Fruit juice quality on-line monitoring method and system based on spectrum and AI

Abstract

The invention discloses a fruit juice quality online monitoring method and system based on spectrum and AI, relates to the technical field of fruit juice quality detection, solves the technical problem that in the prior art, the equipment state of a spectrum detection end cannot be accurately diagnosed, and particularly relates to a method and system for monitoring the quality of fruit juice by using spectrum detection, AI analysis and full-process collaboration as the core. An integrated monitoring system composed of a spectrum detection end, an online monitoring platform, a self-adaptive loading unit and a production line quality control unit is constructed; equipment state diagnosis and multi-dimensional spectrum data acquisition are completed through a spectrum detection end, an on-line monitoring platform issues an overall instruction, processes data and feeds back signals, a self-adaptive loading unit dynamically adapts to scene changes of a production line, and a production line quality control unit focuses on key processes to implement accurate intervention.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fruit juice quality detection, specifically an online monitoring method and system for fruit juice quality based on spectrum and AI. BACKGROUND

[0002] The online monitoring method for fruit juice quality based on spectrum and AI realizes real-time and non-destructive monitoring of key indicators in the fruit juice production process by integrating the rapid detection capability of spectrum technology and the intelligent analysis advantage of artificial intelligence.

[0003] The patent with application number 202311158698.5 discloses a real-time monitoring method for the freshness of compound fruit juice, which includes: for a single type of fruit juice, performing physical and chemical characteristic analysis to obtain accurate fruit juice characteristic parameters; associating the taste data of the fruit juice with the shelf life of the fruit juice to determine the correlation between different taste attributes and the shelf life; obtaining the taste prediction result of the compound fruit juice through a compound fruit juice taste prediction model, and obtaining the shelf life prediction result of the compound fruit juice according to a fruit juice shelf life prediction model; comparing the prediction result with the set shelf life standard and taste standard, and using chi-square test and T-test statistical analysis methods to determine whether the compound fruit juice meets the standard.

[0004] However, the existing technology has the following technical problems:

[0005] Firstly, it is impossible to accurately diagnose the state of the spectrum detection end through horizontal and vertical comparison parameters; secondly, it is impossible to focus on the key quality influencing process of filling to realize full-parameter monitoring of the filling process, covering core indicators such as flow rate, filling amount, and packaging pressure, and it is impossible to comprehensively control the quality risk of the filling link.

[0006] In view of the above technical defects, a solution is proposed. SUMMARY

[0007] The purpose of the present application is to solve the above-mentioned problems by proposing an online monitoring method and system for fruit juice quality based on spectrum and AI.

[0008] The purpose of the present application can be achieved by the following technical solutions:

[0009] The online monitoring system for fruit juice quality based on spectrum and AI includes an online monitoring platform, wherein the online monitoring platform is in communication connection with a spectrum detection end, and is also in communication connection with an adaptive loading unit and a production line quality control unit;

[0010] The spectral detection end performs spectral detection on the juice processing production line, and performs equipment detection and data analysis, specifically: the spectral detection end performs spectral data acquisition on the juice in the production line, and detects the acquisition process to determine whether the equipment detection is qualified, and performs juice quality detection on the spectral equipment collected data after being qualified;

[0011] The adaptive loading unit detects and controls when the production line processing scene changes;

[0012] The production line quality control unit detects and controls the production line processing quality.

[0013] Further, when the spectral equipment decides the spectral type, the efficiency indicators of different spectral technologies in the same production line are collected, the efficiency indicators are based on single detection time consumption, and the spectral efficiency indicators and non-spectral efficiency indicators are selected according to different spectral technology collection; the advance amount of the selected spectral efficiency indicator is obtained by comparing the indicators, and is marked as a horizontal comparison parameter; the running efficiency indicators of the same spectral equipment in different stages are obtained, the running efficiency indicators are based on detection time consumption, the detection delay amount is obtained by difference calculation according to the corresponding running efficiency indicators in different stages, and is marked as a longitudinal comparison parameter.

