the thresholds include the loss of flow threshold T1 and the excessive fluidization threshold T2.

8. The method for determining the fluidization state of a fluidized bed as described in claim 1, characterized in that: A low baseline spectrum percentage higher than the threshold T1 indicates that the fluidized bed is in a non-fluidized state; a high baseline spectrum percentage higher than the threshold T2 indicates that the fluidized bed is in an over-fluidized state; and both the low baseline and high baseline spectrum percentages are lower than the thresholds, indicating that the fluidized bed is in a normal fluidized state.

9. The method for determining the fluidization state of a fluidized bed as described in claim 8, characterized in that: Threshold T1 is 7.3-25.8%; threshold T2 is 2.5-5.6%.

Images