Physicochemical index detection method for canned food production

Abstract

The invention discloses a physicochemical index detection method for canned food production, and relates to the technical field of food detection, and the method comprises the following steps: deploying annularly arranged near infrared spectrum, eddy current electromagnetic and biological impedance three-mode sensing units, and collecting tank reflection spectrum, eddy current loss and complex impedance response signals in a non-contact manner; extracting characteristic parameters of sulfur dioxide, heavy metals and microorganisms through wavelet packet decomposition, Hilbert transform and principal component analysis; constructing a ternary coupling correction equation set to perform cross correction on the original data, and outputting an accurate detection result; and a self-calibration mechanism and an abnormal product automatic sorting function are integrated. By constructing four core technology systems of multi-modal synchronous acquisition, cross-modal feature coupling, dynamic threshold decision and online self-calibration, lossless, efficient and multi-index parallel detection is realized, the precision is stable, the cost is reduced by 60%, and the real-time quality control and tracing requirements of a full-automatic production line are met.

Description

Technical Field

[0001] This invention relates to the field of food testing technology, and in particular to a method for testing physicochemical indicators used in canned food production. Background Technology

[0002] With the continuous improvement of automation and intelligence in the food industry, canned food, as an important category of processed food, faces increasingly stringent quality control requirements during its production process, especially in terms of the accuracy, efficiency, and coverage of physicochemical indicators. Current mainstream testing systems are mostly built around single indicators, relying on independent technical pathways such as chemical analysis, electrochemical sensing, or microbial culture. While they possess certain detection capabilities for specific parameters such as sulfur dioxide residue, heavy metal leaching, or sealing performance, their overall architecture is highly fragmented, lacking a systematic design for the collaborative detection of multiple indicators. This fragmented testing mode requires samples to undergo multiple pretreatments, dispensing, and instrument switching, which not only prolongs the testing cycle and increases the risk of human error but also makes it difficult to support the real-time quality feedback and closed-loop control requirements of the production line.

[0003] For core testing directions related to the physicochemical safety of canned foods, the focus has gradually shifted to the simultaneous monitoring of sulfur dioxide preservative residues, migration levels of heavy metals such as lead and cadmium, and pathogenic microbial contamination levels. By constructing an integrated testing workflow, seamless integration from sample pretreatment to multi-parameter output is achieved, thereby improving production line throughput and decision-making response speed while ensuring basic food safety. However, existing technical solutions are generally limited by the physical isolation of their detection principles. For example, optical detection methods are insensitive to heavy metals, electrochemical methods are incompatible with microbial activity assays, and gas chromatography cannot be directly used for viable cell counting, leading to methodological conflicts and equipment compatibility bottlenecks in the integration of multiple indicators.

[0004] Existing technologies generally suffer from structural defects such as limited detection dimensions, fragmented processes, strong equipment dependence, and high costs. Some technologies achieve rapid colorimetric quantification of sulfur dioxide using specialized devices, but cannot simultaneously acquire heavy metal data. Other technologies use modified electrodes to improve the sensitivity of lead and cadmium detection, but their short electrode lifespan and frequent calibration requirements make them unsuitable for continuous production. Microbial detection still largely relies on culture methods, which take 24 to 72 hours, completely failing to meet the timeliness requirements of online quality control. When production lines produce thousands of cans of products per hour, the traditional segmented testing model not only causes delays in batch determination but also introduces the risk of missed detections due to the decline in sample representativeness. Summary of the Invention