[0014] Further, if the horizontal comparison parameter does not exceed the advance amount threshold, or the longitudinal comparison parameter exceeds the detection delay amount threshold, it is inferred that the equipment detection corresponding to the spectral detection end is unqualified, an equipment maintenance signal is generated and sent to the online monitoring platform, and the online monitoring platform maintains the spectral equipment; if the horizontal comparison parameter exceeds the advance amount threshold, and the longitudinal comparison parameter does not exceed the detection delay amount threshold, it is inferred that the equipment detection corresponding to the spectral detection end is qualified, an equipment qualified signal is generated and sent to the online monitoring platform.

[0015] Further, when the equipment detection is determined to be qualified, the spectral equipment collected data is subjected to juice quality detection;

[0016] The spectral data collected by the spectral equipment is recorded, specifically absorbance, transmittance and emissivity; and the spectral data is subjected to data processing to obtain extension data, specifically characteristic peak position, characteristic peak height, peak area and peak width;

[0017] According to the processing juice type of the current production line, the corresponding juice ingredients are obtained, the spectral data of the corresponding juice is obtained according to the ingredient proportion, that is, the preset spectral data is calibrated; and the preset spectral data is set in a range according to the numerical deviation of the juice ingredient monitoring;

[0018] According to the production line processing progress, real-time spectral data of the juice is obtained, and the real-time spectral data is compared with the corresponding preset spectral data range of the same type, and the spectral data not in the preset range is marked as abnormal data, and the abnormal data corresponding to the juice ingredient is marked as a deviation ingredient.

[0019] Further, the occurrence time of the abnormal data is recorded, and the duration of the abnormal data is also recorded;

[0020] If the generation frequency of the occurrence time of the abnormal data exceeds the set frequency threshold, or the duration of the abnormal data exceeds the set duration threshold, it is inferred that the raw material for processing juice on the production line is abnormal, a raw material abnormality signal is generated and sent to the online monitoring platform together with the corresponding deviation component. After the online monitoring platform receives the raw material abnormality signal, the production speed of the production line is reduced and the raw material is checked and replaced. If the generation frequency of the occurrence time of the abnormal data does not exceed the set frequency threshold, and the duration of the abnormal data does not exceed the set duration threshold, it is inferred that the processing of the juice on the production line is abnormal, a processing abnormality signal is generated and sent to the online monitoring platform together with the corresponding processing procedure of the deviation component. After the online monitoring platform receives it, the production line processing is rectified.

[0021] Further, the process of the adaptive loading unit is as follows:

[0022] When the production line starts or the product switching command is issued, in the initial stage when the juice starts to flow through the online spectral detection pool, spectral data acquisition and production signaling acquisition are performed. The production signaling acquisition is represented by reading the current preset product code from the programmable logic controller (PLC) of the production line in real time through the industrial communication interface;

[0023] A plurality of continuous raw spectral data is acquired for average processing to suppress noise, and the averaged spectral data is input into a lightweight spectral fingerprint recognition convolutional neural network (CNN) model. After the model receives the spectral data, a predicted product type identifier is output;

[0024] If the output predicted product type identifier and the preset product code obtained by the PLC are the same type of juice product, it is determined that the PLC control instruction is correct, and the high-precision quantitative analysis model corresponding to the corresponding product code is automatically loaded. If the output predicted product type identifier and the preset product code obtained by the PLC are not the same type of juice product, it is determined that the PLC control instruction is incorrect, a production line control signal is generated and sent to the online monitoring platform, and after the online monitoring platform receives it, the processing of the juice product type is controlled.

[0025] Further, the process of the production line quality control unit is as follows:

[0026] The filling link in a juice processing production line is analyzed, the flow rate and flow of the juice during filling are recorded through a flow sensor, and the liquid level height of the juice during filling is recorded through a liquid level sensor under the current flow rate; and the beverage packaging pressure is collected through a pressure sensor; the flow rate fluctuation frequency generated by uncontrollable factors of the production line equipment during the production line processing is obtained, wherein the uncontrollable factors represent the running vibration of the production line equipment and the vibration frequency generated by external influences; and the juice amount of the flow regulation amount and the corresponding liquid level height deviation during the flow rate fluctuation, the regulation deviation ratio is calculated according to the ratio, that is, the ratio of the flow regulation amount to the height deviation juice amount.

