A method and system for online monitoring of kitchen fume in a hotel

Abstract

This invention discloses a method and system for online monitoring of cooking fumes in hotels and restaurants, relating to the field of cooking fume monitoring technology. The method includes: deploying a gas-sensitive element array in the cooking fume exhaust duct to collect the response voltage signal sequence corresponding to the cooking fumes, and simultaneously collecting the airflow velocity inside the duct; extracting the voltage transition edge characteristics and their arrival times, generating a first time-series feature vector from the time interval, and obtaining a second time-series feature vector after airflow velocity correction; calculating the dynamic time curvature distance between this vector and the standard time-series vectors of various cooking fume components to identify the types of volatile organic compounds in the cooking fumes, and then matching the corresponding response intensity coefficient to correct the voltage signal amplitude; calculating the short-term increase in cooking fume concentration based on the multi-cycle peak change of the corrected amplitude, and issuing an abnormal cooking fume alarm signal when the increase exceeds a dynamic threshold.

Description

Technical Field

[0001] This invention relates to the field of oil fume monitoring technology, specifically to a method and system for online monitoring of oil fumes in hotels and restaurants. Background Technology

[0002] Hotel and restaurant cooking fume emissions are characterized by intermittent occurrences, complex composition, and dramatic concentration fluctuations, making online monitoring of emissions increasingly urgent. Existing online fume monitoring methods generally employ a single gas sensor to collect response signals, directly converting the signal amplitude into concentration values ​​for threshold comparison. However, this method faces two major technical challenges. First, due to differences in molecular structure and activity, different volatile organic compounds (VOCs) produce significantly different response intensities on the same gas sensor. The specific types of VOCs in the fume cannot be identified solely from the response signal amplitude, leading to severe distortion in concentration estimation under complex, multi-component fume scenarios. Second, the airflow velocity inside the exhaust duct frequently changes due to fan speed adjustments and duct resistance. This transient velocity fluctuation compresses or stretches the time axis of the gas sensor's response signal, distorting the temporal characteristics extracted based on a fixed time reference. This results in the incorrect matching of the same type of fume to different components at different flow rates. While devices based on spectroscopy or mass spectrometry can identify components, they are unsuitable for the high-temperature, high-humidity, and oil-covered environments of hotel exhaust ducts, and are also complex and costly. Current technologies for determining excessive oil fume concentrations largely rely on fixed alarm thresholds, which cannot adapt to the normal fluctuations in emission patterns between dining and non-dining periods in hotels. Even small increases in concentration during non-dining periods may indicate abnormal leaks or equipment malfunctions, and fixed thresholds may be slow to react or produce false alarms. Therefore, accurately identifying the types of volatile organic compounds (VOCs) in dynamically changing exhaust environments and correcting concentration estimates accordingly, as well as setting adaptive alarm mechanisms based on time-varying emission characteristics, have become critical issues that urgently need to be addressed. Summary of the Invention

[0003] This invention provides a method and system for online monitoring of cooking fumes in hotels and restaurants, aiming to solve the technical problems of existing methods being unable to identify the types of volatile organic compounds and perform concentration calibration in exhaust environments with fluctuating airflow speeds, and being unable to adaptively adjust alarm thresholds, leading to inaccurate identification of abnormal states.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for online monitoring of cooking fumes in hotel restaurants, comprising:

[0005] By deploying an array of gas-sensitive elements inside the exhaust duct of a hotel restaurant, the response voltage signal sequence as the fumes pass through is collected, and the internal airflow velocity value of the exhaust duct is acquired simultaneously.

[0006] Based on the rate of change of voltage amplitude at each sampling moment in the response voltage signal sequence, multiple voltage transition edge features are extracted, and the arrival time of each voltage transition edge feature is recorded.

[0007] A first time-series feature vector is obtained based on the time interval between the arrival times of the multiple voltage transition edges, and a second time-series feature vector is obtained by correcting the first time-series feature vector based on the internal airflow velocity value.

[0008] Based on the dynamic time curvature distance between the second time-series feature vector and multiple standard time-series vectors of oil fume components, the type of volatile organic compounds in the oil fume is determined, and the amplitude of the response voltage signal sequence is corrected according to the preset response intensity coefficient corresponding to the type of classification to obtain the corrected response amplitude.

[0009] Based on the periodic peak change of the correction response amplitude over multiple consecutive acquisition cycles, the short-term increment of the fume concentration is calculated, and when the short-term increment exceeds the dynamic adjustment threshold, an abnormal fume alarm signal is triggered.

[0010] Preferably, the gas-sensitive element array consists of multiple metal-oxide-semiconductor gas sensors installed at equal angular intervals on the same cross-sectional circumference of the inner wall of the exhaust duct. Under the condition of applying a constant heating voltage to each sensor, the output resistance value of each sensor is continuously collected as oil fumes pass through and converted into a voltage signal, forming the response voltage signal sequence. A hot-film flow velocity sensor is embedded at the center of the installation cross-section of the gas-sensitive element array to measure the airflow velocity in the exhaust duct in real time and obtain the internal airflow velocity value. This scheme can stably acquire response signals and accurately capture airflow changes.

[0011] Preferably, a voltage difference sequence is generated by calculating the voltage difference between adjacent sampling times in the response voltage signal sequence; sampling times with an absolute voltage difference greater than a preset transition threshold are selected as candidate transition times; within a preset time window before and after each candidate transition time, the time corresponding to the maximum value of the voltage amplitude change rate is searched as the actual arrival time of the voltage transition along the feature, and the direction of voltage amplitude change is recorded. The preset transition threshold is the product of the root mean square value of the voltage difference sequence and a preset adjustment coefficient, thereby adapting to different background noise levels.