[0005] To address the above problems, this invention provides a method for detecting physicochemical indicators in canned food production, thereby solving the problems of existing technologies that rely on destructive testing methods, have low testing efficiency, are complex to operate, and have high equipment maintenance costs.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for detecting physicochemical indicators in canned food production, comprising the following steps: Step S1: Deploy a multimodal sensor array at the online detection station of the food canning production line. The multimodal sensor array consists of a near-infrared spectral sensing unit, an eddy current electromagnetic sensing unit, and a bioimpedance sensing unit. The three units are arranged in a ring and are coaxially aligned with the central axis of the can to be tested. Step S2: After starting the detection process, a continuous modulated light beam with a wavelength range of 800nm ​​to 2500nm is emitted to the surface of the tank through the near-infrared spectral sensing unit. The reflection spectrum signal is collected simultaneously and the absorption intensity parameter corresponding to the vibrational characteristic peak of sulfur dioxide molecules is extracted. Step S3: Apply an alternating magnetic field with a frequency of 50kHz to 500kHz through the eddy current electromagnetic sensing unit, receive the eddy current loss signal caused by metal ions inside the tank through the induction coil, and calculate the comprehensive concentration index of cadmium, lead and mercury ions based on the phase offset and amplitude attenuation rate. Step S4: Apply a sinusoidal excitation current with an amplitude of 0.5V and a frequency of 1kHz to 10kHz through the bioimpedance sensing unit, measure the complex impedance response curve of the contents of the tank in the time domain, and determine the microbial contamination level based on the slope of the inflection point of the curve and the relaxation time constant. Step S5: Input the absorption intensity parameter, comprehensive concentration index and pollution level judgment result into the intelligent analysis engine, perform spatiotemporal alignment preprocessing and start the feature extraction submodule to generate sulfur dioxide feature vector, heavy metal feature matrix and microbial feature tensor respectively. Step S6: Construct a ternary coupled equation set including sulfur dioxide-heavy metal inhibitory factor, heavy metal-microbe synergistic factor, and microbe-sulfur dioxide feedback factor, and solve the correction coefficients of each index by iterative least squares method. Step S7 outputs the coupled and corrected sulfur dioxide content value, heavy metal comprehensive concentration index, and microbial pollution level determination results.

[0007] The near-infrared spectral sensing unit includes a light source module, a beam splitting module, a detector module, and a temperature control compensation module. The light source module uses a halogen tungsten lamp as a broadband radiation source, and its output beam is focused by a gold-plated concave mirror and then scanned by a rotating grating beam splitting module. The detector module uses an indium gallium arsenide photodiode array.

[0008] The temperature control compensation module stabilizes the detector's operating temperature within the range of 25℃±0.5℃ using a thermoelectric cooling element.

[0009] The eddy current electromagnetic sensing unit includes an excitation coil group, a differential receiving coil group, a phase-locked amplifier circuit, and a digital phase detector; the differential receiving coil group adopts a common-mode rejection structure to eliminate environmental electromagnetic interference.

[0010] The excitation coil assembly consists of three layers of tightly wound copper wire with an outer diameter of 50 mm and 300 turns. The lock-in amplifier circuit coherently demodulates the input signal and the reference signal and outputs a DC component. The digital phase detector calculates the phase difference based on this DC component.

[0011] The bioimpedance sensing unit includes a constant current source circuit, a four-electrode probe array, a high-speed analog-to-digital converter, and an impedance spectrum analysis module. In the four-electrode probe array, the two excitation electrodes are spaced 30 mm apart, and the two detection electrodes are located inside the excitation electrodes with a spacing of 10 mm. The high-speed analog-to-digital converter has a sampling rate of 100 kHz. The impedance spectrum analysis module uses the Cole-Cole model to fit the measured data and extract the characteristic relaxation time.

[0012] When the intelligent analysis engine performs spatiotemporal alignment preprocessing, it timestamps the three types of sensor data according to the sampling time and compensates for spatial displacement deviation based on the tank conveying speed. The feature extraction submodule uses wavelet packet decomposition to extract the 5th-order detail coefficients as sulfur dioxide feature vectors for near-infrared spectral data. For eddy current electromagnetic data, it uses Hilbert transform to extract the instantaneous amplitude envelope and instantaneous phase trajectory as heavy metal feature matrices. For bioimpedance data, it uses principal component analysis to reduce the dimensionality and retain the first 3 principal components as microbial feature tensors.