[0027] Further, if the flow rate fluctuation frequency exceeds the set fluctuation frequency threshold, or the regulation deviation ratio does not exceed the regulation ratio threshold, it is inferred that the production line processing has a quality impact, a production line rectification signal is generated and sent to the online monitoring platform;

[0028] If the flow rate fluctuation frequency does not exceed the set fluctuation frequency threshold, and the regulation deviation ratio exceeds the regulation ratio threshold, it is inferred that the production line processing quality is normal, a production line stability signal is generated and sent to the online monitoring platform.

[0029] The spectral detection end performs spectral detection on the juice processing production line, and performs device detection and data analysis,

[0030] Specifically, the spectral detection end performs spectral data acquisition on the juice in the production line, and detects the acquisition process to determine whether the device detection is qualified, and after being qualified, performs juice quality detection on the spectral device collected data;

[0031] The adaptive loading unit detects and controls when the production line processing scene changes;

[0032] The production line quality control unit detects and controls the production line processing quality.

[0033] Compared with the prior art, the beneficial effects of the present application are:

[0034] 1. Through horizontal comparison parameters (differences in different spectral technology efficiency) and vertical comparison parameters (differences in running efficiency of the same device at different stages), the device state of the spectral detection end is accurately diagnosed, the problem of unreasonable spectral technology selection or device performance decline is identified in advance, and the detection failure caused by device failure is avoided; at the same time, spectral core data (absorbance, transmittance, etc.) and extended data (feature peak position, peak area, etc.) are collected, the data dimension is comprehensive, and sufficient basis is provided for juice quality analysis, and the accuracy of component detection is improved.

[0035] Based on the frequency and duration of abnormal data, the raw material abnormality and processing abnormality are accurately distinguished, the production line intervention is more targeted (raw material inspection and replacement or processing procedure rectification), and the production waste caused by blind adjustment is reduced; the preset spectral data range is combined with real-time spectrum comparison to realize rapid positioning of component deviation, especially for monitoring of specific component characteristic wavelengths (such as 1450nm absorption peak corresponding to moisture), so that the quality detection is more targeted.

[0036] 2. For scenarios such as juice type change, the PLC product code is read in real time through an industrial communication interface, and the product type classification is quickly completed by combining a lightweight CNN model, realizing automatic adaptation of scene switching without manual intervention and reducing the detection interruption time during product switching; the average processing of original spectral data effectively suppresses noise and improves data purity, laying a reliable foundation for subsequent model analysis; the lightweight CNN model is responsible for rapid classification, and the high-precision quantitative analysis model (PLS / SVR) is responsible for component quantification, with clear division of labor and consideration of detection speed and analysis accuracy; comparison and calibration of historical reference spectrum and real-time spectrum further corrects detection deviation and ensures the consistency of detection standards for different batches of products; the binding and activation of correction parameters and quantitative analysis model realize dynamic optimization of the detection model, adapt to the differences in spectral characteristics after product switching, and ensure the stability and accuracy of the detection results.

[0037] 3. Focusing on the key quality influencing process of filling, data is collected through multi-dimensional sensors such as flow, liquid level, and pressure to realize full-parameter monitoring of the filling process, covering core indicators such as flow rate, filling volume, and packaging pressure, and fully controlling the quality risk of the filling link; the flow rate floating frequency (considering uncontrollable factors such as equipment vibration and external interference) and adjustment deviation amount ratio analysis are introduced to accurately identify hidden abnormalities in the filling process and provide early warning for problems such as juice spillage and insufficient filling volume, avoiding the flow of unqualified products into the market; the accurate output of production line rectification signals and stable signals realizes dynamic control of the filling process, avoiding the impact of excessive rectification on production efficiency, preventing quality problems caused by abnormal omission, and ensuring product filling consistency and safety. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to facilitate the understanding of those skilled in the art, the present application will be further described below with reference to the accompanying drawings.

