An abnormal electrolysis condition identification method

Abstract

The present invention provides a method for identifying abnormal electrolysis conditions, including: Step 1, obtaining historical infrared thermal images of a target electrolytic cell, and determining a characteristic auxiliary line perpendicular to the plate of the target electrolytic cell in the historical infrared thermal images; Step 2, obtaining temperature peak points on the characteristic auxiliary line to obtain a plurality of abnormal observation points; Step 3, for each abnormal observation point among the plurality of abnormal observation points, extracting a spatial temperature feature sequence in the spatial dimension and a temporal temperature feature sequence in the temporal dimension according to the abnormal observation point; Step 4, respectively performing hierarchical clustering on the plurality of abnormal observation points according to the spatial temperature feature sequence and the temporal temperature feature sequence to obtain a clustering result, and identifying the plate in the abnormal condition according to the clustering result; avoiding the inaccuracy caused by artificially extracting the temperature representative value and setting the temperature threshold when judging the plate condition through the temperature threshold, thereby realizing more accurate and more refined intelligent identification of abnormal electrolysis conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of anomaly recognition, and particularly relates to a method for recognizing abnormal electrolysis conditions. Background Art

[0002] In the smelting of non-ferrous metals such as copper, lead, and zinc, electrolysis is the final process for purifying and refining non-ferrous metals, which directly determines the quality of non-ferrous metal products. At the same time, the electrolysis power consumption accounts for a large proportion in the total energy consumption of non-ferrous metal smelting, and abnormal electrolysis conditions will significantly affect the power consumption cost of non-ferrous metal smelting. Therefore, the intelligent recognition of abnormal electrolysis conditions has great practical significance for improving the quality and reducing the consumption of non-ferrous metal smelting.

[0003] Most of the existing methods identify the abnormal condition of inter-electrode short circuit in abnormal electrolysis conditions based on electrical signals and temperature signals, and rarely classify and identify abnormal electrolysis conditions such as inter-electrode short circuit, abnormal connection of electrode plates, and open circuit. Electrical signals usually include cell voltage signals and electromagnetic signals. The detection of abnormal electrolysis conditions based on cell voltage signals mostly uses sensor contact measurement, resulting in serious corrosion and damage of system equipment, difficult maintenance, and low input-output ratio. The identification of abnormal electrolysis conditions based on electromagnetic signals is usually carried out by workers using a drag meter to scan the electromagnetic field signals slot by slot and electrode plate by electrode plate. Although this method is sensitive in measurement, due to the lack of a large-scale and batch measurement scheme for electrode plate electromagnetic signals, the workload of this method is large and the measurement results lag seriously. The identification of abnormal electrolysis conditions based on temperature signals often identifies abnormal electrolysis conditions based on the temperature value at a single moment or an infrared thermal image. The temperature value at a single moment and the thermal imaging graphic features are used. Under complex on-site interference, the temperature characteristics of the electrolytic cell at a single moment cannot effectively identify all abnormal electrode plates. Summary of the Invention

[0004] The present invention provides a method for recognizing abnormal electrolysis conditions, and its purpose is to improve the accuracy and precision of intelligent recognition of abnormal electrolysis conditions.

[0005] In order to achieve the above purpose, the present invention provides a method for recognizing abnormal electrolysis conditions, including:

[0006] Step 1, obtaining historical infrared thermal images of a target electrolytic cell, and determining a characteristic auxiliary line perpendicular to the direction of the electrode plates of the target electrolytic cell in the historical infrared thermal images;

[0007] Step 2, obtaining temperature peak points on the characteristic auxiliary line to obtain a plurality of abnormal observation points;

[0008] Step 3, for each abnormal observation point among the plurality of abnormal observation points, extracting a spatial temperature feature sequence in the spatial dimension and a temporal temperature feature sequence in the temporal dimension according to the abnormal observation point;

[0009] Step 4: Perform hierarchical clustering on multiple abnormal observation points according to the spatial temperature feature sequence and the time temperature feature sequence respectively to obtain the clustering result, and identify the electrode plates in abnormal working conditions according to the clustering result.

