CN112255364B - A Soft Sensing Method for Real-time Quantitative Judgment of Sintering End Point State - Google Patents

Abstract

The invention discloses a soft measurement method for quantitatively judging the state of a sintering end point in real time, belonging to the field of sintering process control. According to the method, a secondary curve is fitted to a thermocouple insertion position and a waste gas temperature detection value on the basis of a waste gas temperature detection value of an air box in a sintering production line and an image shot by a tail CCD camera, and a quantitative end point position (MBTP) and a quantitative end point temperature (Tmax) are preliminarily obtained by calculating coordinates (X and Y) of a maximum value point of the curve. The air leakage problem of the sintering production line occasionally causes the detection distortion of the exhaust gas temperature of the air box, so that the secondary curve fitting method obtains wrong end point state, and the end point state obtained by the secondary curve fitting method is corrected by adopting a machine tail image recognition result. The terminal position and the temperature obtained by calculation by the method are more real and accurate, and the method can assist field operators in judging the terminal state and the change trend thereof quantitatively in real time.

Description

A Soft Sensing Method for Real-time Quantitative Judgment of Sintering End Point State

technical field

The invention relates to a soft measurement method for quantitatively determining the state of a sintering end point in real time, and belongs to the field of sintering process control.

Background technique

In sintering production, the sintering end point is an important process parameter that affects the quality, output and cost of sintering ore. A suitable and stable end point state is the key to ensuring that the blast furnace can obtain high-quality sintering ore. When the sintering end point is advanced, the production capacity of the sintering machine cannot be fully utilized, resulting in a decrease in the output of sintered ore; when the sintering end point is delayed, the mixture will be unloaded at the end of the machine without fully burning through, resulting in a decrease in the qualified rate of sintered ore and the cost of rise.

The environment of the sintering machine tail is harsh. Under such conditions of high temperature, high humidity, high dust and strong interference, there is currently no instrument to directly detect the end point of sintering. Whether the end position is estimated by the exhaust gas temperature method, the negative pressure method and the exhaust gas composition judgment method, or the red layer is observed at the tail of the machine, the end state can only be judged qualitatively, which will inevitably affect the precise control of the end state and the parameters of the sintering process. Optimization adjustment. Therefore, the study of how to quantitatively and accurately detect the end state has important guiding significance for the refinement of the sintering production process.

The quadratic curve fitting method based on the exhaust gas temperature of the bellows is currently the commonly adopted end-point state soft measurement method. During the sintering production process, there are often problems such as blow-by on the material surface of the trolley, air leakage from the tail of the machine, etc., which cause the exhaust gas temperature of the bellows to be inconsistent with the actual temperature, resulting in jitter or deviation in the end state fitting results based on the exhaust gas temperature of the bellows. The output result cannot effectively guide the on-site operator to accurately judge the end state. In order to obtain more stable and accurate endpoint state detection results, the soft sensing method of endpoint state has been improved and optimized.

SUMMARY OF THE INVENTION

The invention proposes a soft measurement method for quantitatively determining the sintering end state in real time, which can perform real-time, quantitative and stable detection on the end state in the sintering production process, which is important for field operators to control the end state in time and carry out refined operations. guiding role.

Specifically, a soft measurement method for quantitatively determining the state of a sintering end point in real time includes the following steps:

Step 1: Data acquisition, real-time acquisition of the exhaust gas temperature detection value of the bellows in the sintering production line and the image captured by the tail CCD camera;

Step 2: Data fitting, by fitting a quadratic curve between the thermocouple insertion position and the exhaust gas temperature detection value, obtaining the coordinates (X, Y) of the maximum point of the curve, and initially obtaining the quantitative end position (MBTP) and end temperature (Tmax);

Step 3: Classification and judgment of the tail image, the tail image is divided into three categories, namely under-burning, normal and over-burning, and the convolutional neural network model is used to classify and judge the tail image;

Step 4: End point position correction. Based on the recognized category of the tail image, an expert rule is established to correct the end point state obtained by the quadratic curve fitting method, and real and accurate end point state information is obtained in real time.