[0012] Preferably, the arrival times of the multiple voltage transition edges are arranged chronologically to form a time sequence, and the difference between adjacent arrival times is calculated as the arrival time interval. These are then combined to obtain the first time-series feature vector. The average airflow velocity value of the exhaust pipe during the fume collection period is obtained, and the velocity ratio between this average airflow velocity value and a preset standard airflow velocity value is calculated. Each arrival time interval in the first time-series feature vector is multiplied by the velocity ratio to obtain the corrected time interval, which is then sequentially combined to form the second time-series feature vector. Flow velocity correction can eliminate the influence of airflow instability on the time-series features.

[0013] Preferably, the average airflow velocity value is the velocity value calculated by time-weighted moving average over the oil fume collection period.

[0014] Preferably, standard time-series feature vectors of various volatile organic compounds are retrieved from a preset standard database of oil fume components. The dynamic time curvature distance between the second time-series feature vector and each standard time-series feature vector is calculated, and the volatile organic compound type corresponding to the smallest curvature distance is used as the category classification. Based on this category classification, the corresponding response intensity coefficient is found in a preset response intensity coefficient mapping table. The voltage amplitude at each sampling moment in the response voltage signal sequence is divided by this response intensity coefficient to obtain the corrected response amplitude. This method can accurately identify oil fume components and compensate for the response differences between different components.

[0015] Preferably, the calculation of the dynamic time bending distance adopts a global window constraint, and the window width is a preset percentage of the length of the second temporal feature vector, so as to balance matching accuracy and computational efficiency.

[0016] Preferably, the maximum value of the correction response amplitude within each acquisition cycle is taken as the periodic peak value, resulting in a periodic peak value sequence for multiple consecutive acquisition cycles; the difference between adjacent periodic peak values ​​is calculated, and the average of multiple differences is taken as the short-term increment value of the oil fume concentration; multiple historical short-term increment values ​​within the same historical time period are obtained, their standard deviation is calculated, and a dynamic adjustment threshold is set accordingly; when the short-term increment value is greater than the dynamic adjustment threshold, an oil fume abnormality alarm signal is triggered. The dynamic threshold can adapt to normal fluctuations in different time periods.

[0017] Preferably, after triggering the abnormal oil fume alarm signal, the peak values ​​of subsequent collection cycles are collected to obtain a subsequent cycle peak sequence; the peak rise slope of the sequence within the sliding time window is calculated, and the peak rise slope is compared with a preset slope alarm threshold; when the peak rise slope is greater than the preset slope alarm threshold, an oil fume concentration continuous rise warning signal is generated and its urgency level is marked as high risk level to improve the foresight of the warning.

[0018] Preferably, after obtaining the correction response amplitude, the temperature and humidity values ​​of the oil fume inside the exhaust pipe are also obtained. The temperature compensation coefficient and humidity compensation coefficient are obtained by looking up the temperature compensation coefficient table and humidity compensation coefficient table, respectively. The correction response amplitude is multiplied by the product of the temperature compensation coefficient and the humidity compensation coefficient to obtain the temperature and humidity correction amplitude, and the correction response amplitude is updated accordingly to further improve the environmental adaptability of concentration monitoring.

[0019] Preferably, after obtaining the response voltage signal sequence, baseline drift detection is performed on each gas sensor in the gas-sensitive element array, the average baseline resistance during the smoke-free period is calculated, and the resistance offset between the real-time resistance value at the current sampling moment and the average baseline resistance is calculated; when the resistance offset exceeds a preset baseline drift threshold, the real-time resistance value is corrected to the product of the average baseline resistance and a preset resistance correction coefficient, and the response voltage signal sequence is updated based on the corrected resistance value to suppress the interference of long-term sensor drift on the signal.

[0020] Preferably, after determining the type of volatile organic compounds, the response voltage signal sequence is further classified and labeled according to the type, the concentration contribution value of the volatile organic compound is calculated for each category, and the concentration contribution values ​​are sorted from largest to smallest to output an oil fume component analysis report, which facilitates a direct understanding of the composition of pollutants.

[0021] An online monitoring system for kitchen fumes in hotels and restaurants includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the monitoring method described above.

[0022] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0023] To address the distortion of the timing characteristics of the response signal caused by dynamic changes in airflow velocity, this paper proposes a method to extract the arrival times of voltage transition edges from the response voltage signal sequence. This extract constructs a first timing feature vector reflecting the time intervals between adjacent transitions. Simultaneously acquired internal airflow velocity values ​​are then introduced to generate velocity ratios. These velocity ratios are used to perform multiplicative corrections on each time interval in the first timing feature vector, resulting in a second timing feature vector equivalent to that observed under standard flow velocity conditions. This time-domain scaling correction method eliminates the nonlinear compression and stretching effects on the signal time axis caused by sudden changes in fan speed or pipeline pressure fluctuations. This ensures that subsequent matching with the standard timing vector is no longer affected by real-time flow velocity interference, effectively guaranteeing the accuracy and repeatability of volatile organic compound (VOC) identification under variable flow velocity conditions.

