An intelligent control method and system for mineral processing

Abstract

The present invention relates to the field of equipment control, and in particular to an intelligent control method and system for a mineral processing process. An intelligent control system for a mineral processing process comprises: a data acquisition module, a control strategy adjustment input data construction module, a first screening module, a second screening module, an equipment control module, a target recovery rate prediction model update module, an equipment control parameter output model adjustment module, and a target equipment control parameter output model update module. The present invention queries the direction of equipment parameter adjustment from a standard database through the feature vector corresponding to the ore image, and adaptively adjusts the parameters of the equipment control parameter output model based on the direction of equipment parameter adjustment, so that the equipment control parameter set output by the equipment control parameter output model after the adaptive parameter adjustment can better conform to the ore conditions, thereby improving the ore recovery rate during mineral processing.

Description

Technical Field

[0001] The present invention relates to the field of equipment control, and in particular to an intelligent control method and system for a mineral processing process. Background Art

[0002] In existing mineral processing processes, the control of key links such as crushing, grinding, and flotation equipment usually relies on manual adjustment, which has many disadvantages. First, because the properties of the ore and external conditions may change at any time, manual adjustment is difficult to respond in a timely manner and often has a lag, affecting production efficiency and product quality. Second, manual adjustment relies on the experience and judgment of the operator, and the adjustment accuracy is difficult to guarantee. Different personnel may have large differences in the operating results of the same equipment, especially in the process stage where parameters fluctuate frequently. In addition, operators need to continuously monitor and frequently adjust the equipment, which is labor-intensive. Long hours of work can easily lead to fatigue and operational errors, further reducing production stability. At the same time, in the traditional manual adjustment process, production data is not fully utilized, and there is a lack of effective data analysis and feedback mechanisms, making it difficult to achieve optimal control of equipment operation. Summary of the Invention

[0003] The present invention queries the direction of equipment parameter adjustment from a standard database through the feature vector corresponding to the ore image, and adaptively adjusts the parameters of the equipment control parameter output model based on the direction of the equipment parameter adjustment, so that the equipment control parameter set output by the equipment control parameter output model after the adaptive parameter adjustment can be more in line with the ore conditions, thereby improving the ore recovery rate during ore dressing; and the adjustment of the equipment control parameter output model is based on the target equipment control parameter output model, and the target equipment control parameter output model is only adjusted at a specific time point, which can avoid the occurrence of ores with excessively large grade differences in the same batch, which affects the equipment parameter adjustment of other copper ores during ore dressing.

[0004] An intelligent control method for a mineral processing process, comprising:

[0005] Obtain ore images and equipment control parameter sets at monitoring time points;

[0006] The ore image is sent to the ore image feature extraction model for processing, and the ore image feature vector is output. The ore image feature vector and the equipment control parameter set are spliced ​​together to build a control strategy to adjust the input data;

[0007] The control strategy adjustment input data is sent to the target device control parameter output model for processing, and the target device control parameter set is output; the standard ore image feature vectors are selected one by one from the standard database, and the standard database includes the standard ore image feature vector, the standard device control parameter set and the optimal recovery rate; for each selected standard ore image feature vector, the similarity between the standard ore image feature vector and the ore image feature vector is calculated, and it is determined whether the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold; if the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold, the standard device control parameter set corresponding to the standard ore image feature vector is added to the to-be-selected set; if the similarity between the standard ore image feature vector and the ore image feature vector is not higher than the similarity threshold, no operation is performed;

[0008] Traversing the candidate set, for each standard equipment control parameter set selected in the candidate set, concatenating the ore image feature vector with the standard equipment control parameter set to construct recovery rate prediction data, then feeding the recovery rate prediction data into the target recovery rate prediction model for processing, and outputting the target predicted recovery rate; until all standard equipment control parameter sets in the candidate set are selected, recording the highest target predicted recovery rate as the marked recovery rate, and recording the standard equipment control parameter set corresponding to the marked recovery rate as the marked equipment control parameter set; and also including updating the target recovery rate prediction model;

[0009] Calculate the real-time loss value based on the target device control parameter set and the marked device control parameter set, and adjust the parameters of the device control parameter output model through the back propagation algorithm based on the real-time loss value. Send the control strategy adjustment input data to the device control parameter output model after parameter adjustment for processing, and output the predicted device control parameter set. Use the predicted device control parameter set to control the equipment in the mineral processing operation;

[0010] At the target model update time point, the target device control parameter output model is directly replaced by the device control parameter output model.