[0039] Figure 1 The system principle diagram of the present application. DETAILED DESCRIPTION

[0040] In order to enable personnel in the technical field to better understand the present application scheme, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0041] Reference herein to "embodiment" means that the specific features, structures, or characteristics described in connection with the embodiment can be included in at least one embodiment of the present application. The phrase appears at various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It is explicitly and implicitly understood that the embodiments described herein can be combined with other embodiments.

[0042] Please refer to Figure 1 As shown, the online monitoring platform generates a detection instruction to the spectrum detection end, and the spectrum detection end performs spectrum detection on the juice processing production line and executes device detection and data analysis.

[0043] At the same time, the online monitoring platform is in communication connection with an adaptive loading unit to cope with various processing scenarios of the juice processing production line, ensure the accuracy of the spectrum detection of the juice, and improve the monitoring efficiency of the quality of the juice. In addition, the online monitoring platform is in communication connection with a production line quality control unit.

[0044] Embodiment one

[0045] This embodiment detects the quality according to the whole production line in combination with the spectrum detection end. The online monitoring platform generates a detection instruction to the spectrum detection end. The spectrum detection end performs spectrum detection on the juice processing production line and executes device detection and data analysis.

[0046] The specific process is as follows:

[0047] The spectrum detection end performs spectrum data collection on the juice in the production line and detects the collection process.

[0048] When the spectrum device decides the spectrum type, the efficiency index of different spectrum technologies in the same production line is collected. The efficiency index is accurate in terms of single detection time, and the selected spectrum efficiency index and the non-selected spectrum efficiency index are selected according to different spectrum technology collection. The selected spectrum efficiency index is obtained in advance according to the index comparison, and is marked as a horizontal comparison parameter.

[0049] Obtain the running efficiency indicators of the same spectral equipment at different stages, the running efficiency indicators are based on the detection time delay, and the detection delay is obtained by difference calculation of the corresponding running efficiency indicators at different stages, and is marked as a longitudinal comparison parameter;

[0050] Compare the transverse comparison parameter and the longitudinal comparison parameter:

[0051] If the transverse comparison parameter does not exceed the advance amount threshold, or the longitudinal comparison parameter exceeds the detection delay threshold, it is inferred that the corresponding equipment of the spectral detection end is unqualified, that is, the spectral technology type selection is unreasonable or the equipment detection performance is degraded, an equipment maintenance signal is generated and sent to the online monitoring platform, and the online monitoring platform maintains the spectral equipment;

[0052] If the transverse comparison parameter exceeds the advance amount threshold, and the longitudinal comparison parameter does not exceed the detection delay threshold, it is inferred that the corresponding equipment of the spectral detection end is qualified, an equipment qualified signal is generated and sent to the online monitoring platform;

[0053] Determine that the equipment detection is qualified, and perform juice quality detection on the spectral equipment collected data;

[0054] Record the spectral data collected by the spectral equipment, specifically absorbance, transmittance and emissivity; and process the spectral data to obtain extended data, specifically characteristic peak position, characteristic peak height, peak area, peak width, wherein the spectral data and the extended data proposed in the present application are not all types of spectral data, in actual use scenarios, different spectral technology corresponds to different data types, but they are all applicable to the present system;

[0055] According to the processing juice type of the current production line, the corresponding juice components are obtained, the spectral data of the corresponding juice is obtained according to the component proportion, that is, the preset spectral data is calibrated; and the preset spectral data is set in a range according to the numerical deviation of the juice component monitoring value;

[0056] According to the production line processing progress, real-time spectral data of the juice is obtained, and the real-time spectral data is compared with the corresponding preset spectral data range of the same type, and the spectral data not in the preset range is marked as abnormal data, and the corresponding juice component of the abnormal data is marked as a deviation component; For example: characteristic peak position, characteristic wavelength corresponding to a specific component (such as -OH group near near-infrared 1450nm has an absorption peak, corresponding to water); Peak height: the peak value of the light signal at the characteristic wavelength, quantifying the component content (the higher the peak, the higher the component concentration may be); The area under the characteristic peak curve is more stable than the peak height, and is often used for quantitative analysis of complex components (such as when multiple sugars are mixed);

[0057] Record the time when the abnormal data appears, and record the duration of the abnormal data;