[0010] Furthermore, Step 1 includes:

[0011] Continuously collect N infrared thermal images with a period of T;

[0012] For each of the N infrared thermal images, determine multiple characteristic auxiliary lines perpendicular to the direction of the target electrolytic cell electrode plate in the infrared thermal image;

[0013] For each of the multiple characteristic auxiliary lines, extract the temperature change in the time dimension according to the characteristic auxiliary line and form a temperature change matrix.

[0014] Furthermore, Step 2 includes:

[0015] For each of the multiple characteristic auxiliary lines, form a temperature sequence with the points on the characteristic auxiliary line;

[0016] Perform one-sided difference processing on the temperature sequence to obtain a temperature value difference sequence;

[0017] Search for the points that cross the line y = 0 from positive to negative on the temperature value difference sequence to obtain the temperature peak points and get multiple abnormal observation points.

[0018] Furthermore, the temperature sequence T k is:

[0019] T k ={t k (i)|i∈R, i∈[1, N pixel_k}

[0020] Perform one-sided difference processing on the temperature sequence T k to obtain a temperature value difference sequence T' k which is:

[0021] T′ k ={t′ k (i)|i∈N, i∈[1, N pixel_k}

[0022] where t k (i) represents the infrared measurement temperature value at the i-th pixel position on the k-th characteristic auxiliary line, R represents the total number of pixel positions, N pixel_k represents the pixel length value of the k-th characteristic auxiliary line, which is equal to the pixel width perpendicular to the electrode plate direction of the cropped single-cell image, and t' k (i)=t k (i)-tk (i - 1), (i ∈ N, i ∈ [2, N pixel_k ), where N represents the total number of pixel positions after one-sided difference;

[0023] Search for the points where the temperature value difference sequence crosses the line y = 0 from positive to negative, and obtain the set I of temperature upper peak points hight ={i | t' k (i) < 0, t' k (i - 1) > 0, i ∈ [2, N pixel_k}, and obtain multiple abnormal observation points.

[0024] Furthermore, before step 3, it also includes:

[0025] Obtain the set I of lower peak points low ={i | t' k (i) > 0, t' k (i - 1) < 0, i ∈ [2, N pixel_k};

[0026] In the set of upper peak points, search for the first batch of upper peak points with temperatures greater than the preset temperature threshold from both ends of the set of upper peak points, and obtain the first upper peak point I high_1 and the second upper peak point Among them, the first upper peak point I high_1 is the first upper peak point with a temperature greater than the preset temperature threshold searched from the head end of the set of upper peak points, and the second upper peak point is the first upper peak point with a temperature greater than the preset temperature threshold searched from the tail end of the set of upper peak points,

[0027] Between the first upper peak point I high_1 and the second upper peak point , search for the first lower peak point I high_1 that is less than and closest to the first upper peak point I low_1 and the second lower peak point I high_2 that is greater than and closest to I low_2 ;

[0028] Crop the historical infrared thermal image and the temperature change matrix, and retain the image area located between the first lower peak point I low_1 and the second lower peak point I low_2 .

[0029] Furthermore, step 3 includes:

[0030] Extract the thermal image pixel sequences extended in the spatial dimension of all abnormal observation points in the target electrolytic cell, and obtain multiple spatial temperature feature sequences;

[0031] Extract the pixel sequences of thermal images extended in the time dimension for all abnormal observation points in the target electrolytic cell to obtain multiple time-temperature feature sequences.

[0032] Furthermore, step 4 includes:

[0033] For each spatial temperature sequence among the multiple spatial temperature sequences, calculate the similarity between the spatial temperature sequences;

[0034] For each time-temperature sequence among the multiple time-temperature sequences, calculate the similarity between the time-temperature sequences;

[0035] Set the number of classes, construct a hierarchical clustering tree, and cluster the abnormal observation points into multiple groups;

[0036] For each group among the multiple groups, determine whether the electrode plate corresponding to the observation points within the group is abnormal according to the temperature mean value of the abnormal observation points within the group and the number of abnormal observation points;

[0037] When the number of abnormal observation points within multiple groups is less than 1 / p of the total number of electrode plates, calculate the temperature mean value of all abnormal observation points within each group to obtain the temperature mean value corresponding to each group, and arrange the calculated multiple temperature mean values in descending order. Take the group corresponding to the first temperature mean value among the arranged multiple temperature mean values as the abnormal working condition group, determine that the electrode plate corresponding to the area where the abnormal observation points within the abnormal working condition group are located is abnormal, take the group corresponding to the second temperature mean value among the arranged multiple temperature mean values as the suspected abnormal working condition group, and determine that the electrode plate corresponding to the area where the abnormal observation points within the suspected abnormal working condition group are located is suspected to be abnormal.