[0024] To address the issue that a single gas sensor cannot distinguish between different types of volatile organic compounds (VOCs) and that direct amplitude conversion leads to concentration calculation errors, this method calculates the dynamic time curvature distance between the second time-series feature vector (corrected for flow velocity) and the standard time-series vectors of various VOCs. The VOC type corresponding to the smallest curvature distance is selected as the primary component of the current cooking fume. Based on this selection, a preset response intensity coefficient is retrieved, and amplitude correction is applied to each sampling point of the response voltage signal sequence. This method utilizes the characteristic adsorption and desorption dynamic differences of different VOCs on the surface of the gas-sensitive element array, using the shape similarity of the time series as the core criterion for component identification. This overcomes the limitation of traditional methods that rely solely on signal amplitude for discrimination, fundamentally correcting the cross-sensitivity error introduced by differences in component sensitivity. This allows for a more accurate corrected response amplitude that closely approximates the true component concentration, even in situations where multiple components coexist. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0026] Figure 1 This is a flowchart of the online monitoring method for kitchen fumes in hotels and restaurants;

[0027] Figure 2 This is a flowchart of signal processing and airflow correction feature vector extraction for gas-sensitive sensor arrays in hotel and restaurant exhaust ducts;

[0028] Figure 3 This is a flowchart for analyzing the components of cooking fumes and calculating their concentration contribution. Detailed Implementation

[0029] 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 some embodiments of the present invention, not all embodiments. 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.

[0030] See Figure 1This invention provides a method and system for online monitoring of cooking fumes in hotels and restaurants. The method uses an array of gas-sensitive elements arranged within the exhaust duct of the hotel or restaurant to collect response voltage signal sequences as cooking fumes pass through, and simultaneously acquires the internal airflow velocity value of the exhaust duct. Based on the rate of change of voltage amplitude at each sampling moment in the response voltage signal sequence, multiple voltage transition edge features are extracted, and the arrival time of each voltage transition edge feature is recorded. Based on the time interval between the arrival times of the multiple voltage transition edge features, a first time-series feature vector is obtained, and the first time-series feature vector is corrected for flow velocity based on the internal airflow velocity value to obtain a second time-series feature vector. Based on the dynamic time curvature distance between the second time-series feature vector and multiple standard time-series vectors of cooking fume components, the type of volatile organic compounds in the cooking fume is determined, and the amplitude of the response voltage signal sequence is corrected according to the preset response intensity coefficient corresponding to the type classification to obtain a corrected response amplitude. Based on the periodic peak change of the corrected response amplitude over multiple consecutive acquisition cycles, the short-term increment value of the cooking fume concentration is calculated, and when the short-term increment value exceeds a dynamic adjustment threshold, an abnormal cooking fume alarm signal is triggered. The processor in this system implements the steps of the above method when executing a computer program.

[0031] Example 1:

[0032] In specific implementation, please refer to Figure 2The same cross-section of the hotel restaurant's exhaust duct is divided into equally spaced sections. Multiple metal-oxide-semiconductor (MOS) gas sensors are installed at these sections, forming a gas-sensitive element array. The number of MOS gas sensors is determined by the inner diameter of the exhaust duct; for example, eight MOS gas sensors are installed in a circular duct with a nominal diameter of 300 mm, with a central angle of 45 degrees between adjacent sensors. A constant heating voltage of 5.0 volts is applied to the heating electrode of each MOS gas sensor in the array, stabilizing the temperature of the sensitive layer of each sensor within the range of 300 to 400 degrees Celsius. The output resistance value of each MOS gas sensor is continuously collected at a preset sampling frequency of 100 Hz as the exhaust fumes flow through it. The output resistance value collected by each metal-oxide-semiconductor (MOS) gas sensor is converted into a voltage signal through a voltage divider circuit. The voltage divider circuit has a resistance of 10,000 ohms, and the output voltage signal ranges from 0 volts to 5 volts, thus obtaining the response voltage signal sequence of each MOS gas sensor. A hot-film flow velocity sensor is embedded at the center of the cross-section of the gas-sensitive element array, with its sensing probe facing the airflow direction. The hot-film flow velocity sensor incorporates a heating resistor and a temperature-sensing resistor. The heating resistor is maintained at a constant temperature 50 degrees Celsius higher than the incoming flow temperature under the drive of a constant temperature control circuit. As the airflow passes over the heating resistor, it carries away heat. The temperature-sensing resistor senses the temperature change, and the constant temperature control circuit automatically adjusts the heating power according to the temperature change. The internal airflow velocity value is calculated and output in real time based on the calibrated relationship between the heating power and the airflow velocity.

[0033] For each channel of the response voltage signal sequence, calculate the voltage amplitude difference between adjacent sampling times. Let the voltage amplitude corresponding to the k-th sampling time be V(k), and the voltage amplitude corresponding to the (k+1)-th sampling time be V(k+1). Then, the voltage amplitude difference ΔV(k) = V(k+1) - V(k). All ΔV(k) values ​​are arranged in the order of sampling times to form a voltage difference sequence. A preset transition threshold is set to distinguish between noise fluctuations and actual voltage jumps caused by the adsorption of oil fume components. The preset transition threshold Th is calculated according to the following formula:

[0034]

[0035] in, This represents the root mean square value of the voltage difference sequence. The calculation process involves squaring each voltage amplitude difference in the voltage difference sequence, averaging the results, and then taking the square root. This indicates the preset adjustment coefficient. The value is set based on the floor noise amplitude of the metal-oxide-semiconductor gas sensor. In application scenarios where the root mean square value of the floor noise of the metal-oxide-semiconductor gas sensor installed in the exhaust pipe is approximately 0.02 volts, Set to 1.5. In the voltage difference sequence, filter all sampling times k that satisfy the condition that the absolute value of ΔV(k) is greater than the preset transition threshold Th, and mark the selected sampling times as candidate transition times.

[0036] For each candidate transition moment, a preset time window of 50 milliseconds is taken before and after it. Within each preset time window, the rate of change of voltage amplitude corresponding to each sampling moment is calculated. The rate of change of voltage amplitude is determined by the ratio of the voltage amplitude difference between the two points before and after the sampling moment to the time step. The maximum absolute value of the rate of change of voltage amplitude within the preset time window is searched, and the sampling moment corresponding to the maximum absolute value of the rate of change of voltage amplitude is taken as the precise actual arrival moment of the voltage transition edge feature. At the same time, the direction of voltage amplitude change at this actual arrival moment is recorded. When the voltage amplitude difference corresponding to the actual arrival moment is greater than zero, the direction of voltage amplitude change is recorded as the upward direction; when the voltage amplitude difference corresponding to the actual arrival moment is less than zero, the direction of voltage amplitude change is recorded as the downward direction.

