A toxic gas emission monitoring system for electronic component workshops

Abstract

The invention discloses a toxic gas emission monitoring system for an electronic component workshop, relates to the field of gas monitoring, comprises a sampling and analysis unit, a data acquisition unit, an online detection unit, a data storage unit, a data analysis unit, a data processing unit, a dynamic early warning unit and an abnormality processing unit, wherein the sampling and analysis unit is used for low-frequency acquisition of the toxic gas content at an emission port of a workshop in each time series and generates original data; the toxic gas emission monitoring system for an electronic component workshop acquires original data sets at low frequency and accompanying detection data at high frequency in the same time series and at the same acquisition point to establish an accompanying relationship, thereby performing in-situ real-time measurement in a high-frequency monitoring manner, realizing high-frequency monitoring of toxic gas emissions in a workshop production process, reducing the consumption of manpower and material resources for sampling in conventional low-frequency monitoring, and reducing the error in the low-frequency sampling process and the delay problem of data acquisition.

Description

Technical Field

[0001] The invention relates to a gas monitoring technology, and in particular to a toxic gas emission monitoring system for an electronic component workshop. Background Art

[0002] The production of electronic components is often accompanied by the emission of VOCs, which contains a variety of toxic gases. When workers work in the workshop for a long time, it poses a certain threat of occupational diseases to the human body. Therefore, the exhaust gas generated in the workshop is generally discharged after being discharged after exhaust gas treatment. With the development of society, people's concept of environmental protection is getting stronger and stronger, and there are corresponding emission indicators for exhaust gas emissions.

[0003] At present, electronic component production workshops treat toxic gases emitted from the workshop by setting up environmental treatment equipment. However, due to the complexity of gas detection, the traditional periodic sampling requires a lot of material and human resources, and the delay in the detection of some toxic gas components, the detection results of toxic gas emissions are delayed, and the handling of abnormal emissions is also delayed, which in turn affects the compliance production of the workshop and even affects the environment of the workshop personnel and the surrounding areas. Summary of the invention

[0004] The purpose of the present invention is to provide a toxic gas emission monitoring system for an electronic component workshop to solve the above-mentioned deficiencies in the prior art.

[0005] In order to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a toxic gas emission monitoring system for an electronic component workshop, comprising a sampling and analysis unit, a data acquisition unit, an online detection unit, a data storage unit, a data analysis unit, a data processing unit, a dynamic early warning unit and an abnormality processing unit, wherein the sampling and analysis unit is used to collect the toxic gas content of the workshop emission port at a low frequency in each time series and generate raw data, and transmit the raw data to the data storage unit, the online detection unit is used to receive the data information collected by the sensor component in the data acquisition unit at a high frequency in each time series and generate accompanying monitoring data, and transmit the accompanying data to the data storage unit, the data analysis unit establishes an accompanying relationship between the accompanying monitoring data and the original data based on statistical analysis and evaluation, and transmits it to the data processing unit, the data processing unit uses the accompanying relationship data transmitted by the data analysis unit and establishes a prediction model based on a back propagation neural network, and calculates a prediction residual sequence and generates a threshold interval A through the prediction model in combination with the accompanying monitoring data, the dynamic early warning unit is used to receive the detection data transmitted by the online detection unit and calculate the residual sequence at the current moment in combination with the prediction model, and compares it with the threshold interval A for data judgment, and the dynamic early warning unit transmits the abnormal data of the judgment result to the abnormality processing unit.

[0006] Furthermore, there are two groups of sensor components in the data acquisition unit, and they are respectively located on both sides of the gas filtering device at the workshop exhaust port, and the sensor component close to the side inside the workshop is started and controlled by the abnormality processing unit.

[0007] Furthermore, the data analysis unit performs feasible substitution on the original data by accompanying the monitoring data, and establishes a substitution relationship linear equation based on a statistical analysis method, which is expressed as:

[0008] y i =kx i +b+e i i=1,2,3...,n;

[0009] Where: yi is the i-th toxic gas target variable, xi is the i-th toxic gas alternative variable, ei is the random error of the i-th toxic gas alternative variable, k is the equation slope, b is the equation intercept, and n is the number of sampling times.