[0038] Furthermore, the similarity between the spatial temperature sequences calculated using the Manhattan distance is:

[0039] D cityblock =||abs(A - B)||

[0040] Where D cityblock represents the Manhattan distance, A and B respectively represent any two spatial temperature sequences, and the function abs(·) takes the absolute value of each element in the vector;

[0041] The similarity between the time-temperature sequences calculated using the Jaccard distance is:

[0042] D Jaccard =AB T / (||AA T || + ||BB T || - AB T )

[0043] Among them, D Jaccard represents the Jaccard distance, and A and B respectively represent any two time-temperature sequences.

[0044] The above solution of the present invention has the following beneficial effects:

[0045] The present invention obtains the historical infrared thermal images of the target electrolytic cell, and determines the characteristic auxiliary line perpendicular to the plate of the target electrolytic cell in the historical infrared thermal images; obtains the temperature peak points on the characteristic auxiliary line to obtain a plurality of abnormal observation points; for each abnormal observation point among the plurality of abnormal observation points, extracts the spatial temperature characteristic sequence in the spatial dimension and the time-temperature characteristic sequence in the time dimension according to the abnormal observation point; respectively performs hierarchical clustering on the plurality of abnormal observation points according to the spatial temperature characteristic sequence and the time-temperature characteristic sequence to obtain a clustering result, and identifies the plates in abnormal working conditions according to the clustering result; comprehensively combines the temperature distribution characteristics in both the time and space dimensions, and uses an unsupervised clustering method to identify abnormal electrolysis working conditions, avoiding the inaccuracy caused by artificially extracting temperature representative values and setting temperature thresholds when judging the working conditions of the plates through temperature thresholds, thereby improving the accuracy and precision of intelligent identification of abnormal electrolysis working conditions; provides guidance for non-ferrous metal smelting enterprises to timely and accurately detect abnormal electrolysis working conditions, assists in improving product quality, reduces production energy consumption, and improves enterprise production efficiency.

[0046] Other beneficial effects of the present invention will be described in detail in the subsequent specific implementation section. Brief Description of the Drawings

[0047] Figure 1 is a flowchart of an embodiment of the present invention;

[0048] Figure 2 is a processing result diagram of distortion correction and image cropping of the infrared thermal image in an embodiment of the present invention; among them, (a) is the original infrared thermal image; (b) is the infrared thermal image after distortion correction; (c) is the full-tank infrared thermal image after cropping; (d) is the single-tank infrared thermal image after cropping.

[0049] Figure 3 In (a) is a visualization diagram of the temperature change matrix in the time dimension formed on the first characteristic auxiliary line in an embodiment of the present invention; (b) is a visualization diagram of the temperature change matrix in the time dimension formed on the second characteristic auxiliary line in an embodiment of the present invention; (c) is a visualization diagram of the temperature change matrix in the time dimension formed on the third characteristic auxiliary line in an embodiment of the present invention;

[0050] Figure 4Among them, (a) is the temperature distribution curve diagram on the feature auxiliary line in the embodiment of the present invention; (b) is the single-sided difference curve diagram of the temperature curve in the embodiment of the present invention. Specific embodiments

[0051] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of them. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present invention.

[0052] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention. In addition, the terms "first", "second", "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0053] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "installation", "connection", "connection" should be understood in a broad sense. For example, it can be a locking connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0054] In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0055] The present invention provides an abnormal electrolysis condition identification method for existing problems.

[0056] As Figure 1 shown, the embodiment of the present invention provides an abnormal electrolysis condition identification method, including:

[0057] Step 1, obtain historical infrared thermal images of the target electrolytic cell, and determine a feature auxiliary line perpendicular to the direction of the target electrolytic cell plate in the historical infrared thermal images;

[0058] Step 2, obtain the temperature peak points on the feature auxiliary line to obtain a plurality of abnormal observation points;

[0059] Step 3: For each of the multiple abnormal observation points, extract the spatial temperature feature sequence in the spatial dimension and the temporal temperature feature sequence in the temporal dimension according to the abnormal observation point;

[0060] Step 4: Perform hierarchical clustering on the multiple abnormal observation points respectively according to the spatial temperature feature sequence and the temporal temperature feature sequence to obtain the clustering result, and identify the electrode plates in the abnormal working condition according to the clustering result.

