Dehydrated onion drying environment real-time monitoring method based on multi-sensor fusion

Abstract

The present application relates to the field of environmental monitoring, and more particularly to a dehydration onion drying environment real-time monitoring method based on multi-sensor fusion, the method comprising: obtaining multi-dimensional environment data of a sub-region in a drying chamber; calculating spatial dispersion index, state change vector and dynamic consistency index at the current time; calculating an adaptive adjustment coefficient based on the environmental entropy change index, using the adaptive adjustment coefficient to modify the parameters of the Kalman filter, filtering and denoising the normalized state vector to obtain a denoised state vector; determining an operation deviation vector and calculating a local sensitivity enhancement eigenvalue; splicing the denoised state vector and the local sensitivity enhancement eigenvalue to construct an augmented vector, inputting the augmented vector into an anomaly detection model to obtain an anomaly score, and if the anomaly score exceeds a preset threshold and lasts for a preset period, an alarm is triggered. The present application improves the accuracy of early anomaly detection in the drying process.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring, and in particular to a method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion. Background Technology

[0002] As a highly heat-sensitive processed agricultural product, the precision of the drying process for dehydrated onions directly determines the commercial value of the finished product. During industrial continuous drying, subtle fluctuations and changes in environmental parameters such as real-time temperature, relative humidity, circulating air velocity, and volatile organic compound concentration in each temperature zone have a complex nonlinear mapping relationship with the onion's internal crispness, color saturation, and the retention of sulfur compounds. To ensure consistent quality, real-time monitoring of the physical field within the drying chamber is essential.

[0003] Traditional monitoring solutions primarily rely on distributed environmental sensors for data collection and perform data fusion through alarm mechanisms with fixed thresholds or simple algorithms such as weighted averages and arithmetic mean. However, under complex industrial operating conditions, existing technologies suffer from the following problems.

[0004] First, the distribution of the hot air flow field in the drying chamber is affected by the material bulk density, tray arrangement, and duct structure, resulting in sensors in different geographical locations being in heterogeneous sub-environments. Existing technologies often ignore the spatial correlation between sensors; when local wind resistance increases and causes flow field turbulence, the original acquired data will exhibit physical consistency deviations. Second, the drying chamber is in an extreme environment of high temperature, high humidity, and high concentration of volatile gases for extended periods, which easily induces zero-point drift, sensitivity decay, and electromagnetic noise interference in industrial-grade sensors. Traditional filtering algorithms (such as Kalman filtering or low-pass filtering with fixed parameters) have a trade-off between response speed and smoothness: pursuing smoothness will mask real parameter abrupt changes, while pursuing sensitivity will introduce a large amount of random environmental noise, making it difficult to achieve a dynamic balance between noise suppression and preservation of effective process characteristics. Finally, in the early stages of localized charring, browning, or mold growth in onions, the evolution of physical parameters usually exhibits extremely weak, gradual trends rather than abrupt jumps. Existing technologies are limited by feature extraction accuracy, and these crucial early warning signals are often lost in normal process fluctuations. When parameters exceed fixed thresholds and trigger alarms, the material has usually already suffered irreversible quality damage. Therefore, reducing the spatiotemporal gaps between sensor data, achieving in-depth assessment of environmental consistency in high-noise environments, and constructing enhancement mechanisms that can amplify subtle gradual changes have become urgent technical problems to be solved in improving the automation level and yield of dehydrated onion drying. Summary of the Invention

[0005] To address the issue of low accuracy in sensor data fusion during the drying process of dehydrated onions, this invention provides a real-time monitoring method for the drying environment of dehydrated onions based on multi-sensor fusion.

[0006] This invention provides a method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion, employing the following technical solution: A real-time monitoring method for the drying environment of dehydrated onions based on multi-sensor fusion includes the following steps: acquiring multi-dimensional environmental data of each sub-region in the drying chamber and performing normalization preprocessing to obtain the original state vector at the current moment; calculating the spatial dispersion index of the same type of sensors based on the normalized observation values ​​of multiple sensors of the same type at the current moment, and calculating the state change vector based on the original state vectors at the current moment and the previous moment; determining the dynamic consistency index at the current moment by the L2 norm of the spatial dispersion index and the state change vector; the dynamic consistency index is negatively correlated with the mean of the spatial dispersion index, and positively correlated with the L2 norm of the state change vector. The environmental entropy change index is calculated, and an adaptive adjustment coefficient is calculated based on the environmental entropy change index. The parameters of the Kalman filter are corrected using the adaptive adjustment coefficient to filter and denoise the original state vector, resulting in a denoised state vector. The operational deviation vector is determined based on the degree of deviation of the denoised state vector from historical benchmark data, and a local sensitivity enhancement feature value is calculated by combining the dynamic consistency index. The denoised state vector and the local sensitivity enhancement feature value are concatenated to construct an augmented vector. The augmented vector is input into the anomaly detection model to obtain an anomaly score. If the anomaly score exceeds a preset threshold and continues for a preset period, an alarm is triggered.