The first step specifically includes:

Real-time acquisition of the detection value of the exhaust gas temperature of the bellows in the whole sintering production line and the specific placement position of the thermocouple in the bellows, recording the tail image captured by the CCD camera of the tail of the sintering machine during the same period, the exhaust gas temperature of the bellows and the detection image of the tail as the end state The basic data source for real-time judgment.

The second step specifically includes:

The quadratic curve is fitted according to the thermocouple placement position and the exhaust gas temperature detection value, and the end position (MBTP) and end temperature (Tmax) are preliminarily calculated as follows.

2-1. When the exhaust gas temperature of the last bellows is the largest, the sintering end point (BTP) = the last bellows number (Num) - 0.5 (the thermocouple is generally placed in the middle of the bellows, so take 0.5, and this coefficient can be adjusted according to the actual situation. Adjustment), the sintering end position (MBTP) = the sintering end (BTP) * the length of the bellows, and the sintering end temperature (Tmax) is the exhaust gas temperature of the last bellows.

2-2. When the exhaust gas temperature of the penultimate bellows is the largest, take the exhaust gas temperature detection values of the penultimate third bellows, the penultimate bellows and the last bellows (T _Num-2 , T _Num-1 , T) And the thermocouple position (Num-2.5, Num-1.5, Num-0.5) to fit the quadratic curve, and calculate the coordinates (X, Y) of the maximum point of the curve. Assuming the form of the quadratic equation of one variable is aX ² +bX+c=Y, take the above thermocouple position as X, take the detected value of exhaust gas temperature as Y, and bring each set of detected data into the equation to solve the coefficients a, b and c respectively. Sintering end point (BTP) = -b/(2a), sintering end point position (MBTP) = sintering end point (BTP) * length of bellows, sintering end point temperature (Tmax) = a*BTP ² +b*BTP+c.

2-3. Through statistics on the fluctuation range of the historical end point of multiple sintering machines, it is determined that the fluctuation range of the end point is generally between (Num-6) bellows and (Num) bellows. For the end state between (Num-5) bellows and (Num-1) bellows, the methods described in steps 2-2 are used to calculate.

2-4. When (Num-6) the exhaust gas temperature of the bellows is the largest, the sintering end point (BTP) = the number of the bellows (Num-6) - 0.5 (the thermocouple placement position is generally in the middle of the bellows, so take 0.5, this coefficient can be Adjust according to the actual situation), the sintering end position (MBTP) = the sintering end (BTP) * the length of the bellows, and the sintering end temperature (Tmax) is (Num-6) the exhaust gas temperature of the bellows.

The step 3 specifically includes:

According to the different positions of the red fire layer in the tail image, the images are divided into 3 categories, namely under-burning, normal and over-burning. Under-burning means that the red fire layer is at the top of the tail image, and the lower part of the red fire layer is all raw material; normally, the red fire layer is at the bottom of the tail image, accounting for about one-third of the height of the entire section; over-burning means that it is at the bottom of the tail image There is only a narrow red fire layer, and the upper part of the red fire layer is full of sintered blocks. A convolutional neural network (CNN) model is used to classify and judge the tail images.

The step 4 specifically includes:

Based on the category of the tail image recognized by the convolutional neural network (CNN) model, the end point state obtained by the quadratic curve fitting method is corrected. The specific expert rules are as follows.

4-1. When the tail image is recognized as normal by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is within the range of (Num-2) to (Num-1) bellows, Then output the end position and end temperature obtained by the quadratic curve fitting method;

4-2. When the tail image is recognized as normal by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is outside the range of (Num-2) to (Num-1) bellows, Then output the end position and end temperature between (Num-2) and (Num-1) bellows at the last moment;

4-3. When the tail image is identified as underburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is within the range of (Num-6) to (Num-2) bellows , then output the end position and end temperature obtained by the quadratic curve fitting method;

4-4. When the tail image is identified as underburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is in the range from (Num-6) to (Num-2) outside the bellows , then output the end position and end temperature between (Num-6) and (Num-3) bellows at the last moment;