[0061] In step 1 of the embodiment of the present invention, by obtaining multiple historical infrared thermal images within a certain time range from the current moment to the past moment, batch preprocessing such as distortion correction and image cropping is performed on the thermal images, and a characteristic auxiliary line perpendicular to the direction of the target electrolytic cell electrode plate is determined in the historical infrared thermal images, and a temperature change matrix in the temporal dimension is constructed for subsequent processing of the electrode plate temperature data, which specifically includes:

[0062] Continuously collect N infrared thermal images with a period of T to obtain the temperature field state and its change of the large-area electrolytic cell;

[0063] As Figure 2 shown, perform batch distortion correction and single-cell cropping on the N infrared thermal images to facilitate subsequent extraction of the electrode plate spatial temperature sequence;

[0064] For each of the N infrared thermal images, determine K (K≥3) characteristic auxiliary lines perpendicular to the direction of the target electrolytic cell electrode plate, which are dispersedly located in the three regions A, B, and C of a single electrolytic cell, and are used to assist in subsequent extraction of the temporal and spatial temperature sequences; in the embodiment of the present invention, K = 3, the first characteristic auxiliary line is located in the A region of the single electrolytic cell, that is, at the 1 / 4 position of the single electrolytic cell image along the electrode plate direction, the second characteristic auxiliary line is located in the B region of the single electrolytic cell, that is, at the 2 / 4 position of the single electrolytic cell image along the electrode plate direction, and the third characteristic auxiliary line is located in the C region of the single electrolytic cell, that is, at the 3 / 4 position of the single electrolytic cell image along the electrode plate direction;

[0065] For each of the multiple characteristic auxiliary lines, extract its temperature change in the temporal dimension according to the characteristic auxiliary line and form a temperature change matrix, as Figure 3 shown, for facilitating subsequent extraction of the temporal temperature sequence.

[0066] Specifically, step 2 includes:

[0067] For each of the multiple characteristic auxiliary lines, form a temperature sequence with the points on the characteristic auxiliary line;

[0068] Perform one-sided difference processing on the temperature sequence to obtain a temperature value difference sequence;

[0069] Search for the points where the temperature value difference sequence crosses the line y = 0 from positive to negative, and locate them as the peak points on the temperature, obtaining multiple abnormal observation points. As Figure 4 shown, the searched observation points are marked with "*" on the curve.

[0070] In the embodiment of the present invention, the temperature sequence T k is:

[0071] T k ={t k (i)|i∈R, i∈[1, N pixel_k}

[0072] Perform one-sided difference processing on the temperature sequence T k to obtain the temperature value difference sequence T' k as:

[0073] T′ k ={t′ k (i)|i∈N, i∈[1, N pixel_k}

[0074] where, t k (i) represents the infrared measurement temperature value at the i-th pixel position on the k-th characteristic auxiliary line, t' k (1)=t' k (2), R represents the total number of pixel positions, N pixel_k represents the pixel length value of the k-th characteristic auxiliary line, which is equal to the pixel width of the single-slot image perpendicular to the plate direction after cropping, t' k (i)=t k (i)-t k (i - 1), (i∈N, i∈[2, N pixel_k ), N represents the total number of pixel positions after one-sided difference;

[0075] Search for the points where the temperature value difference sequence crosses the line y = 0 from positive to negative in the temperature value difference sequence, and obtain the set I high ={i|t' k (i)<0, t' k (i - 1)>0, i∈[2, N pixel_k} to obtain multiple abnormal observation points.