[0007] This invention addresses the complex conditions of dehydrated onion drying by dynamically evaluating the spatiotemporal consistency of environmental data and adaptively adjusting filtering parameters. This effectively suppresses random noise from sensors caused by high temperature and humidity, and nonlinearly amplifies subtle, gradual abnormalities such as uneven heating or early scorching of onions. This prevents critical warning signals from being masked by normal process fluctuations, thereby improving the accuracy of early abnormality detection during the drying process and ensuring the consistency of the finished onion quality.

[0008] Preferably, the method for obtaining multidimensional environmental data in the drying room includes: the method for obtaining multidimensional environmental data of each sub-area in the drying room includes: by using temperature sensors, humidity sensors, hot air pressure sensors and volatile organic compound sensors deployed in the sub-areas of the drying room to collect real-time temperature values, real-time humidity values, real-time air pressure values ​​and sulfur-containing gas concentration values ​​respectively, and constructing an original state vector.

[0009] By comprehensively collecting multi-dimensional physical and chemical parameters such as temperature, humidity, wind pressure, and sulfur gas concentrations closely related to the volatility characteristics of onions, compared with single-dimensional monitoring, it can more comprehensively and accurately reflect the actual dehydration state and physicochemical evolution process of onions in the hot air flow field, providing reliable multi-source data support for the subsequent construction of high-confidence environmental state characteristics.

[0010] Preferably, the step of calculating the spatial dispersion index is as follows: multiply the weight coefficients of the same type of sensors in different sub-regions by the square of the difference between the corresponding normalized observation value and the normalized mean of the same type of sensors, and sum the product results of all sensors to obtain the spatial dispersion index.

[0011] When the drying room is partially affected by material accumulation or air duct structure interference, resulting in a heterogeneous sub-environment, this calculation method can accurately quantify the degree of spatial difference between different measuring points, sensitively reflect whether the internal hot air flow field is disturbed or the local wind resistance is abnormally increased, thus providing an accurate measure for evaluating the consistency of multi-sensor network sensing data.

[0012] Preferably, the step of determining the dynamic consistency index at the current moment is as follows: calculate the ratio of the mean of the spatial dispersion index to the L2 norm of the state change vector, take the negative value of the ratio and use it as the exponent of the natural constant to perform a power operation to obtain the dynamic consistency index.

[0013] Compared to simple data comparison, this method combines the spatial dispersion with the overall temperature rise or dehumidification dynamic adjustment process in the time dimension, effectively avoiding the misjudgment of reasonable fluctuations within the allowable range of the process as flow field anomalies, and maintaining the stability of the monitoring system in the complex and ever-changing drying cycle.

[0014] Preferably, the step of calculating the environmental entropy change index is as follows: take the natural logarithm of the sum of the dynamic consistency index and the minimum offset, then multiply the logarithm result by the dynamic consistency index and take the negative value to obtain the environmental entropy change index.

[0015] By calculating the environmental entropy change index, the uncertainty of the sensor system's sensing state under high noise background in the drying room can be reflected in real time and sensitively. Slight consistency deviations are transformed into significant numerical changes, providing a reliable decision basis for subsequent dynamic linkage adjustment filtering algorithms.

[0016] Preferably, the step of generating the adaptive adjustment coefficient based on the environmental entropy change index is as follows: calculate the difference between the environmental entropy change index and the benchmark entropy, multiply the difference by a preset sensitivity factor, take the negative value as the exponent of the natural logarithm base, and add the result of the exponentiation to a preset constant and take the reciprocal to obtain the adaptive adjustment coefficient.

[0017] By utilizing the adaptive adjustment coefficient, the sensitivity of the control system to sudden environmental changes can be dynamically adjusted according to the actual fluctuations in the drying environment, ensuring smooth output during normal drying periods and rapid tracking in the event of abnormal changes.