4-5. When the tail image is identified as overburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is within the range of (Num) to (Num-1) bellows, then Output the end position and end temperature obtained by the quadratic curve fitting method;

4-6. When the tail image is identified as overburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is in the range from (Num) to (Num-1) outside the bellows, then Output the end position and end temperature between (Num) and (Num-1) the bellows range at the last moment;

A soft measurement method for quantitatively determining the state of a sintering end point in real time, comprising: a CCD camera, a storage device, a processor, etc.; the processor loads and executes the data, images and rules in the storage device, so as to realize the claims 1-5 A soft sensing method for quantitatively determining the state of the sintering end point in real time.

Compared with the prior art, the advantages of the present invention are:

(1) A soft measurement method of the present invention for quantitatively determining the state of the sintering end point in real time. The sintering end point is multiplied by the length of the bellows to obtain the distance between the sintering end position and the starting point of the No. 1 bellows, which improves the accuracy of the end position judgment.

(2) A soft measurement method of the present invention for quantitatively judging the end state of sintering in real time, selects various data sources such as numerical values (bellows exhaust gas temperature), images (aircraft tail image) and other data sources as the signals for end state judgment, fully considering and Input information related to the end state.

(3) A soft measurement method of the present invention for quantitatively determining the end state of sintering in real time, using the convolutional neural network to recognize the category of the tail image as the benchmark, and correcting the end state obtained by the quadratic curve fitting method, which improves the actual situation. Stability of endpoint state detection during production.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings and embodiments, in which:

FIG. 1 is a flow chart of a soft measurement method for quantitatively determining the state of the sintering end point in real time according to the present invention.

FIG. 2 is a schematic diagram of a tail image category of the present invention.

FIG. 3 is a comparison diagram of the correction of the quadratic curve fitting value using the image recognition result of the tail of the aircraft.

FIG. 4 is an application effect diagram of the sintering end state detection of the present invention.

Detailed ways

In order to make the technical features, objects and effects of the present invention more clearly understood, the specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

A soft measurement method for quantitatively determining the state of the sintering end point in real time, as shown in Figure 1, specifically includes the following steps:

Specifically, a soft measurement method for quantitatively determining the state of a sintering end point in real time includes the following steps:

Step 1: Data acquisition, real-time acquisition of the exhaust gas temperature detection value of the bellows in the sintering production line and the image captured by the tail CCD camera;

Step 2: Data fitting, by fitting a quadratic curve between the thermocouple insertion position and the exhaust gas temperature detection value, obtaining the coordinates (X, Y) of the maximum point of the curve, and initially obtaining the quantitative end position (MBTP) and end temperature (Tmax);

Step 3: Classification and judgment of the tail image, the tail image is divided into three categories, namely under-burning, normal and over-burning, and the convolutional neural network model is used to classify and judge the tail image;

Step 4: End point position correction. Based on the recognized category of the tail image, an expert rule is established to correct the end point state obtained by the quadratic curve fitting method, and real and accurate end point state information is obtained in real time.

The first step specifically includes:

Real-time acquisition of the detection value of the exhaust gas temperature of the bellows in the whole sintering production line and the specific placement position of the thermocouple in the bellows, recording the tail image captured by the CCD camera of the tail of the sintering machine during the same period, the exhaust gas temperature of the bellows and the detection image of the tail as the end state The basic data source for real-time judgment.

The second step specifically includes:

The quadratic curve is fitted according to the thermocouple placement position and the exhaust gas temperature detection value, and the end position (MBTP) and end temperature (Tmax) are preliminarily calculated as follows.

2-1. When the exhaust gas temperature of the last bellows is the largest, the sintering end point (BTP) = the last bellows number (Num) - 0.5 (the thermocouple is generally placed in the middle of the bellows, so take 0.5, and this coefficient can be adjusted according to the actual situation. Adjustment), the sintering end position (MBTP) = the sintering end (BTP) * the length of the bellows, and the sintering end temperature (Tmax) is the exhaust gas temperature of the last bellows.