[0037] The actual arrival times of all voltage transition edges extracted during the acquisition process are arranged in chronological order to form a time sequence. The time difference between any two adjacent actual arrival times in the time sequence is calculated to obtain multiple arrival time intervals. These multiple arrival time intervals are arranged in the order of their generation and combined to form the first time-series feature vector. The average airflow velocity value of the exhaust pipe during the fume collection period is obtained. The average airflow velocity value is calculated as follows: during the fume collection period, the internal airflow velocity value sequence output by the hot-film flow sensor is processed by a time-weighted moving average. The window width of the time-weighted moving average is 10 seconds, with each second as a time step. The weighting function adopts a linearly decreasing form, with higher weights assigned to internal airflow velocity values ​​closer to the current time. The average airflow velocity value is divided by a preset standard airflow velocity value, which is set to 5.0 m / s, to obtain the velocity ratio. Each arrival time interval contained in the first time-series feature vector is multiplied by the velocity ratio to obtain multiple corrected time intervals. These multiple corrected time intervals are combined in the original chronological order to form the second time-series feature vector.

[0038] Example 2:

[0039] In practice, the construction of the pre-set oil fume component standard database is completed offline. For each target volatile organic compound (VOC), in a standard experimental environment, using the same type of metal oxide semiconductor gas sensor array as used in online monitoring, under conditions of 25 degrees Celsius, 50% relative humidity, and a standard airflow velocity of 5.0 m / s, a single-type VOC sample of known concentration is introduced, and the response voltage signal sequence is collected. Following the same processing method as online monitoring, the arrival time of the voltage transition edge characteristics is extracted, the arrival time interval is calculated, and combined into a standard time series feature vector. The name of each VOC type is associated with its corresponding standard time series feature vector and stored to form the pre-set oil fume component standard database. During online monitoring, the standard time series feature vectors corresponding to various VOCs such as ethyl acetate, ethanol, propanol, butanol, formaldehyde, acetaldehyde, benzene, toluene, xylene, limonene, and α-pinene are retrieved from the pre-set oil fume component standard database to construct multiple oil fume component standard time series vectors.

[0040] The second time-series feature vector is denoted as the query sequence Q. The query sequence Q is composed of p corrected time intervals arranged sequentially, and the form of the query sequence Q is Q= , ,..., The m-th standard time series feature vector among multiple standard time series vectors of oil fume components is denoted as the reference sequence. Reference sequence The reference sequence consists of r standard time intervals arranged in order. The form is = , ,..., The dynamic time-bending distance is calculated using a recursive dynamic programming method, constructing a p-row, r-column cumulative distance matrix D. The element in the a-th row and b-th column of the cumulative distance matrix D... The calculation is performed according to the following iterative formula:

[0041]

[0042] in, This represents the a-th corrected time interval in the query sequence Q; Represents the reference sequence The b-th standard time interval; | - | indicates and The absolute value of the difference is used as the local distance; This represents the value of the element in the (a-1)th row and (b-1)th column of the cumulative distance matrix D; This represents the value of the element in the (a-1)th row and bth column of the cumulative distance matrix D; This represents the value of the element in the (a)th row and (b-1)th column of the cumulative distance matrix D; `min{}` indicates that the minimum value among the three values ​​within the curly braces is selected as the operand. The cumulative distance matrix D is initialized by setting the value of the element in the 0th row and 0th column... Setting it to 0 will change the value of the element in the b-th column of the 0th row. Set to infinity, and set the value of the element in row a, column 0 to infinity. Set to infinity. Iterative calculations begin with a=1 and b=1, and proceed sequentially up to a=p and b=r, ultimately obtaining the element value in the p-th row and r-th column of the cumulative distance matrix D. ,Will As the query sequence Q and the reference sequence The dynamic time bending distance value.

[0043] The calculation path for dynamic time bending distance introduces a global window constraint. The width of the global window is set as a preset percentage of the length p of the query sequence Q, with the preset percentage set to 10%. This global window constraint means that during the iterative calculation of the cumulative distance matrix D, for the a-th row, the iteration range of column index b is limited to the interval [a-\lfloorp\times0.1\rfloor,a+\lfloorp\times0.1\rfloor]. Elements exceeding [1,r] are automatically truncated. If the lower limit of the interval is less than 1, it is set to 1; if the upper limit is greater than r, it is set to r. Matrix elements outside the interval do not participate in the iterative calculation. For each standard time series feature vector in the multiple standard time series vectors of oil fume components, the above dynamic time bending distance calculation process is performed to obtain multiple bending distance values, each corresponding to a type of volatile organic compound.

[0044] Multiple bending distance values ​​are compared numerically, and the minimum bending distance value is selected. The type of volatile organic compound corresponding to the minimum bending distance value is determined as the type of volatile organic compound in the current oil fume.

[0045] The construction of the preset response intensity coefficient mapping table was completed offline. For each target volatile organic compound (VOC), a single-type VOC sample with a constant concentration of 50 ppm was introduced into a standard experimental environment. The steady-state output voltage amplitude of the metal oxide semiconductor gas sensor array during the saturation response phase was recorded. The ratio of the steady-state output voltage amplitude corresponding to each VOC to the steady-state output voltage amplitude of the reference VOC was used as the preset response intensity coefficient. Ethanol was selected as the reference VOC. The type name of each VOC and its corresponding preset response intensity coefficient were associated and stored to form the preset response intensity coefficient mapping table. During online monitoring, based on the determined type of VOC, the corresponding preset response intensity coefficient was retrieved from the preset response intensity coefficient mapping table. The voltage amplitude at each sampling moment in the response voltage signal sequence was divided by the retrieved preset response intensity coefficient, and the quotient was the corrected response amplitude at each sampling moment.