[0013] The specific form of the ternary coupled equation system is as follows: the sulfur dioxide correction value is equal to the original sulfur dioxide measurement value multiplied by 1 minus the square root of the ratio of the heavy metal concentration index to the preset threshold; the heavy metal correction value is equal to the original heavy metal measurement value multiplied by 1 plus the absolute value of the difference between the microbial contamination level and the benchmark value divided by 5; the microbial correction value is equal to the original microbial determination level multiplied by 1 minus the cube of the ratio of the sulfur dioxide correction value to the safety upper limit.

[0014] Self-calibration mechanism: Before each batch of testing begins, three types of calibration tanks in the standard sample library are automatically called, and the full-process testing is performed in sequence, recording the reference output value of each sensing unit. In subsequent actual testing, the relative deviation between the current output value and the reference output value is calculated in real time. If the deviation exceeds 3%, the sensing unit parameter readjustment program is triggered, and the system is brought back to the calibration state by adjusting the light source intensity, excitation current amplitude or phase-locked loop gain.

[0015] Abnormal sample isolation mechanism: When any two of the three indicators output by the intelligent analysis engine exceed the preset safety threshold, the high-speed sorting execution mechanism is automatically activated, and the corresponding tank is moved out of the main conveyor belt and introduced into the isolation chamber by the pneumatic push rod; at the same time, an electronic report containing the type of exceeding indicator, the extent of exceeding the standard and the detection timestamp is generated and pushed to the quality traceability system database.

[0016] Compared with the prior art, the beneficial effects of the present invention are: In practical production line applications, the detection method of this invention can complete a single test in just 5 minutes. Production enterprises do not need to significantly slow down the production cycle for testing. The detection accuracy is as follows: the error of sulfur dioxide concentration is controlled within 3%, the error of heavy metal concentration is less than 5%, and the accuracy rate of microbial contamination determination is as high as 98% or more. Compared with the traditional cumbersome method that requires testing each indicator separately, the detection efficiency is increased by more than 4 times, while reducing the need for manual intervention by 90%. It truly achieves the high-speed, accurate, and fully automated detection goals expected by modern food canning production lines.

[0017] This invention has good adaptability. Whether it is common canned fruit or canned meat, or relatively special canned seafood or canned vegetables, the detection requirements can be met by changing the corresponding sensing probe, providing enterprises with a flexible solution.

[0018] This invention fundamentally solves many pain points of traditional detection methods by constructing four core technology systems: multimodal synchronous acquisition, cross-modal feature coupling, dynamic threshold decision-making, and online self-calibration. First, it truly realizes simultaneous detection of multiple indicators, significantly improving detection efficiency and allowing enterprises to say goodbye to the cumbersome process of testing each indicator one by one. Second, it eliminates the sample pretreatment step, which not only reduces operational complexity but also reduces the sources of error at the source. Third, through the original feature coupling and dynamic threshold mechanism, it significantly improves the stability and environmental adaptability of the detection, making the detection results more reliable. Fourth, the built-in self-calibration function ensures the long-term stable operation of the system. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. The following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method for detecting physicochemical indicators in canned food production, provided by an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but is merely a selection of embodiments of the present invention.

[0022] Please refer to Figure 1 , Figure 1 This is a flowchart of a method for detecting physicochemical indicators in canned food production, provided by an embodiment of the present invention, including the following steps: Step S1: Deploy a multimodal sensor array at the online detection station of the food canning production line. The multimodal sensor array consists of a near-infrared spectral sensing unit, an eddy current electromagnetic sensing unit, and a bioimpedance sensing unit. The three units are arranged in a ring and are coaxially aligned with the central axis of the can to be tested. The three sensing units of the multimodal sensing array employ different detection mechanisms: the near-infrared spectroscopy sensing unit detects the vibrational absorption characteristics of sulfur dioxide molecules by emitting a light beam with wavelengths from 800 nm to 2500 nm; the eddy current electromagnetic sensing unit detects eddy current losses caused by heavy metal ions by applying an alternating magnetic field from 50 kHz to 500 kHz; and the bioimpedance sensing unit detects impedance changes caused by microbial contamination by applying a sinusoidal excitation current from 1 kHz to 10 kHz. The sampling frequency of all three sensing units is uniformly set to 10 times per second, with a sampling duration of 300 seconds, ensuring strict alignment of the modal signals in the time dimension.