2-2. When the exhaust gas temperature of the penultimate bellows is the largest, take the exhaust gas temperature detection values of the penultimate third bellows, the penultimate bellows and the last bellows (T _Num-2 , T _Num-1 , T) And the thermocouple position (Num-2.5, Num-1.5, Num-0.5) to fit the quadratic curve, and calculate the coordinates (X, Y) of the maximum point of the curve. Assuming the form of the quadratic equation of one variable is aX ² +bX+c=Y, take the above thermocouple position as X, take the detected value of exhaust gas temperature as Y, and bring each set of detected data into the equation to solve the coefficients a, b and c respectively. Sintering end point (BTP) = -b/(2a), sintering end point position (MBTP) = sintering end point (BTP) * length of bellows, sintering end point temperature (Tmax) = a*BTP ² +b*BTP+c.

2-3. Through statistics on the fluctuation range of the historical end point of multiple sintering machines, it is determined that the fluctuation range of the end point is generally between (Num-6) bellows and (Num) bellows. For the end state between (Num-5) bellows and (Num-1) bellows, the methods described in steps 2-2 are used to calculate.

2-4. When (Num-6) the exhaust gas temperature of the bellows is the largest, the sintering end point (BTP) = the number of the bellows (Num-6) - 0.5 (the thermocouple placement position is generally in the middle of the bellows, so take 0.5, this coefficient can be Adjust according to the actual situation), the sintering end position (MBTP) = the sintering end (BTP) * the length of the bellows, and the sintering end temperature (Tmax) is (Num-6) the exhaust gas temperature of the bellows.

The step 3 specifically includes:

According to the different positions of the red fire layer in the tail image, the images are divided into 3 categories, namely under-burning, normal and over-burning. Under-burning means that the red fire layer is at the top of the tail image, and the lower part of the red fire layer is all raw material; normally, the red fire layer is at the bottom of the tail image, accounting for about one-third of the height of the entire section; over-burning means that it is at the bottom of the tail image There is only a narrow red fire layer, and the upper part of the red fire layer is full of sintered blocks. A convolutional neural network (CNN) model is used to classify and judge the tail images.

The step 4 specifically includes:

Based on the category of the tail image recognized by the convolutional neural network (CNN) model, the end point state obtained by the quadratic curve fitting method is corrected. The specific expert rules are as follows.

4-1. When the tail image is recognized as normal by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is within the range of (Num-2) to (Num-1) bellows, Then output the end position and end temperature obtained by the quadratic curve fitting method;

4-2. When the tail image is recognized as normal by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is outside the range of (Num-2) to (Num-1) bellows, Then output the end position and end temperature between (Num-2) and (Num-1) bellows at the last moment;

4-3. When the tail image is identified as underburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is within the range of (Num-6) to (Num-2) bellows , then output the end position and end temperature obtained by the quadratic curve fitting method;

4-4. When the tail image is identified as underburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is in the range from (Num-6) to (Num-2) outside the bellows , then output the end position and end temperature between (Num-6) and (Num-3) bellows at the last moment;

4-5. When the tail image is identified as overburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is within the range of (Num) to (Num-1) bellows, then Output the end position and end temperature obtained by the quadratic curve fitting method;

4-6. When the tail image is identified as overburned by the convolutional neural network (CNN) model, if the end position state obtained by the quadratic curve fitting method is in the range from (Num) to (Num-1) outside the bellows, then Output the end position and end temperature between (Num) and (Num-1) the bellows range at the last moment;

The above method takes the exhaust gas temperature of the bellows and the tail image as the input signals, and the sintering end position and the end temperature as the output results, and realizes the online and quantitative detection of the end state, which has important application value for the field operator to guide the sintering production.