[0076] The embodiment of the present invention obtains the set I of lower peak points low ={i|t' k (i)>0, t' k (i - 1)<0, i∈[2, N pixel_k}. Among the upper peak point sets, search for the first points with temperatures greater than \(t_h\) starting from both ends of the upper peak point set. The two ends of the set refer to the first upper peak point and the last upper peak point in the upper peak point set. The search ends until the temperature of the upper peak point is less than the preset temperature threshold \(t_h\), and the first upper peak point \(I\) is obtained. high_1 and the second upper peak point wherein, the first upper peak point \(I\) high_1 is the first upper peak point with a temperature greater than the preset temperature threshold obtained by searching from the head end of the upper peak point set, and the second upper peak point is the first upper peak point with a temperature greater than the preset temperature threshold obtained by searching from the tail end of the upper peak point set. Between the first upper peak point \(I\) high_1 and the second upper peak point search for the first lower peak point \(I\) that is less than and closest to the first upper peak point \(I\) high_1 and the second lower peak point \(I\) that is greater than and closest to \(I\) low_1 high_2 and the second lower peak point \(I\) low_2 .

[0077] Crop the historical infrared thermal image and the temperature change matrix according to the upper and lower peak points of the temperature and the preset temperature threshold \(t_h\), and retain the image area between the first lower peak point \(I\) low_1 and the second lower peak point \(I\) low_2 , and remove the two end regions.

[0078] wherein, the preset temperature threshold \(t_h\in(0, 255]\). Since different non-ferrous metals use different smelting temperatures, the specific setting value is determined according to the actual application scenario.

[0079] Specifically, step 3 includes:

[0080] Extract the pixel sequences of the thermal images extended in the spatial dimension of all abnormal observation points in the target electrolytic cell to obtain multiple spatial temperature feature sequences \(\{t\) spc (i)|i\in N^+, i\in[1, N obv \}\), wherein, \(t\) spc (i) represents the time-temperature sequence obtained at the \(i\)-th abnormal observation point, and \(N\) obv represents the number of abnormal observation points;

[0081] Extract the pixel sequences of the thermal images extended in the time dimension of all abnormal observation points in the target electrolytic cell to obtain multiple time-temperature feature sequences \(\{t\) tim (i)|i\in N^+, i\in[1, N obv \}\), wherein, \(t\) tim (i) represents the time-temperature sequence obtained at the \(i\)-th abnormal observation point, and \(N\) obv ​Indicates the number of abnormal observation points.

[0082] Step 4 in the embodiment of the present invention, according to the extracted spatial temperature feature sequence and time temperature feature sequence, performs hierarchical clustering on multiple abnormal observation points to obtain a clustering result, and identifies the electrode plates in abnormal working conditions according to the clustering result; specifically as follows:

[0083] The spatial temperature feature sequence t is normalized by the following formula spc and the time temperature feature sequence t tim as follows:

[0084]

[0085] where mean(·) calculates the average value of the sequence elements, and std(·) calculates the standard deviation of the sequence elements;

[0086] For each spatial temperature sequence among multiple spatial temperature sequences, the Manhattan distance (City Block Distance) is used to calculate the similarity between the spatial temperature sequences as:

[0087] D cityblock = ||abs(A - B)||

[0088] where D cityblock represents the Manhattan distance, A and B respectively represent any two spatial temperature sequences, and the function abs(·) takes the absolute value of each element in the vector;

[0089] For each time temperature sequence among multiple time temperature sequences, the Jaccard distance (Jaccard Distance) is used to calculate the similarity between the time temperature sequences;

[0090] D Jaccard = AB T / (||AA T || + ||BB T || - AB T )

[0091] where D Jaccard represents the Jaccard distance, A and B respectively represent any two time temperature sequences, and T is the transpose symbol;

[0092] Set the number of categories C (C ≥ 3), construct a hierarchical clustering tree based on the hierarchical clustering method, and cluster the abnormal observation points into C ethnic groups; in the embodiment of the present invention, the centroid method is used to form the hierarchical clustering tree to realize the clustering of the abnormal observation points and obtain the clustering result;

[0093] For each of multiple ethnic groups, determine whether the electrode plates corresponding to the observation points within the ethnic group are abnormal according to the average temperature of the abnormal observation points within the ethnic group and the number of abnormal observation points.

[0094] When the number of abnormal observation points in multiple ethnic groups is less than 1 / p of the total number of electrode plates, where p ≤ 4, calculate the average temperature of all abnormal observation points within each ethnic group to obtain the average temperature corresponding to each ethnic group, and arrange the calculated multiple average temperatures in descending order. The ethnic group corresponding to the first average temperature in the arranged multiple average temperatures is regarded as the abnormal working condition ethnic group, and the electrode plates corresponding to the areas where the abnormal observation points in this abnormal working condition ethnic group are located are determined to be abnormal. The ethnic group corresponding to the second average temperature in the arranged multiple average temperatures is regarded as the suspected abnormal working condition ethnic group, and the electrode plates corresponding to the areas where the abnormal observation points in this suspected abnormal working condition ethnic group are located are determined to be suspected abnormal. In the embodiment of the present invention, p = 2 is selected, and it is considered based on experience that the number of normal electrode plates in the electrolytic cell is more than half.