[0058] If the generation frequency of the abnormal data occurrence time exceeds the set frequency threshold, or the duration of the abnormal data exceeds the set duration threshold, it is inferred that the raw material for processing juice on the production line is abnormal, a raw material abnormality signal is generated and sent to the online monitoring platform together with the corresponding deviation component, and after the online monitoring platform receives the raw material abnormality signal, the production line production speed is reduced and the raw material is checked and replaced;

[0059] If the generation frequency of the abnormal data occurrence time does not exceed the set frequency threshold, and the duration of the abnormal data does not exceed the set duration threshold, it is inferred that the processing of the juice on the production line is abnormal, a processing abnormality signal is generated and sent to the online monitoring platform together with the corresponding deviation component, and after the online monitoring platform receives, the production line processing is rectified;

[0060] Embodiment two

[0061] This embodiment detects and controls when the production line processing scene changes. If the juice type is changed, the online monitoring platform generates an adaptive loading signal and sends it to the adaptive loading unit;

[0062] When the production line starts or the product switching command is issued, in the initial stage of the juice flowing through the online spectrum detection pool (for example, within the first 30 seconds after starting to flow), spectrum data acquisition and production signaling acquisition are performed. Production signaling acquisition means reading the current product code set through the industrial communication interface (such as OPCUA, ModbusTCP) from the programmable logic controller (PLC) of the production line in real time; This code uniquely corresponds to a juice product (such as "1001" representing 100% pure apple juice, "1002" representing orange juice beverage);

[0063] Acquire multiple continuous raw spectrum data for average processing to suppress noise, and input the averaged spectrum data into a lightweight spectrum fingerprint recognition convolutional neural network (CNN) model; The core task of this lightweight CNN model is to perform fast product type classification, rather than accurate quantitative analysis; Its training data contains standard spectra of all possible produced juice types, and has learned to identify key distinguishing features (for example, carotenoid absorption region of orange juice, specific organic acid feature region of apple juice);

[0064] The model outputs a predicted product type identifier after receiving the spectrum data;

[0065] If the output predicted product type identifier and the preset product code obtained by the PLC are the same type of juice product, it is determined that the PLC control instruction is correct, and a high-precision quantitative analysis model corresponding to the product code (such as a PLS or SVR model for predicting sugar content and acidity) is automatically loaded from the model library;

[0066] The average initial spectrum of the current batch is then compared with the historical reference spectrum used when establishing the standard model corresponding to the product code; a set of correction parameters (e.g., baseline offset, multiplicative scatter correction coefficient) is automatically calculated by an algorithm (such as spectral offset algorithm or univariate linear regression); the loaded high-precision quantitative analysis model is bound and activated with the calculated correction parameters; in the subsequent continuous production process, all real-time collected juice spectrum data are first preprocessed by applying the correction parameters, and then input into the activated quantitative analysis model;

[0067] Embodiment Three

[0068] In this embodiment, the line filling process is taken as the quality influence detection process, and the juice quality is detected and controlled; the online monitoring platform produces a line quality control signal and sends it to the line quality control unit; after receiving the line quality control signal, the line quality control unit detects and controls the quality of the line processing product;

[0069] The filling link in the juice processing line is analyzed, the flow rate and flow of the juice during filling are recorded by a flow sensor, and the liquid level of the juice during filling is recorded by a liquid level sensor at the current flow rate;

[0070] The beverage packaging pressure is also collected by a pressure sensor;

[0071] The flow rate fluctuation frequency caused by uncontrollable factors of the line equipment during the line processing process is obtained, where the uncontrollable factors refer to the running vibration of the line equipment and the vibration frequency caused by external influences;

[0072] The juice quantity corresponding to the flow adjustment amount and the height deviation is calculated according to the ratio, and the adjustment deviation ratio is obtained, i.e., the ratio of the flow adjustment amount to the height deviation juice quantity;

[0073] If the flow rate fluctuation frequency exceeds the set fluctuation frequency threshold, or the adjustment deviation ratio does not exceed the adjustment ratio threshold, it is inferred that the line processing has a quality influence, a line rectification signal is generated and sent to the online monitoring platform, and after receiving the signal, the online monitoring platform rectifies the line filling process to control the juice filling influence, avoid juice spilling and polluting the filling environment, and avoid insufficient juice filling quantity affecting the quality of the juice product;

[0074] If the flow rate fluctuation frequency does not exceed the set fluctuation frequency threshold, and the adjustment deviation ratio exceeds the adjustment ratio threshold, it is inferred that the line processing quality is normal, a line stability signal is generated and sent to the online monitoring platform.