[0046] Example 3:

[0047] In practice, the acquisition period is set to 60 seconds. Within one acquisition period, the correction response amplitude is continuously recorded at a sampling frequency of 100 Hz, resulting in 6000 correction response amplitude data points. For each acquisition period, all correction response amplitude data points are iterated through, and the maximum value is found through numerical comparison. This maximum value is taken as the period peak value for that acquisition period. The period peak values ​​of N consecutive acquisition periods are arranged in chronological order to obtain a period peak value sequence. N is set to 5, meaning that 5 consecutive acquisition periods form a period peak value sequence, which is represented as... , , , , .

[0048] The difference between the peak values ​​of two adjacent acquisition cycles in the periodic peak sequence is calculated to obtain the four periodic peak value changes. The calculation method for the periodic peak value changes is as follows:

[0049]

[0050] Where u takes the values ​​1, 2, 3, 4; This represents the periodic peak value in the (u+1)th acquisition cycle of the periodic peak sequence; This represents the peak value of the u-th acquisition cycle in the periodic peak sequence; This represents the change in the peak value of the (u+1)th acquisition cycle relative to the utth acquisition cycle. The peak value changes over the four cycles are then considered. , , , Take the arithmetic mean, and use the resulting arithmetic mean as the short-term increment of the oil fume concentration.

[0051] The dynamic threshold setting relies on historical short-term incremental value data within the same historical time period. The historical time period is defined as follows: using the date and time of the current collection period as a baseline, all historical short-term incremental values ​​within the same weekday type and the same hourly time period within the past 30 days are retrieved. The weekday type is divided into weekdays and rest days, and the hourly time period uses the hour as the statistical unit. K historical short-term incremental values ​​recorded within the historical time period are retrieved from the storage device. K is a natural number and changes dynamically with the accumulation of historical data; K is not less than 20 when 30 days of historical data have been accumulated. The standard deviation σ of the K historical short-term incremental values ​​is calculated. The formula for the standard deviation is the sum of the squares of the deviations obtained by subtracting the arithmetic mean of the K historical short-term incremental values ​​from each of the K historical short-term incremental values, divided by (K-1), and then taking the square root. The dynamic adjustment threshold is set to 3 times the standard deviation σ, i.e., dynamic adjustment threshold = 3σ. The basis for setting 3 times is that, under the assumption of stable fluctuation of oil fume emission concentration, the distribution of short-term incremental values ​​approximately follows a normal distribution. Events greater than 3 times the standard deviation are statistically low-probability abnormal events, and their probability of natural occurrence is less than 0.3%.

[0052] The calculated short-term increment value is compared with the dynamic adjustment threshold. When the short-term increment value exceeds the dynamic adjustment threshold, an abnormal increase in oil fume concentration is detected, and an abnormal oil fume alarm signal is immediately triggered. The abnormal oil fume alarm signal includes the timestamp of the trigger time, the value of the current short-term increment value, the value of the dynamic adjustment threshold, and a time-series curve thumbnail of the current periodic peak sequence.

[0053] After the abnormal oil fume alarm signal is triggered, starting from the moment the alarm signal is triggered, the peak values ​​of the subsequent M sampling cycles are continuously collected at 60-second intervals, with M set to 10. This results in a sequence of peak values ​​for the subsequent cycles, represented as follows: , ,..., The sliding time window is set to a window length L of 3 acquisition cycles, and the sliding step size is 1 acquisition cycle. Starting from the first acquisition cycle of the subsequent periodic peak sequence, the sliding time window is slid forward sequentially, obtaining a periodic peak subsequence within the window with each slide, for a total of 8 periodic peak subsequences within the window. For each periodic peak subsequence within the window, the peak rise slope is calculated using the least squares linear fitting method. The independent variables of the least squares linear fitting are the acquisition cycle numbers 1, 2, and 3 within the window, and the dependent variables are the three periodic peaks of the periodic peak subsequence within the window. The least squares estimate of the peak rise slope is determined by the ratio of the covariance to the variance of the independent variables.

[0054] The preset slope alarm threshold is set based on the calibration results of the response slope of the metal oxide semiconductor gas sensor in a standard gas of volatile organic compounds with a known concentration gradient. During the calibration process, the rising slope of the periodic peak value is recorded as the concentration linearly increases from 10 ppm to 100 ppm. The 50th percentile value of the calibrated slope is taken as the preset slope alarm threshold, which is 0.85 volts per acquisition cycle. The peak rising slope calculated for each window is compared with the preset slope alarm threshold. When the peak rising slope of any window is greater than the preset slope alarm threshold, a warning signal for continuously rising oil fume concentration is generated, and the urgency level of the warning signal is marked as high risk. The high risk level marking will trigger the linkage activation of the audible and visual alarm devices in the alarm system and push the high risk level information to the hotel and catering management terminal equipment.

[0055] Example 4:

[0056] In practice, after obtaining the correction response amplitude, a thermocouple temperature sensor and a capacitive humidity sensor are embedded 100 mm upstream of the gas-sensitive element array mounting section inside the exhaust duct. The thermocouple temperature sensor has a measurement range of -20°C to 300°C with an accuracy of ±0.5°C, while the capacitive humidity sensor has a measurement range of 0%RH to 100%RH with an accuracy of ±2%RH. The thermocouple temperature sensor and the capacitive humidity sensor synchronously collect the oil fume temperature and humidity values ​​inside the exhaust duct at a sampling frequency of once per second.