[0095] The embodiment of the present invention can also increase the clustering category C to achieve a more refined classification of abnormal working conditions.

[0096] The embodiment of the present invention can also distinguish abnormal working condition situations such as inter-pole short circuits and abnormal electrode plate overlaps through the area where the corresponding feature auxiliary line of the detected abnormal electrode plate is located. If the corresponding feature auxiliary line is located in the middle area of the electrode plate, it is an inter-pole short circuit situation. If an abnormality is detected only on the feature extraction auxiliary line located at one end of the electrode plate, it is an abnormal electrode plate overlap situation.

[0097] The embodiment of the present invention obtains the historical infrared thermal image of the target electrolytic cell and determines the feature auxiliary line perpendicular to the direction of the electrode plates of the target electrolytic cell in the historical infrared thermal image; obtains the temperature peak points on the feature auxiliary line to obtain multiple abnormal observation points; for each abnormal observation point among the multiple abnormal observation points, extracts the spatial temperature feature sequence in the spatial dimension and the time temperature feature sequence in the time dimension according to the abnormal observation point; performs hierarchical clustering on the multiple abnormal observation points respectively according to the spatial temperature feature sequence and the time temperature feature sequence to obtain the clustering result, and identifies the electrode plates in the abnormal working condition according to the clustering result; comprehensively integrates the temperature distribution characteristics in both the time and space dimensions, and uses an unsupervised clustering method to identify abnormal electrolytic working conditions, avoiding the inaccuracy caused by artificially extracting the temperature representative value and setting the temperature threshold when judging the working condition of the electrode plate through the temperature threshold, thereby realizing a more accurate and refined intelligent identification of abnormal electrolytic working conditions; provides guidance for non-ferrous metal smelting enterprises to timely and accurately detect abnormal electrolytic working conditions, assists in improving product quality, reducing production energy consumption, and improving the production efficiency of the enterprise.

[0098] The above are the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention.

Claims

1. A method for identifying abnormal electrolysis conditions, characterized in that, Including: Step 1: Obtain historical infrared thermal images of the target electrolytic cell, and determine characteristic auxiliary lines perpendicular to the plate direction of the target electrolytic cell in the historical infrared thermal images; Step 2: Obtain temperature peak points on the characteristic auxiliary lines to obtain a plurality of abnormal observation points; Step 3: For each abnormal observation point among the plurality of abnormal observation points, extract a spatial temperature feature sequence in the spatial dimension and a temporal temperature feature sequence in the temporal dimension according to the abnormal observation point; Step 4: Perform hierarchical clustering on the plurality of abnormal observation points respectively according to the spatial temperature feature sequence and the temporal temperature feature sequence to obtain a clustering result, and identify the plates in abnormal working conditions according to the clustering result, including: For each spatial temperature sequence among the plurality of spatial temperature sequences, calculate the similarity between the spatial temperature sequences; For each temporal temperature sequence among the plurality of temporal temperature sequences, calculate the similarity between the temporal temperature sequences; Set the number of categories, construct a systematic clustering tree, and cluster the abnormal observation points into multiple groups; For each group among the plurality of groups, determine whether the plate corresponding to the observation points in the group is abnormal according to the temperature mean value of the abnormal observation points in the group and the number of the abnormal observation points; When the number of abnormal observation points in each of the plurality of groups is less than 1 / p of the total number of plates, calculate the temperature mean value of all abnormal observation points in each group to obtain the temperature mean value corresponding to each group, and arrange the calculated plurality of temperature mean values in descending order, and use the group corresponding to the first temperature mean value in the arranged plurality of temperature mean values as the abnormal working condition group, and determine the plate corresponding to the area where the abnormal observation points in the abnormal working condition group are located as abnormal, and use the group corresponding to the second temperature mean value in the arranged plurality of temperature mean values as the suspected abnormal working condition group, and determine the plate corresponding to the area where the abnormal observation points in the suspected abnormal working condition group are located as suspected abnormal.