[0010] Furthermore, the data processing unit uses wavelet analysis technology to decompose the data into new data of different resolutions, and removes the noise introduced into the time series of the accompanying monitoring data due to sensor errors;

[0011] The denoised data is used for neural network prediction, which uses the first n values ​​in the time series as the input layer of the neural network to predict the n+1th value, and continuously introduces new monitoring data to update the input layer neural network to achieve data-driven dynamic prediction of toxic gas time series;

[0012] The predicted value sequence is reconstructed by wavelet reconstruction method and compared with the actual monitoring value sequence to obtain the predicted residual sequence. The residual distribution equation is obtained by fitting with the distribution function, and the threshold interval A for distinguishing toxic gas data is calculated.

[0013] Furthermore, the one-dimensional discrete time series of the accompanying detection data time series is expressed as:

[0014] W(i)=f(t)+σ×e(t)t=0,1,...n-1;

[0015] Where f(t) is the actual time series, e(t) is the noise time series, i is the sampling time, and σ is the noise coefficient;

[0016] When denoising, the noise coefficient σ is normalized to zero and the discrete wavelet function is obtained:

[0017]

[0018] in is the proportional factor, q 0is the time factor, t is the sampling time, and j is the number of the element in the time series;

[0019] The discretized wavelet coefficients are expressed as:

[0020]

[0021] Wavelet reconstruction is expressed as:

[0022] B is a constant;

[0023] The network training uses the gradient descent algorithm to change the weights and deviations in the direction where the error function is reduced the fastest, and iterative training is performed until the prediction accuracy is met. The iterative calculation formula is expressed as:

[0024] xk+1=xk-akgk;

[0025] where x k is the current weight and bias, x k+1 is the weight and bias of the next moment generated by iteration, a k is the learning rate, g k is the gradient of the current error function.

[0026] Furthermore, the dynamic warning unit outputs the judgment result that the residual is within the threshold range as a normal toxic gas emission event, and transmits the judgment result that the residual is not within the threshold range to the next judgment module to judge whether the duration of the residual not within the threshold range is greater than a preset time, and the preset time is 0.2-1h. If the duration of the residual not within the threshold range is greater than the preset time, it is judged as an abnormal toxic gas emission event, and it is transmitted to the abnormality processing unit. If the duration of the residual not within the threshold range is less than or equal to the preset time, it is judged as a normal toxic gas emission event.

[0027] Furthermore, the abnormality handling unit performs online detection by controlling and starting a sensor component close to one side of the workshop, and compares the difference in monitoring data on both sides of the gas filtering device at the exhaust port of the workshop.

[0028] Compared with the prior art, the toxic gas emission monitoring system for electronic component workshops provided by the present invention has the following beneficial effects:

[0029] 1. The toxic gas emission monitoring system for the electronic component workshop establishes a companion relationship by collecting the original data set at a low frequency and the accompanying detection data at a high frequency in the same time series and the same collection point, so as to perform in-situ real-time measurement through high-frequency monitoring, realize high-frequency monitoring of toxic gas emissions in the workshop production process, reduce the manpower and material consumption of sampling in conventional low-frequency monitoring, and reduce the errors and data acquisition delays in the low-frequency sampling process.

[0030] 2. The toxic gas emission monitoring system used in the electronic component workshop removes noise through wavelet denoising and calculates the threshold interval A used for judgment. The denoised toxic gas time series is trained using a back propagation neural network, and combined with actual toxic gas monitoring data to calculate a better prediction accuracy. The prediction accuracy is improved through iterative training, and the dynamic update of the prediction accuracy is achieved.