[0057] The construction of the temperature compensation coefficient table was completed during the offline calibration phase. In a standard experimental environment, a metal-oxide-semiconductor (MOS) gas sensor array was placed in a temperature-controlled test chamber, and a standard ethanol gas with a constant concentration of 50 ppm was introduced. With the oil fume humidity value fixed at 50% RH, the temperature of the test chamber was adjusted from -10°C to 80°C in 10-degree increments, and the steady-state voltage amplitude output by the MOS gas sensor array at each temperature node was recorded. The steady-state voltage amplitude at the standard temperature of 25°C was used as the reference temperature response value. The steady-state voltage amplitude at each temperature node was divided by the reference temperature response value to obtain its reciprocal, which is the temperature compensation coefficient corresponding to that temperature node. The temperature compensation coefficient table uses temperature values ​​as index keys and temperature compensation coefficients as storage values, employing linear interpolation to cover all integer temperature values ​​within the range of -10°C to 80°C.

[0058] The construction of the humidity compensation coefficient table was also completed during the offline calibration phase. In a standard experimental environment, the metal oxide semiconductor gas sensor array was placed in a constant temperature and humidity test chamber, and a standard ethanol gas with a constant concentration of 50 ppm was introduced. With the oil fume temperature fixed at 25 degrees Celsius, the relative humidity of the test chamber was adjusted from 10%RH to 90%RH in increments of 10%RH. The steady-state voltage amplitude output by the metal oxide semiconductor gas sensor array at each humidity node was recorded. The steady-state voltage amplitude at the standard relative humidity of 50%RH was used as the reference humidity response value. The steady-state voltage amplitude at each humidity node was divided by the reference humidity response value to obtain its reciprocal, which is the humidity compensation coefficient corresponding to that humidity node. The humidity compensation coefficient table uses humidity values ​​as index keys and humidity compensation coefficients as storage values, employing linear interpolation to cover all integer humidity values ​​within the range of 10%RH to 90%RH.

[0059] During online monitoring, the real-time oil fume temperature value collected by the thermocouple temperature sensor and the real-time oil fume humidity value collected by the capacitive humidity sensor are used as query conditions to retrieve the temperature compensation coefficient table and humidity compensation coefficient table respectively, and obtain the temperature compensation coefficient. Humidity compensation coefficient When the queried oil fume temperature value is exactly equal to an integer temperature index key in the temperature compensation coefficient table, the corresponding temperature compensation coefficient is read directly. When the queried oil fume temperature value falls between two integer temperature index keys, linear interpolation is performed based on the temperature compensation coefficients corresponding to the two index keys to obtain the temperature compensation coefficient. Humidity compensation coefficient Acquisition method and temperature compensation coefficient The acquisition method is the same.

[0060] The calculation of temperature and humidity correction for the correction response amplitude is performed according to the following formula:

[0061]

[0062] in, This indicates the amplitude of the correction response before temperature and humidity correction; This indicates the temperature compensation coefficient obtained from the temperature compensation coefficient table based on the real-time oil fume temperature value; This indicates the humidity compensation coefficient obtained by querying the humidity compensation coefficient table based on the real-time oil fume humidity value; This represents the temperature and humidity correction amplitude obtained after temperature and humidity correction. (The temperature and humidity correction amplitude is then displayed.) The updated correction response amplitude is used to cover the original correction response amplitude and is used for subsequent periodic peak extraction and short-term incremental value calculation.

[0063] After obtaining the response voltage signal sequence and before extracting the voltage transition edge features, baseline drift detection is performed on the output resistance value of each metal-oxide-semiconductor (MOS) gas sensor in the gas-sensitive element array. During the initial operation phase after the monitoring system is started, a 120-second smoke-free period is set. During this smoke-free period, only clean air purified by high-efficiency smoke extraction flows through the exhaust duct. The arithmetic mean of the output resistance value sequence collected by each MOS gas sensor during this smoke-free period is calculated to obtain the baseline resistance average value of each MOS gas sensor. The baseline resistance average value of each MOS gas sensor is then stored in non-volatile memory.

[0064] During online monitoring, at each sampling moment, the real-time resistance value of each metal-oxide-semiconductor (MOS) gas sensor is acquired, and the resistance offset between the real-time resistance value and the average baseline resistance of the corresponding MOS gas sensor is calculated. The absolute value of the resistance offset reflects the degree of baseline drift that may occur in the sensitive layer of the MOS gas sensor due to long-term exposure to oil fume. The preset baseline drift threshold is set at 20% of the average baseline resistance of the corresponding MOS gas sensor. The basis for setting the preset baseline drift threshold to 20% is that in the aging characteristic test of the MOS gas sensor, when the resistance value offset of the sensitive layer exceeds 20% of the initial baseline resistance, the MOS gas sensor's response sensitivity to volatile organic compounds shows a significant nonlinear decay, requiring resistance value correction.

[0065] When the absolute value of the resistance offset exceeds the preset baseline drift threshold, the real-time resistance value is replaced with the product of the baseline resistance average and the preset resistance correction factor. The preset resistance correction factor is set to 1.05. The basis for this setting is that in accelerated aging experiments, after 1000 hours of equivalent operation, the resistance value of the metal oxide semiconductor gas sensor in clean air fluctuates by approximately 5% compared to the initial baseline resistance average. Using 1.05 as the correction factor can pull the drifted resistance value back to an equivalent level close to the initial state. The corrected real-time resistance value obtained after replacement is converted into a voltage signal through a voltage divider circuit. The voltage signal corresponding to the corrected real-time resistance value replaces the voltage amplitude at the corresponding sampling time in the original response voltage signal sequence, thus completing the update of the response voltage signal sequence. For sampling times where the absolute value of the resistance offset does not exceed the preset baseline drift threshold, the original real-time resistance value and the corresponding voltage amplitude are retained without replacement.