[0018] Preferably, the method for correcting the parameters of the Kalman filter using the adaptive adjustment coefficient is as follows: multiply the preset basic process noise covariance constant matrix with the adaptive adjustment coefficient to obtain the corrected process noise covariance matrix; calculate the difference between the preset constant and the adaptive adjustment coefficient; multiply the difference with the preset basic measurement noise covariance matrix to obtain the corrected measurement noise covariance matrix.

[0019] In extreme application scenarios where onion drying chambers are subjected to long-term high temperature and humidity and accompanied by electromagnetic noise, this method enables the system to dynamically balance between trusting real-time sensor measurements and trusting mathematical model predictions. It effectively eliminates zero-point drift and random noise interference that are easily generated by industrial-grade sensors, and restores the most realistic evolution trajectory of the drying physical environment.

[0020] Preferably, the step of calculating the local sensitivity enhancement feature value is as follows: The denoising state vector is subtracted from the baseline state vector to obtain the running deviation vector; the Euclidean norm of the running deviation vector is calculated and divided by the dynamic consistency index to obtain the local sensitivity enhancement feature value.

[0021] Compared to existing technologies that suffer from delayed early warnings due to limitations in feature extraction accuracy, this method utilizes the physical phenomenon of decreased perception consistency when minor anomalies occur as a denominator reduction term. It nonlinearly and dramatically amplifies the extremely weak temperature and humidity field deviations caused by the physicochemical reactions on the onion surface, enabling the early scorching or moldy signals hidden in the global background noise to be successfully extracted in advance.

[0022] Preferably, the monitoring method further includes: after the alarm is triggered, the logic control unit adjusts the frequency of the inverter of the hot air circulation fan to reduce the wind speed or lower the power compensation value of the electric heater.

[0023] After accurately identifying early subtle anomalies, physical compensation was quickly implemented by adjusting wind speed or heat power, directly preventing irreversible quality damage to the highly heat-sensitive onion material.

[0024] Preferably, the state change vector is obtained by subtracting the original state vector of the current time from the original state vector of the previous time.

[0025] It can intuitively and efficiently capture the overall fluctuation intensity and evolution direction of the hot air flow field and temperature and humidity in the drying room over time during the current sampling period, providing basic state evolution data for the system to determine whether the equipment is currently in a normal process adjustment period or experiencing a sudden anomaly.

[0026] The present invention has the following technical effects: This invention introduces spatial dispersion and environmental entropy change indices to achieve adaptive and linked adjustment of Kalman filter parameters, effectively filtering out spatiotemporal noise. Simultaneously, a local sensitivity enhancement mechanism is constructed to amplify the extremely subtle nonlinear gradual changes in onions, such as early scorching and browning, solving the problem that key early warning signals are easily masked by normal fluctuations and improving the accuracy of monitoring results. Attached Figure Description

[0027] Figure 1 This is a flowchart of the real-time monitoring method for the drying environment of dehydrated onions based on multi-sensor fusion, according to the present invention.

[0028] Figure 2 This is a diagram showing the effect of the adaptively improved Kalman filter on data denoising in this invention.

[0029] Figure 3 This is a comparison chart showing the effectiveness of the isolated forest model in augmented vector detection using this invention. Detailed Implementation

[0030] 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 some, not all, of the embodiments of the present invention. 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.

[0031] This invention discloses a method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion, referring to... Figure 1 The process includes steps S1-S5, as detailed below: Step S1: Obtain multidimensional environmental data and perform preprocessing.

[0032] The drying chamber of the onion drying production line was divided into multiple sub-regions. Temperature sensors, humidity sensors, hot air pressure sensors, and volatile organic compound (VOC) sensors were deployed in each sub-region to acquire real-time temperature, humidity, air pressure, and sulfur gas concentration values. During the onion drying process, the acquisition frequency was set to 1Hz. A linear normalization method was used to process the data in each dimension, mapping them to the range [0,1]. Temperature sensors in different sub-regions were categorized as the same type of sensor; similarly, humidity sensors, hot air pressure sensors, and VOC sensors in different sub-regions were all categorized as the same type of sensor.

[0033] For each sub-zone in the drying chamber, The original state vector at time t is denoted as , ,in, Indicates in The original state vector collected at each moment; This represents the real-time temperature value collected by the temperature sensor in the nth sub-region of the drying chamber; This represents the real-time humidity value collected by the humidity sensor in the nth sub-region of the drying chamber; This represents the real-time air pressure value inside the drying chamber collected by the hot air pressure sensor in the nth sub-region of the drying chamber; This represents the concentration value of sulfur-containing gas collected by the volatile organic compound sensor in the nth sub-region of the drying chamber.