The embodiment of the present invention is based on the historical bellows exhaust gas temperature detection value of a 360m ² sintering machine in a sintering plant and the CCD camera at the rear of the machine, and then constructs a soft measurement method for judging the state of the sintering end point according to the above steps 1 to 4, and finally adopts this The actual production data of the sintering machine is tested. Specific steps include:

(1) Data acquisition

Take the application of a ^360m2 sintering machine in a domestic sintering plant as an example. The sintering machine has an effective sintering area of 80 meters and is equipped with 22 bellows. A thermocouple is installed in the middle of the bellows 1#, 2#, 3#, 5#, 7#, 9#, 11#, 13#, 15#, 16#, 18#, 20#, 21# and 22#. , used for second-level continuous temperature measurement of the exhaust gas temperature in the bellows. The detection value of historical exhaust gas temperature and the detection video of the tail CCD camera are obtained from the on-site automation system. The time span is from June 2019 to June 2020, and the frequency of exhaust gas temperature collection is 1 minute.

(2) Data fitting

Based on the above historical data, the quadratic curve is fitted according to the thermocouple placement position of 16#, 18#, 20#, 21# and 22# bellows and the detected value of exhaust gas temperature, and the end position (MBTP) and end temperature ( Tmax).

2-1. When the exhaust gas temperature of the 22# bellows is the largest, the sintering end point (BTP) = 22 (Num)-0.5 (the thermocouple is placed in the middle of the bellows, so take 0.5) = 21.5, and the sintering end point (MBTP) = 21.5 (BTP) * 3.636 (length of bellows) = 78.174 (m), the sintering end temperature (Tmax) is the detected value of exhaust gas temperature of 22# bellows.

2-2. When the exhaust gas temperature of the 21# bellows is the largest, take the thermocouple positions (19.5, 20.5, 21.5) of the 20#, 21# and 22# bellows and the detection value of the exhaust gas temperature (T _Num-2 , T _Num-1 , T) fitting a quadratic curve. Using the above three sets of points to obtain three quadratic equations in one variable, the coefficients of the equations are calculated: a=(T _Num-2 +T-2*T _Num-1 )/2, b= 41*T _Num-1 -21 *T _Num-2 -20*T, c=(4*T _Num-2 -1521*a-78*b)/4. Sintering end point (BTP) = -b/(2a), sintering end point position (MBTP) = sintering end point (BTP) * 3.636 (length of bellows), sintering end point temperature (Tmax) = a*BTP ² +b*BTP+c .

2-3. When the exhaust gas temperature of the 20# bellows is the largest, take the thermocouple positions (17.5, 19.5, 20.5) of the 18#, 20# and 21# bellows and the detection value of the exhaust gas temperature (T _Num-4 , T _Num-2 , T _Num-1 ) to fit a quadratic curve. Using the above three sets of points to obtain three quadratic equations in one variable, the coefficients of the equations are calculated: a=(T _Num-4 +2*T _Num-1 -3*T _Num-2 )/6, b=(57* T _Num-2 -20*T _Num-4 -37*T _Num-1 )/3, c=(4* T _Num-4 -1225*a-70*b)/4. Sintering end point (BTP) = -b/(2a), sintering end point position (MBTP) = sintering end point (BTP) * 3.636 (length of bellows), sintering end point temperature (Tmax) = a*BTP ² +b*BTP+c .

2-4. When the exhaust gas temperature of the 18# bellows is the largest, take the thermocouple positions (15.5, 17.5, 19.5) of the 16#, 18# and 20# bellows and the detection value of the exhaust gas temperature (T _Num-6 , T _Num-4 , T _Num-2 ) to fit a quadratic curve. Using the above three sets of points to obtain three quadratic equations in one variable, the coefficients of the equations are calculated: a=(T _Num-6 + T _Num-2 -2* T _Num-4 )/8, b=(70*T _{Num -4} -37* T _Num-6 -33* T _Num-2 )/8, c=(4* T _Num-6 -961*a-62*b)/4. Sintering end point (BTP) = -b/(2a), sintering end point position (MBTP) = sintering end point (BTP) * 3.636 (length of bellows), sintering end point temperature (Tmax) = a*BTP ² +b*BTP+c .