[0066] Example 5:

[0067] In specific implementation, please refer to Figure 3After determining the type of volatile organic compounds (VOCs) in the cooking fumes, one or more VOC types are identified in the current cooking fume sample. From a pre-defined cooking fume component standard database, the standard time-series feature vector corresponding to each identified VOC is retrieved, and the name of each identified VOC type is used as a category label. The response voltage signal sequence is unfolded according to the temporal order of the sampling times. For each sampling time in the response voltage signal sequence, the voltage transition edge feature interval to which the corrected time interval corresponding to that sampling time belongs during the construction of the second time-series feature vector is retrieved. When a voltage transition edge feature interval is determined to correspond to a certain VOC type, the response voltage signals of all sampling times contained within that voltage transition edge feature interval are labeled with the category label corresponding to that VOC type. When a sampling time in the response voltage signal sequence falls into the overlapping area of ​​multiple voltage transition edge feature intervals, and the multiple voltage transition edge feature intervals correspond to different VOC types, that sampling time is simultaneously labeled with multiple category labels. The response voltage signals of all sampling times with the same category label are extracted in time sequence to form a category label signal sequence, thereby obtaining multiple category label signal sequences that are the same number as the number of volatile organic compounds that have been determined to belong.

[0068] For each category-labeled signal sequence, calculate the concentration contribution value of the corresponding volatile organic compounds. The concentration contribution value is calculated according to the following formula:

[0069]

[0070] in, Indicates the first The types of volatile organic compounds whose classifications have been determined. The value of is any natural number from 1 to the total number of volatile organic compound species that have been identified; Indicates the first The concentration contribution value corresponding to each volatile organic compound; Indicates the first The first category marker signal sequence corresponding to a certain volatile organic compound The response voltage amplitude at each sampling time. This is the index of the sampling time within the category-labeled signal sequence. The value range is 1 to All integers between; Indicates the first The total number of sampling times contained in the category label signal sequence corresponding to a certain volatile organic compound; This indicates all of the category-labeled signal sequences. Sum the magnitudes of the response voltages; The arithmetic mean of the response voltage amplitudes of the category-labeled signal sequence; Indicates the first Concentration conversion coefficient of a certain volatile organic compound.

[0071] Concentration conversion coefficient The method for obtaining the data is as follows: During the construction of the preset response intensity coefficient mapping table, the steady-state output voltage amplitude of each target volatile organic compound is recorded in a standard experimental environment at a constant concentration of 50 ppm. Simultaneously, steady-state output voltage amplitudes are also collected at concentrations of 10 ppm, 20 ppm, 50 ppm, and 100 ppm. A linear relationship between the volatile organic compound concentration and the steady-state output voltage amplitude is established through least-squares linear fitting. The slope of the linear relationship is the concentration conversion coefficient. Concentration conversion factor The physical meaning is the change in response voltage amplitude caused by a unit change in concentration (ppm), the concentration conversion coefficient. The unit is volts per ppm. The conversion factor between the name of each volatile organic compound and its corresponding concentration is used. The associated data is stored in the extended fields of the preset response intensity coefficient mapping table, and is directly retrieved according to the type of volatile organic compound when calculating the concentration contribution value.

[0072] All identified volatile organic compounds were assigned according to their respective concentration contribution values. Sort the volatile organic compounds (VOCs) in descending order of their concentration to obtain a sorted list. Output the sorted results as a report on the analysis of oil fume components. This report includes: sorting number, name of the VOC type, and concentration contribution value. The report includes the numerical value of the concentration contribution value, its percentage contribution value relative to the total concentration contribution value of all volatile organic compounds (VOCs), and the time interval in which each VOC type appears in the response voltage signal sequence. The oil fume composition analysis report is generated in a structured data format, specifically JSON. The top-level element of the JSON format is an array, and each element in the array is an object containing six fields: "Serial Number," "Type Name," "Concentration Contribution Value," "Percentage," "Start Time," and "End Time." The oil fume composition analysis report is automatically generated and attached to the alarm information whenever an abnormal oil fume alarm signal is triggered, and sent to the hotel's catering management terminal equipment. Simultaneously, a summary oil fume composition analysis report for that natural hour is generated at the end of each natural hour and stored in the historical database.

[0073] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for online monitoring of cooking fumes in hotel and catering establishments, characterized in that, The method includes: By deploying an array of gas-sensitive elements in the exhaust ducts of hotel restaurants, the response voltage signal sequence as the fumes pass through is collected, and the internal airflow velocity value of the exhaust duct is acquired simultaneously. Based on the rate of change of voltage amplitude at each sampling moment in the response voltage signal sequence, multiple voltage transition edge features are extracted, and the arrival time of each voltage transition edge feature is recorded. A first time-series feature vector is obtained based on the time interval between the arrival times of the multiple voltage transition edges, and a second time-series feature vector is obtained by correcting the first time-series feature vector based on the internal airflow velocity value. Based on the dynamic time curvature distance between the second time-series feature vector and multiple standard time-series vectors of oil fume components, the type of volatile organic compounds in the oil fume is determined, and the amplitude of the response voltage signal sequence is corrected according to the preset response intensity coefficient corresponding to the type of classification to obtain the corrected response amplitude. Based on the periodic peak change of the correction response amplitude over multiple consecutive acquisition cycles, the short-term increment of the fume concentration is calculated, and when the short-term increment exceeds the dynamic adjustment threshold, an abnormal fume alarm signal is triggered.