[0034] Step S2: Calculate the dynamic consistency index.

[0035] In the dehydrated onion drying process, the uniformity of the hot air flow field directly determines the quality of the finished product. If the differences in measurement data from different sub-regions of the drying chamber increase abnormally, it usually means that the internal flow field has become turbulent or that excessive material accumulation has led to increased air resistance. In this case, the reliability of single-point sensor data decreases, and a dynamic consistency index needs to be calculated. For example, when the real-time temperature difference between different sub-regions is large, it means that the internal flow field has become turbulent or that excessive material accumulation has led to increased air resistance.

[0036] First, acquire the data from K sensors of the same type inside the drying chamber. The spatial dispersion index is calculated using the normalized observations at time t, expressed as:

[0037] in, Indicates in Spatial dispersion index of data from the same type of sensor at any given time; This represents the total number of sensors of the same type in the drying chamber, and its value is the same as the number of sub-regions, n. Indicates the first among sensors of the same type The weighting coefficient of each sensor, in this embodiment, is [value]. ; Indicates the first in the same type One sensor in Normalized observations at time; Indicates that the same type of sensor is in The normalized mean at any given time. It can be understood that the temperature sensor, humidity sensor, hot air pressure sensor, and volatile organic compound sensor within the drying chamber each correspond to a spatial dispersion index.

[0038] Subsequently, for the drying chamber, in terms of time... and For each node, a continuous state vector is extracted, and the gradient over time is calculated. The expression is as follows:

[0039] in, This represents the state change vector between the current time and the previous sampling time. express The original state vector collected at each moment; express The original state vector collected at each moment.

[0040] Finally, the dynamic consistency index is calculated based on the spatial dispersion index and the state change vector, and its expression is:

[0041] in, express Dynamic consistency indicators of the indoor environment during continuous drying; Indicates in The mean of the spatial dispersion index of all types of sensor data at any given time; The L2 norm represents the vector of state changes, characterizing the overall fluctuation intensity of the environmental state in the drying room over time. This represents a tiny positive number to prevent the denominator from being zero; in this embodiment, it is set to 0.0001.

[0042] When spatial discreteness An increase in the value reflects uneven temperature and humidity distribution in different sub-regions of the drying chamber, leading to an increase in negative values ​​within the index term, thus affecting the dynamic consistency index. Decrease; when the L2 norm of the state change vector decreases. When the denominator increases, it reflects that the corresponding sub-region of the drying chamber is in a dynamic adjustment period of heating or dehumidification. At this time, the larger spatial difference is within the allowable range of the process, and the increase in the denominator makes... The decline was suppressed, maintaining a high level of dynamic consistency. In summary, The spatiotemporal reliability of the perceived data was quantified; the larger the value, the higher the confidence level of the multidimensional environmental data.

[0043] Step S3: Adaptively improve the Kalman filter algorithm.

[0044] This step introduces environmental entropy change to achieve dynamic and coordinated adjustment of Kalman filter parameters. First, the environmental entropy change index is calculated, expressed as:

[0045] in, express The environmental entropy change index at any given time; express Dynamic consistency indicators of the indoor environment during continuous drying; This represents the smallest offset, with a value of [value]. The environmental entropy change index reflects the degree of uncertainty of the sensor system. When the consistency index deviates from the steady state, the environmental entropy change index increases.

[0046] Calculate the adaptive adjustment coefficient :

[0047] in, express The adaptive adjustment coefficient at any given time; This represents the sensitivity factor, which typically ranges from [5, 15]. In this embodiment, it is set to 10 and is used to control the gain intensity of filter parameter adjustment due to environmental changes. express The environmental entropy change index at any given time; The baseline entropy representing the stable drying period is taken as 0.35 in this embodiment.

[0048] The process noise covariance and measurement noise covariance in the Kalman filter algorithm are corrected using adaptive adjustment coefficients. The method is as follows:

[0049]

[0050] in, This represents the process noise covariance matrix at time t after correction. This represents the preset basic process noise covariance constant matrix; This represents the measurement noise covariance matrix at time t after correction; This represents the preset basic measurement noise covariance constant matrix; express The adaptive adjustment coefficient at any given time. For example, ; .