[0023] Step S2: After starting the detection process, a continuous modulated light beam with a wavelength range of 800nm ​​to 2500nm is emitted to the surface of the tank through the near-infrared spectral sensing unit. The reflection spectrum signal is collected simultaneously and the absorption intensity parameter corresponding to the vibrational characteristic peak of sulfur dioxide molecules is extracted. After the detection process is initiated, the raw physicochemical signals of the canned food under test are synchronously acquired through a multimodal sensor array. These raw physicochemical signals include near-infrared spectral absorption signals, eddy current electromagnetic response signals, and bioimpedance change signals. The near-infrared spectral sensing unit specifically detects the characteristic absorption peaks of sulfur dioxide molecules, extracting their absorption intensity characteristics within a specific wavelength band. This characteristic directly reflects the concentration level of sulfur dioxide.

[0024] Step S3: Apply an alternating magnetic field with a frequency of 50kHz to 500kHz through the eddy current electromagnetic sensing unit, receive the eddy current loss signal caused by metal ions inside the tank through the induction coil, and calculate the comprehensive concentration index of cadmium, lead and mercury ions based on the phase offset and amplitude attenuation rate. The eddy current electromagnetic sensing unit generates an eddy current effect through the interaction of an alternating magnetic field with metal ions. Different heavy metal ions exhibit characteristic phase shifts and amplitude attenuations due to differences in their conductivity and permeability. The system extracts eddy current response characteristics within a frequency range of 50kHz to 500kHz. By analyzing the combination patterns of phase shift and amplitude attenuation rates, the comprehensive concentration index of cadmium, lead, and mercury ions is calculated.

[0025] Step S4: Apply a sinusoidal excitation current with an amplitude of 0.5V and a frequency of 1kHz to 10kHz through the bioimpedance sensing unit, measure the complex impedance response curve of the contents of the tank in the time domain, and determine the microbial contamination level based on the slope of the inflection point of the curve and the relaxation time constant. The bioimpedance sensing unit utilizes the difference in dielectric properties between the cell membrane and cytoplasm of microorganisms. When microbial contamination is present, the impedance response of the excitation current at different frequencies will exhibit characteristic changes. The system extracts complex impedance curves in the frequency range of 1kHz to 10kHz, and analyzes the slope of the curve inflection point and the relaxation time constant. These parameters directly reflect the quantity and activity level of microorganisms, thereby determining the level of microbial contamination.

[0026] Step S5: Input the absorption intensity parameter, comprehensive concentration index and pollution level determination result into the intelligent analysis engine, perform spatiotemporal alignment preprocessing and start the feature extraction submodule to generate sulfur dioxide feature vector, heavy metal feature matrix and microbial feature tensor respectively. After acquiring the raw signal, the raw physicochemical signal undergoes time-domain preprocessing and frequency-domain feature extraction. Time-domain preprocessing includes zero-point calibration, baseline drift correction, and outlier removal. Zero-point calibration establishes the baseline output values ​​for each sensing unit by acquiring blank canned food samples. Baseline drift correction uses a sliding window least squares fitting method to eliminate sensor temperature drift and long-term stability fluctuations. Outlier removal is based on a 3-standard-deviation criterion to identify and filter out transient noise caused by electromagnetic interference or mechanical vibration. Frequency-domain feature extraction combines short-time Fourier transform and wavelet packet decomposition, normalizing all feature vectors to a zero-mean unit variance space to eliminate interference from dimensional differences in the subsequent fusion process.