2. The method for online monitoring of cooking fumes in hotels and restaurants according to claim 1, characterized in that, The method involves using an array of gas-sensitive elements arranged within the exhaust duct of a hotel restaurant to collect the response voltage signal sequence as fumes pass through, and simultaneously acquiring the internal airflow velocity value of the exhaust duct, including: Multiple metal oxide semiconductor gas sensors are installed at equal angular intervals on the same cross-sectional circumference of the inner wall of the exhaust pipe to form the gas-sensitive element array. A constant heating voltage is applied to each gas sensor in the gas-sensitive element array, and the output resistance value of each gas sensor is continuously acquired at a preset sampling frequency when oil fumes pass by. Each output resistance value is converted into a voltage signal to obtain the response voltage signal sequence. A hot-film flow velocity sensor is embedded at the center of the mounting section of the gas-sensitive element array. The internal airflow velocity value is obtained by measuring the airflow velocity in the exhaust pipe in real time through the hot-film flow velocity sensor.

3. The method for online monitoring of cooking fumes in hotels and restaurants according to claim 2, characterized in that, Based on the rate of change of voltage amplitude at each sampling moment in the response voltage signal sequence, multiple voltage transition edge features are extracted, and the arrival time of each voltage transition edge feature is recorded, including: Calculate the voltage amplitude difference between adjacent sampling times in the response voltage signal sequence to obtain a voltage difference sequence; In the voltage difference sequence, multiple sampling moments with an absolute voltage difference greater than a preset transition threshold are selected as candidate transition moments; Within a preset time window before and after each candidate transition moment, the moment corresponding to the maximum value of the voltage amplitude change rate is searched and taken as the actual arrival moment of the voltage transition edge feature. The direction of voltage amplitude change at the actual arrival moment is recorded, including the rising direction and the falling direction.

4. The method for online monitoring of cooking fumes in hotels and restaurants according to claim 3, characterized in that, The preset transition threshold is the product of the root mean square value of the voltage difference sequence and the preset adjustment coefficient.

5. The method for online monitoring of cooking fumes in hotels and restaurants according to claim 3, characterized in that, A first time-series feature vector is obtained based on the time interval between the arrival times of the plurality of voltage transition edges. A second time-series feature vector is then obtained by performing flow velocity correction on the first time-series feature vector based on the internal airflow velocity value, including: Arrange the arrival times of the multiple voltage transition edge features into a time sequence according to the chronological order, calculate the difference between two adjacent arrival times in the time sequence to obtain multiple arrival time intervals, and combine the multiple arrival time intervals in time sequence to obtain the first time sequence feature vector. The average airflow velocity value of the exhaust pipe during the oil fume collection period is obtained, and the velocity ratio of the average airflow velocity value to the preset standard airflow velocity value is calculated. Each arrival time interval in the first time-series feature vector is multiplied by the velocity ratio to obtain multiple corrected time intervals. The multiple corrected time intervals are then combined in time sequence to obtain the second time-series feature vector.

6. The method for online monitoring of cooking fumes in hotels and restaurants according to claim 4, characterized in that, The average airflow velocity value is the velocity value calculated by time-weighted moving average of the exhaust pipe during the oil fume collection period.

7. The method for online monitoring of cooking fumes in hotels and restaurants according to claim 5, characterized in that, Based on the dynamic time curvature distance between the second time-series feature vector and multiple standard time-series vectors of oil fume components, the classification of volatile organic compounds in the oil fume is determined, and the amplitude of the response voltage signal sequence is corrected according to the preset response intensity coefficient corresponding to the classification to obtain the corrected response amplitude, including: From the preset standard database of oil fume components, retrieve the standard time-series feature vectors corresponding to various volatile organic compounds to form the multiple standard time-series vectors of oil fume components; The dynamic time bending distance between the second time-series feature vector and each of the multiple standard time-series feature vectors of the oil fume components is calculated to obtain multiple bending distance values. The type of volatile organic compound corresponding to the smallest bending distance value among the multiple bending distance values ​​is selected as the type of volatile organic compound; Based on the classification of the volatile organic compounds, the corresponding response intensity coefficient is found from the preset response intensity coefficient mapping table. The voltage amplitude at each sampling moment in the response voltage signal sequence is divided by the response intensity coefficient to obtain the corrected response amplitude at each sampling moment.

8. The method for online monitoring of cooking fumes in hotels and restaurants according to claim 7, characterized in that, The path constraint for calculating the dynamic time bending distance is a global window constraint, and the width of the global window is a preset percentage of the length of the second temporal feature vector.

9. A method for online monitoring of cooking fumes in hotels and restaurants according to claim 7, characterized in that, Based on the periodic peak change of the correction response amplitude over multiple consecutive acquisition cycles, a short-term increment value of the oil fume concentration is calculated. When the short-term increment value exceeds the dynamic adjustment threshold, an oil fume abnormality alarm signal is triggered, including: Within each acquisition cycle, the maximum value of the correction response amplitude in the time domain is obtained as the periodic peak value of that acquisition cycle, and a sequence of periodic peak values ​​for multiple consecutive acquisition cycles is obtained. Calculate the difference between the periodic peak values ​​of two adjacent collection periods in the periodic peak sequence to obtain multiple periodic peak value changes, and take the average of the multiple periodic peak value changes as the short-term increment value of the oil fume concentration; Multiple historical short-term incremental values ​​within the same historical time period are obtained, the standard deviation of the multiple historical short-term incremental values ​​is calculated, and the dynamic adjustment threshold is set according to the standard deviation. The relationship between the short-term incremental value and the dynamic adjustment threshold is compared. When the short-term incremental value is greater than the dynamic adjustment threshold, the oil fume abnormality alarm signal is triggered.

10. An online monitoring system for kitchen fumes in hotels and restaurants, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the online monitoring method for hotel and catering fumes as described in any one of claims 1 to 9.

Images