[0011] Preferably, updating the target recovery rate prediction model specifically includes the following steps:

[0012] At the monitoring time point, the ore image feature vector is spliced ​​with the target equipment control parameter set and then fed into the recovery rate prediction model for processing, outputting the predicted recovery rate; the recovery rate loss value is calculated based on the marked recovery rate and the predicted recovery rate, and the parameters of the recovery rate prediction model are adjusted based on the recovery rate loss value through the back propagation algorithm;

[0013] At the target model update time point, the target recovery rate prediction model is directly replaced by the recovery rate prediction model.

[0014] Preferably, the construction of the standard database specifically includes the following methods:

[0015] Acquire several ore images, process each ore image through an ore image feature extraction model, and output a corresponding ore image feature vector; perform cluster analysis on all ore image feature vectors to construct several clusters, and select N ore image feature vectors from each cluster as standard ore image feature vectors;

[0016] For each standard ore image feature vector, the corresponding standard equipment control parameter set and optimal recovery rate are obtained through a swarm optimization algorithm; and the standard ore image feature vector, standard equipment control parameter set and optimal recovery rate are mapped one by one and stored in a standard database.

[0017] Preferably, a swarm optimization algorithm is used to obtain a set of standard equipment control parameters and an optimal recovery rate corresponding to a standard ore image feature vector, specifically comprising the following steps:

[0018] Construct several device control parameter simulation individuals, and form a population set of all device control parameter simulation individuals, and set the maximum number of iterations;

[0019] Calculate the fitness of the device control parameter simulation individual. The specific calculation method is to splice the standard ore image feature vector and the device control parameter simulation individual, and then send them into the target recovery rate prediction model for processing, output the target predicted recovery rate, and record the target predicted recovery rate as the fitness of the device control parameter simulation individual.

[0020] The control parameters of the equipment in the population set are simulated and updated iteratively through the group optimization algorithm;

[0021] Until the number of iterations reaches the maximum number of iterations, the device control parameter simulation individual with the largest fitness is recorded as the standard device control parameter set, and the fitness corresponding to the standard device control parameter set is recorded as the optimal recovery rate.

[0022] Preferably, training the device control parameter output model specifically includes the following steps:

[0023] Acquire several equipment control parameter training samples, which include control strategy adjustment input data and its corresponding predicted equipment control parameter set; combine all equipment control parameter training samples into a first training set, and send the first training set to a parameter-initialized equipment control parameter output model for training, using the control strategy adjustment input data as the input of the equipment control parameter output model and the predicted equipment control parameter set as the target, calculate a first loss value, and determine whether the first loss value is within a first preset range. If the first loss value is within the first preset range, output the trained equipment control parameter output model; otherwise, continue to train the equipment control parameter output model using the first training set.

[0024] Preferably, training the recovery rate prediction model specifically includes the following steps:

[0025] A plurality of recovery prediction training samples are obtained, the recovery prediction training samples including ore image feature vectors and equipment control parameter sets, and the recovery prediction training samples are labeled by recovery rate; all labeled recovery prediction training samples are combined into a second training set, and the second training set is sent to a parameter-initialized recovery prediction model for training, with the recovery rate as a target, a second loss value is calculated, and it is determined whether the second loss value is within a second preset range; if the second loss value is within the second preset range, the trained recovery prediction model is output; otherwise, the recovery prediction model is continued to be trained using the second training set.

[0026] Preferably, the similarity calculation adopts a cosine similarity algorithm.