[0031] 3. The toxic gas emission monitoring system used in the electronic component workshop uses neural network data to predict data and continuously update the threshold range A to achieve dynamic early warning of abnormal toxic gas emission events, so as to monitor the reliability of toxic gas emissions during production in the electronic component workshop. At the same time, the sensor component close to one side of the workshop is started through the abnormal processing unit for online detection, and the difference in monitoring data on both sides of the gas filter device at the workshop exhaust port is compared to determine whether the gas filter device has failed or whether there is any abnormality in the production of the electronic component workshop. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0033] Figure 1 A principle block diagram of a toxic gas emission monitoring system for an electronic component workshop provided by an embodiment of the present invention;

[0034] Figure 2 A flowchart of a dynamic early warning unit provided in an embodiment of the present invention.

[0035] Description of reference numerals:

[0036] 1. Sampling and analysis unit; 2. Data acquisition unit; 3. Online detection unit; 4. Data storage unit; 5. Data analysis unit; 6. Data processing unit; 7. Dynamic warning unit; 8. Abnormal processing unit. DETAILED DESCRIPTION

[0037] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0038] Embodiment 1:

[0039] See also Figure 1 and Figure 2A toxic gas emission monitoring system for an electronic component workshop comprises a sampling and analysis unit 1, a data acquisition unit 2, an online detection unit 3, a data storage unit 4, a data analysis unit 5, a data processing unit 6, a dynamic early warning unit 7 and an abnormality processing unit 8. The sampling and analysis unit 1 is used to collect the toxic gas content at the workshop discharge port at a low frequency in each time series and generate raw data, and transmit the raw data to the data storage unit 4. The low-frequency collection is generally performed by sampling and detecting the toxic gas at the discharge port at a fixed time and point by the detection equipment. The online detection unit 3 is used to receive the data signal collected by the sensor component in the data acquisition unit 2 at a high frequency in each time series. The information is collected and the accompanying monitoring data is generated, and the accompanying data is transmitted to the data storage unit 4. The high-frequency acquisition generally uses a sensor component for real-time monitoring. The sensor component is a gas refractive index sensor of model FRI-NP or a gas nano-mass density sensor of model TrueDyne. The gas refractive index sensor is used to monitor the refractive index change of the exhaust gas in the workshop in real time. The gas nano-mass density sensor directly measures the gas density through the oscillation frequency. There are two groups of sensor components in the data acquisition unit 2, and they are respectively located on both sides of the gas filter device at the exhaust port of the workshop. The sensor component close to the side in the workshop is started and controlled by the abnormal processing unit 8;

[0040] The data analysis unit 5 establishes the accompanying relationship between the accompanying monitoring data and the original data based on statistical analysis and evaluation, and transmits it to the data processing unit 6. The data analysis unit 5 performs feasible substitution of the original data through the accompanying monitoring data, and establishes the substitution relationship linear equation based on the statistical analysis method, which is expressed as:

[0041] y i =kx i +b+e i i=1,2,3...,n;

[0042] Where: yi is the i-th toxic gas target variable, xi is the i-th toxic gas alternative variable, ei is the random error of the i-th toxic gas alternative variable, k is the equation slope, b is the equation intercept, and n is the number of sampling times. By collecting the original data set at a low frequency and the accompanying detection data at a high frequency in the same time series and at the same collection point, an accompanying relationship is established, so that in-situ real-time measurement is performed through a high-frequency monitoring method, and high-frequency monitoring of toxic gas emissions in the workshop production process is realized, reducing the manpower and material consumption of sampling in conventional low-frequency monitoring, and reducing the errors in the low-frequency sampling process and the delay problem of data acquisition;

[0043] The data processing unit 6 uses the accompanying relationship data transmitted by the data analysis unit 5 and establishes a prediction model based on a back propagation neural network, and calculates the prediction residual sequence and generates a threshold interval A by combining the prediction model with the accompanying monitoring data. The data processing unit 6 uses wavelet analysis technology to decompose the data into new data with different resolutions, and removes the noise introduced by the sensor error in the time series of the accompanying monitoring data;

[0044] The denoised data is used for neural network prediction, which uses the first n values ​​in the time series as the input layer of the neural network to predict the n+1th value, and continuously introduces new monitoring data to update the input layer neural network to achieve data-driven dynamic prediction of toxic gas time series;