2-5. When the exhaust gas temperature of the 16# bellows is the highest, the sintering end point (BTP) = 16 (Num-6)-0.5 (the thermocouple is placed in the middle of the bellows, so take 0.5) = 15.5, the sintering end point (MBTP) =15.5(BTP)*3.636(length of bellows)=56.358(m), the sintering end temperature (Tmax) is the detected value of exhaust gas temperature of 16# bellows.

(3) Classification and judgment of tail images

According to the collection frequency of the exhaust gas temperature detection value, the convolutional neural network model is used to classify and judge the detection video of the tail of the machine in the same period to determine whether the position status of the end point is under-burning, normal, or over-burning.

(4) End position correction

Based on the recognized category of the tail image, the end point state obtained by the quadratic curve fitting method is corrected.

The original end position state obtained by quadratic curve fitting is compared with the end position state after correction based on tail image recognition, as shown in Figure 3. At positions ① and ② in the figure, the convolutional neural network model is used to identify that the tail image is in the over-burned and normal states, respectively. Therefore, the process of 4-6 and 4-2 in step 4 of the soft measurement method according to the sintering end state Rule, the end position state of the quadratic curve fitting result is corrected.

(5) Application effect

The soft measurement model of the sintering end state established by the above method was applied to the latest production data on site for testing. The actual production data of the above 360m ² sintering machine in August 2020 was selected, and the results of about 1000 groups of test samples were intercepted for display. ,As shown in Figure 4. In Figure 4, the abscissa is the number of test samples, and the ordinate is the end point position and end point temperature. In the figure, the original end point state of quadratic curve fitting and the end point state after correction based on image recognition are drawn at the same time, and curves with different marks are drawn. expressed. It can be seen from Figure 4 that the corrected end-point state based on image recognition can not only display the actual situation of the end-point of on-site sintering more realistically and stably, but also has higher accuracy. Therefore, the soft measurement method for real-time quantitative determination of the sintering end state proposed by the present invention has a good application effect, and can be well applied and extended to the automation systems of other sintering machines.

The specific embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above-mentioned specific embodiments, which are merely illustrative rather than restrictive. Under the inspiration of the present invention, those skilled in the art can also make many forms without departing from the scope of the present invention and the protection scope of the claims, which all belong to the protection scope of the present invention.

Claims (4)