[0051] When drastic environmental fluctuations lead to changes in environmental entropy indicators Exceed At this time, the value of the adaptive adjustment coefficient increases. Increase A decrease in the speed of the filter indicates that the system places greater trust in real-time sensor measurements, enabling the filter to quickly track sudden environmental changes; conversely, when the drying environment becomes stable, a decrease in the speed of the filter indicates a greater trust in real-time sensor measurements. Reduced, leading to Decrease The increase physically reflects the system's greater trust in the mathematical model's predictions, improving the smoothness of the output state. An adaptively improved Kalman filter is used to filter and denoise the multidimensional environmental data, further obtaining the denoised original state vector, which is denoised as the denoised state vector.

[0052] Step S4: Construct local sensitivity enhancement feature values.

[0053] Using the current time t as the endpoint, stable data from the first 1000 sampling points of the current batch are extracted along the historical direction and a sliding window is constructed. The mean of each state dimension within the sliding window is calculated as the historical benchmark. And calculate the running deviation vector. :

[0054] in, This represents the deviation vector of the current time t relative to the historical baseline. This represents the denoised state vector at time t after Kalman filtering. This represents the baseline state vector composed of the mean values ​​of data in each dimension within the sliding window. In other words, the mean values ​​of data in each dimension within the sliding window are calculated, and multiple mean values ​​constitute the baseline state vector.

[0055] The expression for calculating the local sensitivity enhancement feature value is as follows:

[0056] in, The value representing the local sensitivity enhancement eigenvalue at time t; The Euclidean norm of the deviation vector of the current operating condition from the historical benchmark at time t quantifies the total intensity of the deviation from the benchmark. express Dynamic consistency indicators of the indoor environment during continuous drying; This represents a tiny constant with a value of 0.01.

[0057] This step, by constructing local sensitivity enhancement eigenvalues, achieves nonlinear amplification of weak abnormal signals in the early stages of onion drying. Specifically, during the stable operation phase of the drying process, the hot air flow field and temperature and humidity distribution within the drying chamber exhibit a high degree of physical consistency, at which point the dynamic consistency index approaches 1. Under these conditions, the value of the local sensitivity enhancement eigenvalue is primarily influenced by the Euclidean norm of the operational deviation vector, which characterizes the degree to which the current state deviates from the normal baseline. At a low level, the output local sensitivity enhancement feature value is low and remains stable, reflecting that the production environment is near a controlled steady-state baseline.

[0058] When onions experience uneven heating or early scorching, the physicochemical reactions on the material surface cause deviations in the local temperature and humidity fields. These deviations are often masked by global background noise in the early stages. However, such subtle anomalies inevitably lead to a decrease in the sensitivity of sensor network perception consistency, resulting in a reduction in dynamic consistency indicators. As the formula definition shows, due to... Located in the denominator, the decrease in its value amplifies the Euclidean norm of the deviation vector, further increasing the local sensitivity enhancement eigenvalue.

[0059] Step S5: Anomaly detection based on augmented vectors.

[0060] The denoised state vector With local sensitivity enhancement feature value Concatenate to construct augmenting vectors An isolation forest model is constructed and trained using augmented vectors from historical time points to obtain a trained isolation forest model. The augmented vector at the current time point is then obtained and input into the trained isolation forest model to obtain anomaly scores.

[0061] If abnormal scores If this state persists for three sampling cycles, an alarm is triggered. Upon alarm, the logic control unit adjusts the inverter frequency of the hot air circulating fan, reducing the fan speed or lowering the power compensation value of the electric heater, thereby preventing the onions from burning. In this embodiment... The value is 0.7.

[0062] The effects of this invention can also be shown in conjunction with the accompanying drawings. Figure 2To demonstrate the effectiveness of the adaptively improved Kalman filter in denoising data, this figure visually illustrates its ability to suppress environmental noise. The original data contained significant electronic noise and environmental interference, resulting in drastic fluctuations. While the standard Kalman filter smooths the signal, it is relatively slow to react to abrupt changes in environmental conditions (such as controlled temperature regulation). In contrast, the adaptively improved Kalman filter of this invention dynamically adjusts the covariance matrix based on environmental entropy changes, achieving more thorough filtering during stable periods and faster tracking of changes in real physical quantities during periods of fluctuation.