2. The abnormal electrolysis condition recognition method according to claim 1, wherein The said Step 1 includes: Continuously collect N infrared thermal images with a period of T; For each of the N infrared thermal images, determine a plurality of characteristic auxiliary lines perpendicular to the plate direction of the target electrolytic cell in the infrared thermal image; For each of the plurality of characteristic auxiliary lines, extract its temperature change in the temporal dimension according to the characteristic auxiliary line and form a temperature change matrix.

3. The abnormal electrolysis condition identification method according to claim 2, wherein, The said Step 2 includes: For each of the plurality of characteristic auxiliary lines, form a temperature sequence with the points on the characteristic auxiliary line; Perform unilateral difference processing on the temperature sequence to obtain a temperature value difference sequence; Search for the points that cross the straight line y = 0 from positive to negative on the temperature value difference sequence to obtain temperature peak points and obtain a plurality of abnormal observation points.

4. The abnormal electrolytic working condition identification method according to claim 3, characterized in that The temperature sequence T k is as follows: T k = {t k (i) | i ∈ R, i ∈ [1, N pixel_k} Perform a unilateral difference operation on the temperature sequence T k to obtain the temperature value difference sequence T' k as follows: T′ k ={t′ k (i)|i∈N,i∈[1,N pixel_k} where t k (i) represents the infrared measured temperature value at the i-th pixel position on the k-th feature auxiliary line, R represents the total number of pixel positions, N pixel_k represents the pixel length value of the k-th feature auxiliary line, which is equal to the pixel width of the single-slot image perpendicular to the plate direction after cropping, t′ k (i) = t k (i) - t k (i - 1), (i ∈ N, i ∈ [2, N pixel_k ), N represents the total number of pixel positions after unilateral differencing; Search for the points where the temperature value difference sequence crosses the line y = 0 from positive to negative, and obtain the set I of temperature upper peak points high ={i|t′ k (i)<0,t′ k (i - 1)>0,i∈[2,N pixel_k}, and obtain multiple abnormal observation points.

5. The abnormal electrolysis condition identification method according to claim 4, wherein Before Step 3, it further includes: Obtain the set I of the lower peak points low ={i|t′ k (i)>0, t′ k (i - 1)<0, i ∈ [2, N pixel_k}; Among the upper peak point set, search for the first batch of upper peak points with temperatures greater than the preset temperature threshold starting from both ends of the upper peak point set to obtain the first upper peak point I high_1 and the second upper peak point I high_2 , where the first upper peak point I high_1 is the first upper peak point with a temperature greater than the preset temperature threshold searched from the head end of the upper peak point set, and the second upper peak point I high_2 is the first upper peak point with a temperature greater than the preset temperature threshold searched from the tail end of the upper peak point set, I high_1 <I high_2 ; At the first upper peak point I high_1 and the second upper peak point I high_2 search for the first lower peak point I high_1 that is less than and closest to the first upper peak point I low_1 and the second lower peak point I high_2 that is greater than and closest to I low_2 ; The historical infrared thermal image and the temperature change matrix are cropped to retain the first lower peak point I low_1 and the second lower peak point I low_2 The image area between.

6. The abnormal electrolysis condition identification method according to claim 5, wherein The said Step 3 includes: Extract the pixel sequences of the thermal images extended in the spatial dimension for all abnormal observation points in the target electrolytic cell to obtain multiple spatial temperature feature sequences; Extract the pixel sequences of the thermal images extended in the time dimension for all abnormal observation points in the target electrolytic cell to obtain multiple temporal temperature feature sequences.

7. The abnormal electrolysis condition identification method according to claim 6, characterized in that Including: The similarity between the spatial temperature sequences is calculated using the Manhattan distance as: D cityblock = ||abs(A - B)|| Among them, D cityblock represents the Manhattan distance, A and B respectively represent any two spatial temperature sequences, and the function abs(·) takes the absolute value of each element in the vector; The similarity between the temporal temperature sequences is calculated using the Jaccard distance as: D Jaccard = AB T / (‖AA T ‖ + ‖BB T ‖ - AB T ) Among them, D Jaccard represents the Jaccard distance, and A and B respectively represent any two time-temperature sequences.

Images