[0027] The present invention also provides an intelligent control system for a mineral processing process, comprising:

[0028] Data acquisition module, used to obtain ore images and equipment control parameter sets at the monitoring time point;

[0029] A control strategy adjustment input data construction module is used to send the ore image to the ore image feature extraction model for processing, output the ore image feature vector, splice the ore image feature vector and the equipment control parameter set, and construct the control strategy adjustment input data;

[0030] a first screening module, configured to input control strategy adjustment input data into a target device control parameter output model for processing, and output a target device control parameter set; select standard ore image feature vectors one by one from a standard database, wherein the standard database includes standard ore image feature vectors, standard device control parameter sets, and optimal recovery rates; calculate, for each selected standard ore image feature vector, the similarity between the standard ore image feature vector and the ore image feature vector; and determine whether the similarity between the standard ore image feature vector and the ore image feature vector is higher than a similarity threshold; if the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold, add the standard device control parameter set corresponding to the standard ore image feature vector to a to-be-selected set; and if the similarity between the standard ore image feature vector and the ore image feature vector is not higher than the similarity threshold, perform no operation;

[0031] The second screening module is used to traverse the candidate set, and for each standard equipment control parameter set selected in the candidate set, concatenate the ore image feature vector with the standard equipment control parameter set to construct recovery rate prediction data, and then feed the recovery rate prediction data into the target recovery rate prediction model for processing, and output the target predicted recovery rate; until all standard equipment control parameter sets in the candidate set are selected, the highest target predicted recovery rate is recorded as the marked recovery rate, and the standard equipment control parameter set corresponding to the marked recovery rate is recorded as the marked equipment control parameter set;

[0032] The equipment control module is used to input the control strategy adjustment input data into the equipment control parameter output model after parameter adjustment for processing, output the predicted equipment control parameter set, and control the equipment in the mineral processing operation through the predicted equipment control parameter set;

[0033] A target recovery rate prediction model updating module is used to update the target recovery rate prediction model;

[0034] The device control parameter output model adjustment module calculates the real-time loss value based on the target device control parameter set and the marked device control parameter set, and adjusts the parameters of the device control parameter output model through the back propagation algorithm based on the real-time loss value;

[0035] The target device control parameter output model updating module is used to update the target device control parameter output model.

[0036] The present invention has the following advantages:

[0037] The present invention queries the direction of equipment parameter adjustment from a standard database through the feature vector corresponding to the ore image, and adaptively adjusts the parameters of the equipment control parameter output model based on the direction of the equipment parameter adjustment, so that the equipment control parameter set output by the equipment control parameter output model after the adaptive parameter adjustment can be more in line with the ore conditions, thereby improving the ore recovery rate during ore dressing; and the adjustment of the equipment control parameter output model is based on the target equipment control parameter output model, and the target equipment control parameter output model is only adjusted at a specific time point, which can avoid the occurrence of ores with excessively large grade differences in the same batch, which affects the equipment parameter adjustment of other copper ores during ore dressing. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic diagram of the structure of the intelligent control system for the mineral processing process adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0040] Example 1, an intelligent control method for a mineral processing process, comprising:

[0041] At the monitoring time point, ore images and equipment control parameter sets are acquired. The ore images are images taken of the ore during the beneficiation operation. The equipment control parameter sets include the operating parameters of various equipment in the beneficiation process, such as the speed of the grinding mill, the amount of reagent added, and the stirring speed of the flotation cell. These equipment control parameters will affect the recovery rate of the beneficiation operation. The recovery rate of the beneficiation operation refers to the proportion of valuable metals extracted from the original ore.

[0042] The ore image is sent to the ore image feature extraction model for processing. The ore image feature extraction model is established based on the pre-trained YoLov7 model. The pre-training process uses the ImageNet dataset to output the ore image feature vector. The output of the second-to-last layer of the pre-trained YoLov7 model is selected as the ore image feature vector. The ore image feature vector can represent the grade of the ore. For example, for the copper ore targeted in this embodiment, the ore image feature vector represents the copper content in the copper ore. The ore image feature vector and the equipment control parameter set are spliced ​​to construct a control strategy to adjust the input data.