[0063] Figure 3 To compare the performance of state vector detection and augmented vector detection using the isolated forest model, existing technologies detect anomalies by using the original state vector. However, the improvement in anomaly scores is not significant when local fluctuations occur, posing a risk of false negatives. This invention introduces a dynamic consistency index as the denominator in the augmented vector calculation, non-linearly amplifying the anomaly signal. Within the anomaly range, the score rapidly exceeds the threshold and triggers a red alarm zone. This demonstrates that this method can issue timely warnings in the early stages of slight scorching or uneven heating of onions, protecting material quality by adjusting the inverter and heater power.

[0064] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion, characterized in that, The steps include: acquiring multidimensional environmental data for each sub-region within the drying chamber and performing normalization preprocessing to obtain the original state vector at the current moment; calculating the spatial dispersion index of the same type of sensors based on the normalized observations of multiple sensors at the current moment, and calculating the state change vector based on the original state vectors at the current moment and the previous moment; determining the dynamic consistency index at the current moment by using the spatial dispersion index and the L2 norm of the state change vector; the dynamic consistency index is negatively correlated with the mean of the spatial dispersion index, and positively correlated with the L2 norm of the state change vector. The environmental entropy change index is calculated, and the adaptive adjustment coefficient is calculated based on the environmental entropy change index. The parameters of the Kalman filter are corrected using the adaptive adjustment coefficient to filter and denoise the original state vector, resulting in a denoised state vector. The running deviation vector is determined based on the degree of deviation of the denoised state vector from the historical benchmark data, and the local sensitivity enhancement feature value is calculated in combination with the dynamic consistency index. The denoised state vector is concatenated with the local sensitivity enhancement feature value to construct an augmented vector. The augmented vector is then input into the anomaly detection model to obtain an anomaly score. If the anomaly score exceeds a preset threshold and continues for a preset period, an alarm is triggered.

2. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 1, characterized in that, The method for obtaining multidimensional environmental data for each sub-area in the drying chamber includes: collecting real-time temperature, real-time humidity, real-time air pressure and sulfur-containing gas concentration values ​​by deploying temperature sensors, humidity sensors, hot air pressure sensors and volatile organic compound sensors in the sub-areas of the drying chamber, and constructing an original state vector.

3. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 1, characterized in that, The steps for calculating the spatial dispersion index are as follows: multiply the weight coefficients of the same type of sensors in different sub-regions by the square of the difference between the corresponding normalized observation value and the normalized mean of the same type of sensors, and sum the product results of all sensors to obtain the spatial dispersion index.

4. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 1, characterized in that, The step of determining the dynamic consistency index at the current moment is as follows: calculate the ratio of the mean of the spatial dispersion index to the L2 norm of the state change vector, take the negative value of the ratio and use it as the exponent of the natural constant to perform a power operation to obtain the dynamic consistency index.

5. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 1, characterized in that, The steps for calculating the environmental entropy change index are as follows: take the natural logarithm of the sum of the dynamic consistency index and the minimum offset, multiply the logarithmic result by the dynamic consistency index and take the negative value to obtain the environmental entropy change index.

6. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 5, characterized in that, The step of generating the adaptive adjustment coefficient based on the environmental entropy change index is as follows: calculate the difference between the environmental entropy change index and the benchmark entropy, multiply the difference by a preset sensitivity factor, take the negative value as the exponent of the natural logarithm base, and perform a power operation. Add the power operation result to a preset constant and take the reciprocal to obtain the adaptive adjustment coefficient.

7. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 6, characterized in that, The method for correcting the parameters of the Kalman filter using the adaptive adjustment coefficient is as follows: multiply the preset basic process noise covariance constant matrix with the adaptive adjustment coefficient to obtain the corrected process noise covariance matrix. Calculate the difference between the preset constant and the adaptive adjustment coefficient; The difference is multiplied by the preset baseline measurement noise covariance matrix to obtain the corrected measurement noise covariance matrix.

8. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 1, characterized in that, The steps for calculating the local sensitivity enhancement eigenvalue are as follows: The denoising state vector is subtracted from the baseline state vector to obtain the running deviation vector; the Euclidean norm of the running deviation vector is calculated and divided by the dynamic consistency index to obtain the local sensitivity enhancement feature value.

9. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 1, characterized in that, The monitoring method also includes: after the alarm is triggered, the logic control unit adjusts the frequency of the inverter of the hot air circulation fan to reduce the wind speed or lower the power compensation value of the electric heater.

10. The method for real-time monitoring of the drying environment of dehydrated onions based on multi-sensor fusion according to claim 1, characterized in that, The state change vector is obtained by subtracting the original state vector from the previous state vector at the current time.

Images