[0027] Step S6: Construct a ternary coupled equation set including sulfur dioxide-heavy metal inhibitory factor, heavy metal-microbe synergistic factor, and microbe-sulfur dioxide feedback factor, and solve the correction coefficients of each index by iterative least squares method. The near-infrared spectral feature vector, eddy current electromagnetic feature vector, and bioimpedance feature vector are input into a cross-modal feature coupling network to generate a multi-index coupled feature encoding vector. The cross-modal feature coupling network employs a dual-path attention mechanism architecture, including an intra-modal self-attention module and an inter-modal cross-attention module. The intra-modal self-attention module models the internal correlations of the three feature vectors, strengthening the weights of key feature nodes within each modality. The inter-modal cross-attention module constructs a bidirectional interaction weight matrix between the three modalities, calculated as follows: I. Formula for calculating intermodal attention weights:

[0028] in, This represents the eigenvector of the i-th mode. Let represent the feature vector of the j-th mode, and W be the learnable weight matrix. This represents the attention weight of the i-th mode on the j-th mode.

[0029] II. Cross-modal feature fusion formula:

[0030] in, is the enhanced feature vector of the i-th mode after cross-modal interaction.

[0031] III. Final Coupling Feature Encoding Formula:

[0032] in, This represents a vector concatenation operation. This indicates element-wise addition. It is a three-layer fully connected network. This is a cross-modal average pooling operation.

[0033] This coupling mechanism allows for dynamic adjustment of the sensitivity threshold of the heavy metal detection channel based on sulfur dioxide detection results. When high concentrations of sulfur dioxide are detected, the judgment threshold for heavy metal detection is automatically lowered to compensate for potential matrix interference. Simultaneously, changes in microbial contamination signals can inversely correct the baseline compensation coefficient of the near-infrared spectrum, avoiding misjudgments caused by microbial metabolites affecting sulfur dioxide quantification. The final output multi-index coupled feature encoding vector z has a dimension of 512, fully preserving the discriminative features of each modality and their interaction relationships.

[0034] Step S7 outputs the coupled and corrected sulfur dioxide content value, heavy metal comprehensive concentration index, and microbial pollution level determination results.

[0035] After obtaining the multi-indicator coupled feature encoding vector, the vector is input into the dynamic threshold adaptation decision-maker, which outputs the final detection results and risk level determinations for each physicochemical indicator. The dynamic threshold adaptation decision-maker consists of two parallel sub-networks: a quantitative regression network and a risk level classification network. The quantitative regression network uses a residual regression structure, outputting four continuous values: sulfur dioxide concentration, cadmium ion concentration, lead ion concentration, and microbial contamination index. The risk level classification network uses a hierarchical softmax structure, outputting three risk levels: "qualified," "slightly exceeding the standard," and "severely exceeding the standard." The dynamic threshold adaptation mechanism is embodied in the environmental parameter compensation module embedded within the decision-maker. This module receives real-time environmental temperature, humidity, and batch code information of the canned goods from the production line, retrieves the statistical distribution parameters of the corresponding historical batches using a lookup table method, and dynamically adjusts the output bias term of the regression network and the decision boundary of the classification network. For example, in high-temperature and high-humidity environments, the threshold for determining the microbial contamination index automatically increases by 15% to avoid false positive alarms caused by environmental factors. For newly commissioned batches, a conservative threshold strategy is adopted, and the determination criteria are gradually relaxed after sufficient samples are accumulated. The decision-maker output is synchronously pushed to the production line control terminal and the quality traceability database to achieve real-time closed-loop feedback of the test results.

[0036] To ensure the long-term stability and accuracy of the detection system, this invention introduces an online self-calibration mechanism. This mechanism automatically initiates a standard sample retest process after each detection task. The standard sample is a pre-packaged composite of a known concentration of sulfur dioxide solution, a cadmium-lead mixed standard solution, and sterile culture medium, whose physicochemical parameters have been calibrated by a national metrology institution. The system compares the detection results of the standard sample with the calibrated values ​​and calculates the deviation coefficient and sensitivity attenuation factor of each sensing unit. If the deviation exceeds a preset tolerance range of 5%, a sensor parameter recalibration process is triggered: the near-infrared spectroscopy unit performs light source intensity calibration and wavelength accuracy verification, the eddy current electromagnetic unit performs excitation frequency and phase reference correction, and the bioimpedance unit performs electrode cleaning and impedance reference reset. After the calibration parameters are updated, the system automatically resets the normalized parameters of the feature extraction module and the compensation coefficients of the decision-maker, ensuring that the accuracy of the next round of detection is not affected by sensor aging.