[0075] The juice quality online monitoring method based on spectrum and AI is as follows:

[0076] The spectrum detection end performs spectrum detection on the fruit juice processing production line, and performs equipment detection and data analysis,

[0077] Specifically, the spectrum detection end collects spectrum data of the fruit juice in the production line, detects the collection process to determine whether the equipment detection is qualified, and performs fruit juice quality detection on the spectrum data collected by the spectrum equipment after the equipment detection is qualified.

[0078] The adaptive loading unit detects and controls when the production line processing scene changes.

[0079] The production line quality control unit detects and controls the production line processing quality.

[0080] The threshold or the preset value, the preset range and the like are set for result comparison and analysis, so as to determine whether it is good or bad, and the size of the threshold is set according to sample data large model analysis and artificial experience combination to set input storage, and can be adjusted appropriately according to seasonal or common influence conditions.

[0081] The preferred embodiments of the present application disclosed above are only used to help explain the present application. The preferred embodiments do not describe all the details and limit the present application to the specific embodiments. Obviously, many modifications and changes can be made according to the content of the present application. The present application is selected and described in detail to better explain the principles and practical application of the present application, so that those skilled in the art can well understand and use the present application. The present application is limited by the claims and their entire scope and equivalents.

Claims

1. A juice quality online monitoring system based on spectroscopy and AI, characterized in that, It includes an online monitoring platform, which is connected to the spectral detection terminal and also has an adaptive loading unit and a production line quality control unit. The spectral detection end performs spectral detection on the juice processing production line and performs equipment detection and data analysis. Specifically, the spectral detection end collects spectral data of the juice in the production line and detects the collection process to determine whether the equipment detection is qualified. After the equipment is qualified, the data collected by the spectral equipment is used to detect the quality of the juice. The adaptive loading unit performs detection and control when the production line processing scenario changes; The production line quality control unit detects and controls the processing quality of the production line.

2. The online fruit juice quality monitoring system based on spectroscopy and AI according to claim 1, characterized in that, When deciding on the spectral type for the spectroscopic equipment, efficiency indicators of different spectral technologies on the same production line are collected. The efficiency indicators are based on the time consumed in a single detection. Selective spectral efficiency indicators and non-selective spectral efficiency indicators are collected according to different spectral technologies. The lead amount of the selected spectral efficiency indicator is obtained by comparing the indicators and marked as a horizontal comparison parameter. The operating efficiency indicators of the same spectroscopic equipment at different stages are obtained. The operating efficiency indicators are based on the extension of detection time. The detection delay is calculated by the difference between the corresponding operating efficiency indicators at different stages and marked as a vertical comparison parameter.

3. The online fruit juice quality monitoring system based on spectroscopy and AI according to claim 2, characterized in that, If the horizontal comparison parameter does not exceed the lead threshold, or the vertical comparison parameter exceeds the detection delay threshold, it is inferred that the corresponding equipment of the spectral detection end is unqualified, an equipment maintenance signal is generated and sent to the online monitoring platform, and the online monitoring platform performs maintenance on the spectral equipment. If the horizontal comparison parameter exceeds the lead threshold and the vertical comparison parameter does not exceed the detection delay threshold, it is inferred that the corresponding equipment of the spectral detection end is qualified, and an equipment qualification signal is generated and sent to the online monitoring platform.