1. A soft sensing method for quantitatively determining the state of a sintering end point in real time, comprising the following steps: Step 1: Data acquisition, real-time acquisition of the exhaust gas temperature detection value of the bellows in the sintering production line and the image captured by the tail CCD camera; Step 2: Data fitting, by fitting a quadratic curve to the thermocouple insertion position and the exhaust gas temperature detection value, obtaining the coordinates (X, Y) of the maximum point of the curve, and initially obtaining the quantitative end position MBTP and End point temperature Tmax, 2-1. When the exhaust gas temperature of the last bellows is the highest, the sintering end point BTP = the last bellows number Num-0.5, the thermocouple is generally placed in the middle of the bellows, so take 0.5, this coefficient can be adjusted according to the actual situation, the sintering end point Position MBTP = sintering end point BTP * length of bellows, sintering end point temperature Tmax is the exhaust gas temperature of the last bellows, 2-2. When the exhaust gas temperature of the penultimate bellows is the largest, take the exhaust gas temperature detection values of the penultimate third bellows, the penultimate bellows and the last bellows (T _Num-2 , T _Num-1 , T) And the thermocouple position (Num-2.5, Num-1.5, Num-0.5) to fit the quadratic curve, calculate the coordinates (X, Y) of the maximum point of the curve, assuming the form of the quadratic equation is aX ² +bX+ c=Y, take the position of the above thermocouple as X, take the detected value of exhaust gas temperature as Y, bring each set of detected data into the equation to solve the coefficients a, b and c respectively, the sintering end point BTP=-b/(2a), the sintering end point Position MBTP=sintering end point BTP* length of bellows, sintering end point temperature Tmax=a*BTP ² +b*BTP+c, 2-3. Through the statistics of the fluctuation range of the historical end points of multiple sintering machines, it is determined that the fluctuation range of the end point is generally between the Num-6 bellows and the Num bellows, and the end point state between the Num-5 bellows and the Num-1 bellows is determined. , are calculated by the method described in 2-2 steps, 2-4. When the exhaust gas temperature of the Num-6 bellows is the highest, the sintering end point BTP = the bellows number Num-6-0.5. The thermocouple is generally placed in the middle of the bellows, so take 0.5. This coefficient can be adjusted according to the actual situation. The end position MBTP = the sintering end BTP * the length of the bellows, and the sintering end temperature Tmax is the exhaust gas temperature of the Num-6 bellows; Step 3: Classification and judgment of the tail image, the tail image is divided into three categories, namely under-burning, normal and over-burning, and the convolutional neural network model is used to classify and judge the tail image; Step 4: End-point position correction. Based on the recognized category of the tail image, establish expert rules to correct the end-point state obtained by the quadratic curve fitting method, and obtain real and accurate end-point state information in real time. The specific expert rules are as follows , 4-1. When the tail image is recognized as normal by the convolutional neural network CNN model, if the end position state obtained by the quadratic curve fitting method is within the range of Num-2 to Num-1 bellows, the quadratic curve fitting will be output. legally obtained end position and end temperature, 4-2. When the tail image is recognized as normal by the convolutional neural network CNN model, if the end position state obtained by the quadratic curve fitting method is outside the range of Num-2 to Num-1 bellows, the output is at the last moment. the end position and end temperature between the bellows range Num-2 to Num-1, 4-3. When the tail image is identified as under-burned by the convolutional neural network CNN model, if the end position state obtained by the quadratic curve fitting method is between the bellows range of Num-6 and Num-2, the quadratic curve will be output. The end position and end temperature obtained by the fitting method, 4-4. When the tail image is identified as under-burned by the convolutional neural network CNN model, if the end position state obtained by the quadratic curve fitting method is outside the range of Num-6 to Num-2 bellows, output the last moment the end position and end temperature between the bellows range Num-6 to Num-3, 4-5. When the tail image is identified as overburned by the convolutional neural network CNN model, if the end position state obtained by the quadratic curve fitting method is between Num and Num-1 bellows, the quadratic curve fitting method is output. obtained end position and end temperature, 4-6. When the tail image is identified as overburned by the convolutional neural network CNN model, if the end position state obtained by the quadratic curve fitting method is in the range from Num to Num-1 bellows, the output is located at Num at the last moment. To Num-) the end position and end temperature of the bellows range. 2. The soft measurement method for quantitatively determining the sintering end state in real time according to claim 1, wherein step 1 specifically comprises: Real-time acquisition of the detection value of the exhaust gas temperature of the bellows in the whole sintering production line and the specific placement position of the thermocouple in the bellows, recording the tail image captured by the CCD camera of the tail of the sintering machine during the same period, the exhaust gas temperature of the bellows and the detection image of the tail as the end state The basic data source for real-time judgment. 3. A kind of soft measurement method for quantitatively determining sintering end state in real time according to claim 1, it is characterized in that, step 3 specifically comprises: According to the different positions of the red fire layer in the tail image, the image is divided into 3 categories, namely under-burning, normal and over-burning. Under-burning means that the red fire layer is located at the top of the tail image, and the lower part of the red fire layer is all raw meal; normal; The red fire layer is at the bottom of the tail image, accounting for one-third of the height of the entire section; over-burning means that there is only a narrow red fire layer at the bottom of the tail image, and the upper part of the red fire layer is full of sintered blocks, using the convolutional neural network CNN model Classify and judge the tail image. 4. A kind of soft measurement method for quantitatively determining sintering end state in real time according to claim 1, characterized in that, comprising: a CCD camera, a storage device and a processor; the processor loads and executes the data in the storage device, The image and the rules are used to realize the soft measurement method for quantitatively determining the sintering end state in real time according to any one of claims 1 to 3.

Images