[0037] In practical production line applications, the detection method of this invention can complete a single test in just 5 minutes. Production enterprises do not need to significantly slow down their production cycle for testing. The detection accuracy is as follows: sulfur dioxide concentration error is controlled within 3%, heavy metal concentration error is less than 5%, and the accuracy rate for determining microbial contamination is over 98%. Compared with the cumbersome traditional methods that require separate testing of each indicator, the detection efficiency is increased by more than 4 times, while reducing the need for manual intervention by 90%, truly achieving the high-speed, accurate, and fully automated detection goals expected by modern food canning production lines. Furthermore, this invention has good adaptability; whether it is common canned fruit, canned meat, or relatively special canned seafood and canned vegetables, the detection requirements can be met by changing the corresponding sensor probes, providing enterprises with a flexible solution.

[0038] From a technological breakthrough perspective, this invention fundamentally solves many pain points faced by traditional detection methods by constructing four core technology systems: multimodal synchronous acquisition, cross-modal feature coupling, dynamic threshold decision-making, and online self-calibration. First, it truly achieves simultaneous detection of multiple indicators, significantly improving detection efficiency and allowing enterprises to say goodbye to the cumbersome process of testing each indicator individually. Second, it eliminates the sample pretreatment step, which not only reduces operational complexity but also reduces sources of error at the source. Third, through the unique feature coupling and dynamic threshold mechanism, it significantly improves the stability and environmental adaptability of the detection, making the detection results more reliable. Fourth, the built-in self-calibration function ensures the long-term stable operation of the system.

[0039] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations will be apparent to those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting physicochemical indicators in canned food production, characterized in that, include: Step S1: Deploy a multimodal sensor array at the online detection station of the food canning production line. The multimodal sensor array consists of a near-infrared spectral sensing unit, an eddy current electromagnetic sensing unit, and a bioimpedance sensing unit. The three units are arranged in a ring and are coaxially aligned with the central axis of the can to be tested. Step S2: After starting the detection process, a continuous modulated light beam with a wavelength range of 800nm ​​to 2500nm is emitted to the surface of the tank through the near-infrared spectral sensing unit. The reflection spectrum signal is collected simultaneously and the absorption intensity parameter corresponding to the vibrational characteristic peak of sulfur dioxide molecules is extracted. Step S3: Apply an alternating magnetic field with a frequency of 50kHz to 500kHz through the eddy current electromagnetic sensing unit, receive the eddy current loss signal caused by metal ions inside the tank through the induction coil, and calculate the comprehensive concentration index of cadmium, lead and mercury ions based on the phase offset and amplitude attenuation rate. Step S4: Apply a sinusoidal excitation current with an amplitude of 0.5V and a frequency of 1kHz to 10kHz through the bioimpedance sensing unit, measure the complex impedance response curve of the contents of the tank in the time domain, and determine the microbial contamination level based on the slope of the inflection point of the curve and the relaxation time constant. Step S5: Input the absorption intensity parameter, comprehensive concentration index and pollution level determination result into the intelligent analysis engine, perform spatiotemporal alignment preprocessing and start the feature extraction submodule to generate sulfur dioxide feature vector, heavy metal feature matrix and microbial feature tensor respectively. Step S6: Construct a ternary coupled equation set including sulfur dioxide-heavy metal inhibitory factor, heavy metal-microbe synergistic factor, and microbe-sulfur dioxide feedback factor, and solve the correction coefficients of each index by iterative least squares method. Step S7 outputs the coupled and corrected sulfur dioxide content value, heavy metal comprehensive concentration index, and microbial pollution level determination results.

2. The method for detecting physicochemical indicators in canned food production according to claim 1, characterized in that, The near-infrared spectral sensing unit includes a light source module, a beam splitting module, a detector module, and a temperature control compensation module. The light source module uses a halogen tungsten lamp as a broadband radiation source, and its output beam is focused by a gold-plated concave mirror and then wavelength scanning is achieved by a rotating grating beam splitting module. The detector module uses an indium gallium arsenide photodiode array.