[0045] The predicted value sequence is reconstructed by wavelet reconstruction method and compared with the actual monitoring value sequence to obtain the predicted residual sequence, and the residual distribution equation is obtained by fitting with the distribution function, and the threshold interval A for distinguishing toxic gas data is calculated;

[0046] The one-dimensional discrete time series of the accompanying detection data time series is expressed as:

[0047] W(i)=f(t)+σ×e(t)t=0,1,...n-1;

[0048] Where f(t) is the actual time series, e(t) is the noise time series, i is the sampling time, and σ is the noise coefficient;

[0049] When denoising, the noise coefficient σ is normalized to zero and the discrete wavelet function is obtained:

[0050]

[0051] in is the proportional factor, q 0 is the time factor, t is the sampling time, and j is the number of the element in the time series;

[0052] The discretized wavelet coefficients are expressed as:

[0053]

[0054] Wavelet reconstruction is expressed as:

[0055] B is a constant;

[0056] The network training uses the gradient descent algorithm to change the weights and deviations in the direction where the error function is reduced the fastest, and iterative training is performed until the prediction accuracy is met. The iterative calculation formula is expressed as:

[0057] xk+1=xk-akgk;

[0058] where x k is the current weight and bias, x k+1 is the weight and bias of the next moment generated by iteration, a k is the learning rate, g k is the gradient of the current error function; the noise is removed by wavelet denoising and the threshold interval A for judgment is calculated; the denoised toxic gas time series is trained using a back propagation neural network, and a better prediction accuracy is calculated in combination with the actual toxic gas monitoring data, and the prediction accuracy is improved through iterative training to achieve dynamic update of the prediction accuracy;

[0059] The dynamic warning unit 7 is used to receive the detection data transmitted by the online detection unit 3 and calculate the residual sequence at the current moment in combination with the prediction model, and compare it with the threshold interval A to make a data judgment. The dynamic warning unit 7 outputs the judgment result that the residual is within the threshold interval as a normal emission event of toxic gas, and transmits the judgment result that the residual is not within the threshold interval to the next judgment module to judge whether the duration of the residual not being within the threshold interval is greater than a preset time, and the preset time is 0.2-1h. If the duration of the residual not being within the threshold interval is greater than the preset time, it is judged as an abnormal emission event of toxic gas, and it is transmitted to the abnormal processing unit 8. If the duration of the residual not being within the threshold interval is less than or equal to the preset time, it is judged as a normal emission event of toxic gas. The abnormality handling unit 8 starts the sensor component close to one side of the workshop for online detection by controlling and comparing the difference in monitoring data on both sides of the gas filter device at the workshop exhaust port. The neural network data is used to perform data prediction and continuously update the threshold interval A to realize dynamic early warning of abnormal toxic gas emission events, so as to monitor the reliability of toxic gas emissions during the production of electronic components workshops. At the same time, the abnormality handling unit 8 starts the sensor component close to one side of the workshop for online detection, and compares the difference in monitoring data on both sides of the gas filter device at the workshop exhaust port to determine whether the gas filter device fails or whether there is an abnormality in the production of the electronic components workshop. When the difference is small, the gas filter device may have lost its filtering effect. When the difference is large, the production process of the electronic components workshop may be abnormal.