4. The online fruit juice quality monitoring system based on spectroscopy and AI according to claim 3, characterized in that, After confirming that the equipment has passed inspection, the data collected by the spectral equipment is used to test the quality of the juice. The system records spectral data collected by the spectrometer, specifically absorbance, transmittance, and emissivity; it also processes the spectral data to obtain extended data, specifically characteristic peak position, characteristic peak height, peak area, and peak width; it obtains the corresponding juice components based on the type of juice being processed in the current production line, and then obtains the spectral data of the corresponding juice based on the component ratio, which is calibrated as preset spectral data; and it sets the range of the preset spectral data based on the numerical deviation of the juice components being monitored. Based on the processing progress of the production line, real-time spectral data of the juice is obtained, and the real-time spectral data is compared with the corresponding preset spectral data range. Spectral data that are not within the preset range are marked as abnormal data, and the juice components corresponding to the abnormal data are marked as deviation components.

5. The online fruit juice quality monitoring system based on spectroscopy and AI according to claim 4, characterized in that, Record the time when the abnormal data occurs, and also record the duration of the abnormal data; If the frequency of abnormal data occurrences exceeds a set frequency threshold, or the duration of abnormal data exceeds a set duration threshold, it is inferred that the raw materials used in the juice processing on the production line are abnormal. An abnormal raw material signal is generated and sent to the online monitoring platform along with the corresponding deviation component. Upon receiving the abnormal raw material signal, the online monitoring platform reduces the production line speed and inspects and replaces the raw materials. If the frequency of abnormal data occurrences does not exceed the set frequency threshold, and the duration of abnormal data does not exceed the set duration threshold, it is inferred that the juice processing on the production line is abnormal. An abnormal processing signal is generated and sent to the online monitoring platform along with the corresponding processing step of the deviation component. Upon receiving the signal, the online monitoring platform performs production line rectification.

6. The online fruit juice quality monitoring system based on spectroscopy and AI according to claim 1, characterized in that, The adaptive loading process is as follows: When the production line starts or the product switching command is issued, in the initial stage when the juice begins to flow through the online spectral detection pool, spectral data acquisition and production signal acquisition are performed. Production signal acquisition means reading the currently set product code from the programmable logic controller of the production line in real time through the industrial communication interface. Multiple consecutive raw spectral data are acquired, averaged to suppress noise, and the averaged spectral data is input into a lightweight spectral fingerprint recognition convolutional neural network model; after receiving the spectral data, the model outputs a predicted product type identifier. If the output predicted product type identifier and the preset product code obtained by the PLC are the same type of juice product, then the PLC control instruction is correct, and the high-precision quantitative analysis model corresponding to the product code is automatically loaded; if the output predicted product type identifier and the preset product code obtained by the PLC are not the same type of juice product, then the PLC control instruction is incorrect, the production line control signal is generated and sent to the online monitoring platform, and the online monitoring platform receives it and performs processing juice product type control.

7. The online fruit juice quality monitoring system based on spectroscopy and AI according to claim 1, characterized in that, The process of the production line quality control unit is as follows: The filling process in a juice processing line is analyzed. Flow sensors record the flow rate and volume of the juice during filling, and level sensors record the juice level height at the current flow rate. Pressure sensors collect the beverage packaging pressure. The frequency of flow rate fluctuations caused by uncontrollable factors in the production line equipment is obtained, including vibrations from equipment operation and external influences. The ratio of the flow rate adjustment to the corresponding level height deviation is calculated to obtain the adjustment deviation ratio, i.e., the ratio of the flow rate adjustment to the level height deviation of the juice.

8. The online fruit juice quality monitoring system based on spectroscopy and AI according to claim 7, characterized in that, If the flow rate fluctuation frequency exceeds the set fluctuation frequency threshold, or the adjustment deviation ratio does not exceed the adjustment ratio threshold, it is inferred that there is a quality impact on the production line processing, and the production line rectification signal is sent to the online monitoring platform. If the flow rate fluctuation frequency does not exceed the set fluctuation frequency threshold, and the adjustment deviation ratio exceeds the adjustment ratio threshold, it is inferred that the production line processing quality is normal, and the production line stability signal is sent to the online monitoring platform.

9. A method for online monitoring of fruit juice quality based on spectroscopy and AI, characterized in that, The online monitoring system for juice quality based on spectroscopy and AI as described in any one of claims 1-8 above.

Images