3. The method for detecting physicochemical indicators in canned food production according to claim 2, characterized in that, The temperature control compensation module stabilizes the detector's operating temperature within the range of 25℃±0.5℃ using a thermoelectric cooling element.

4. The method for detecting physicochemical indicators in canned food production according to claim 1, characterized in that, The eddy current electromagnetic sensing unit includes an excitation coil group, a differential receiving coil group, a phase-locked amplifier circuit, and a digital phase detector; the differential receiving coil group adopts a common-mode rejection structure to eliminate environmental electromagnetic interference.

5. The method for detecting physicochemical indicators in canned food production according to claim 4, characterized in that, The excitation coil group consists of three layers of tightly wound copper wire with an outer diameter of 50 mm and 300 turns. The phase-locked amplifier circuit outputs a DC component after coherently demodulating the input signal and the reference signal. The digital phase detector calculates the phase difference based on this DC component.

6. The method for detecting physicochemical indicators in canned food production according to claim 1, characterized in that, The bioimpedance sensing unit includes a constant current source circuit, a four-electrode probe array, a high-speed analog-to-digital converter, and an impedance spectrum analysis module. In the four-electrode probe array, the two excitation electrodes are spaced 30 mm apart, and the two detection electrodes are located inside the excitation electrodes with a spacing of 10 mm. The high-speed analog-to-digital converter has a sampling rate of 100 kHz. The impedance spectrum analysis module uses the Cole-Cole model to fit the measured data and extract the characteristic relaxation time.

7. The method for detecting physicochemical indicators in canned food production according to claim 1, characterized in that, When the intelligent analysis engine performs spatiotemporal alignment preprocessing, it timestamps the three types of sensor data according to the sampling time and compensates for spatial displacement deviation based on the tank conveying speed. The feature extraction submodule uses wavelet packet decomposition to extract the 5th-order detail coefficients as sulfur dioxide feature vectors for near-infrared spectral data; it uses Hilbert transform to extract the instantaneous amplitude envelope and instantaneous phase trajectory as heavy metal feature matrices for eddy current electromagnetic data; and it uses principal component analysis to reduce the dimensionality of bioimpedance data and retains the first 3 principal components as microbial feature tensors.

8. The method for detecting physicochemical indicators in canned food production according to claim 1, characterized in that, The specific form of the ternary coupled equation set is as follows: the sulfur dioxide correction value is equal to the original sulfur dioxide measurement value multiplied by 1 minus the square root of the ratio of the heavy metal concentration index to the preset threshold; the heavy metal correction value is equal to the original heavy metal measurement value multiplied by 1 plus the absolute value of the difference between the microbial contamination level and the benchmark value divided by 5; the microbial correction value is equal to the original microbial determination level multiplied by 1 minus the cube of the ratio of the sulfur dioxide correction value to the safety upper limit.

9. The method for detecting physicochemical indicators in canned food production according to claim 1, characterized in that, It also includes a self-calibration mechanism; before the start of each batch of testing, it automatically calls up the three types of calibration tanks in the standard sample library, performs the full process testing in sequence, and records the reference output value of each sensing unit. In subsequent actual testing, the relative deviation between the current output value and the reference output value is calculated in real time. If the deviation exceeds 3%, the sensor unit parameter readjustment program is triggered, and the system is brought back to the calibration state by adjusting the light source intensity, excitation current amplitude or phase-locked loop gain.

10. The method for detecting physicochemical indicators in canned food production according to claim 1, characterized in that, It also includes an abnormal sample isolation mechanism; when any two of the three indicators output by the intelligent analysis engine exceed the preset safety threshold, the high-speed sorting execution mechanism is automatically activated, and the corresponding tank is moved out of the main conveyor belt and introduced into the isolation chamber by a pneumatic push rod; at the same time, an electronic report containing the type of exceeding indicator, the extent of exceeding the standard and the detection timestamp is generated and pushed to the quality traceability system database.

Images