[0060] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A toxic gas emission monitoring system for an electronic component workshop, comprising a sampling and analysis unit (1), characterized in that: It also includes a data acquisition unit (2), an online detection unit (3), a data storage unit (4), a data analysis unit (5), a data processing unit (6), a dynamic early warning unit (7) and an abnormality processing unit (8); A sampling and analysis unit (1), which is used to collect the toxic gas content at the workshop exhaust port at a low frequency in each time series and generate raw data, and transmit the raw data to a data storage unit (4); An online detection unit (3) is used to receive data information collected at high frequency by the sensor component in the data collection unit (2) in each time series and generate accompanying monitoring data, and transmit the accompanying monitoring data to the data storage unit (4); A data analysis unit (5) which establishes a relationship between the accompanying monitoring data and the original data based on statistical analysis and evaluation, and transmits the relationship to a data processing unit (6); A data processing unit (6) uses the accompanying relationship data transmitted by the data analysis unit (5) to establish a prediction model based on a back propagation neural network, and calculates a prediction residual sequence and generates a threshold interval A by combining the prediction model with the accompanying monitoring data; A dynamic early warning unit (7) is used to receive the detection data transmitted by the online detection unit (3) and calculate the residual sequence at the current moment in combination with the prediction model, and compare it with the threshold interval A to make a data judgment. The dynamic early warning unit (7) transmits the abnormal data of the judgment result to the abnormality processing unit (8); The data analysis unit (5) performs feasible substitution on the original data by accompanying the monitoring data, and establishes a substitution relationship linear equation based on a statistical analysis method, which is expressed as: y i =kx i +b+e i ,i=1,2,3...,n; Where: y i is the ith toxic gas target variable, x i is the ith toxic gas substitution variable, e i is the random error of the ith toxic gas substitute variable, k is the equation slope, b is the equation intercept, and n is the number of sampling times; The data processing unit (6) uses wavelet analysis technology to decompose the data into new data with different resolutions, and removes the noise introduced into the time series of the accompanying monitoring data due to sensor errors; The denoised data is used for neural network prediction, which uses the first n values ​​in the time series as the input layer of the neural network to predict the n+1th value, and continuously introduces new monitoring data to update the input layer neural network to achieve data-driven dynamic prediction of toxic gas time series; The predicted value sequence is reconstructed by wavelet reconstruction method and compared with the actual monitoring value sequence to obtain the predicted residual sequence, and the residual distribution equation is obtained by fitting with the distribution function, and the threshold interval A for distinguishing toxic gas data is calculated; The one-dimensional discrete time series of the accompanying monitoring data time series is expressed as: W(i)=f(t)+σ×e(t) t=0,1,...n-1; Where f(t) is the actual time series, e(t) is the noise time series, t is the sampling time, and σ is the noise coefficient; When denoising, the noise coefficient σ is normalized to zero and the discrete wavelet function is obtained: in is the scale factor, q0 is the time factor, t is the sampling time, and j is the number of the element in the time series; The discretized wavelet coefficients are expressed as: Wavelet reconstruction is expressed as: B is a constant; The network training adopts the gradient descent algorithm, changes the weights and deviations along the direction where the error function is reduced the fastest, and performs iterative training until the prediction accuracy is met. The iterative calculation formula is expressed as: x k+1 =x k -a k g k ; where x k is the current weight and bias, x k+1 is the weight and bias of the next moment generated by iteration, a k is the learning rate, g k is the gradient of the current error function.

2. A toxic gas emission monitoring system for an electronic component workshop according to claim 1, characterized in that: The data acquisition unit (2) has two groups of sensor components, which are respectively located on both sides of the gas filter device at the workshop exhaust port, and the sensor component close to the workshop side is activated and controlled by the abnormality processing unit (8).

3. The toxic gas emission monitoring system for electronic component workshop according to claim 1 is characterized in that: The dynamic warning unit (7) outputs the judgment result that the residual is within the threshold range as a normal toxic gas emission event, and transmits the judgment result that the residual is not within the threshold range to the next judgment module, and judges whether the duration of the residual not being within the threshold range is greater than a preset time, and the preset time is 0.2-1h. If the duration of the residual not being within the threshold range is greater than the preset time, it is judged as an abnormal toxic gas emission event, and it is transmitted to the abnormal processing unit (8). If the duration of the residual not being within the threshold range is less than or equal to the preset time, it is judged as a normal toxic gas emission event.

4. The toxic gas emission monitoring system for electronic component workshop according to claim 1 is characterized in that: The abnormality processing unit (8) performs online detection by controlling and starting the sensor component close to one side of the workshop, and compares the difference in monitoring data on both sides of the gas filtering device at the workshop exhaust port.

Images