[0043] The control strategy adjustment input data is sent to the target device control parameter output model for processing, and the target device control parameter set is output; the standard ore image feature vectors are selected one by one from the standard database, and the standard database includes the standard ore image feature vector, the standard device control parameter set and the optimal recovery rate. For each selected standard ore image feature vector, the similarity between the standard ore image feature vector and the ore image feature vector is calculated, and it is determined whether the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold. The similarity calculation adopts the cosine similarity algorithm. The similarity threshold is determined by the developer and is generally set to 0.85. If the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold, it means that the standard ore image feature vector and the ore image feature vector are relatively similar, and the corresponding standard device control parameter set can provide a reference for the intelligent control of the equipment. The standard device control parameter set corresponding to the standard ore image feature vector is added to the to-be-selected set; if the similarity between the standard ore image feature vector and the ore image feature vector is not higher than the similarity threshold, no operation is performed;

[0044] In particular, if the similarity between the standard ore image feature vector and the ore image feature vector is higher than the extreme similarity value, the standard equipment control parameter set corresponding to the standard ore image feature vector can be directly used for intelligent control. The extreme similarity value is set by the developer and is generally 0.95;

[0045] Traversing the candidate set, for each standard equipment control parameter set selected in the candidate set, concatenating the ore image feature vector with the standard equipment control parameter set to construct recovery rate prediction data, then feeding the recovery rate prediction data into the target recovery rate prediction model for processing, and outputting the target predicted recovery rate; until all standard equipment control parameter sets in the candidate set are selected, the highest target predicted recovery rate is recorded as the marked recovery rate, and the standard equipment control parameter set corresponding to the marked recovery rate is recorded as the marked equipment control parameter set, where the marked recovery rate represents the direction in which the equipment control parameter set should be adjusted with the assistance of the standard equipment control parameter set; and also includes updating the target recovery rate prediction model;

[0046] The real-time loss value is calculated based on the target device control parameter set and the marked device control parameter set, and the parameters of the device control parameter output model are adjusted through the back propagation algorithm based on the real-time loss value. The control strategy adjustment input data is sent to the device control parameter output model after parameter adjustment for processing, and the predicted device control parameter set is output. The equipment in the mineral processing operation is controlled by the predicted device control parameter set. The specific control is achieved through the PID algorithm;

[0047] At the target model update time point, the target device control parameter output model is directly replaced by the device control parameter output model. That is, whenever the target model is updated, the target device control parameter output model is adjusted, and remains unchanged at other times. This ensures that when the equipment parameters of the same batch of copper ore are adaptively adjusted, the adjustment will not be too large, avoiding the occurrence of ores with too large grade differences in the same batch, which will affect the equipment parameter adjustment of other copper ores during mineral processing;

[0048] The present application queries the direction of equipment parameter adjustment from a standard database through the feature vector corresponding to the ore image, and adaptively adjusts the parameters of the equipment control parameter output model based on the direction of the equipment parameter adjustment, so that the equipment control parameter set output by the equipment control parameter output model after the adaptive adjustment of the parameters can be more in line with the ore conditions, thereby improving the recovery rate of the ore during mineral processing; and the adjustment of the equipment control parameter output model is based on the target equipment control parameter output model, and the target equipment control parameter output model will only be adjusted at a specific time point, which can avoid the occurrence of ores with excessively large grade differences in the same batch, affecting the equipment parameter adjustment of other copper mines during mineral processing.

[0049] The target recovery rate prediction model is updated, specifically including the following steps:

[0050] At the monitoring time point, the ore image feature vector is spliced ​​with the target equipment control parameter set and then fed into the recovery rate prediction model for processing, outputting the predicted recovery rate; the recovery rate loss value is calculated based on the marked recovery rate and the predicted recovery rate, and the parameters of the recovery rate prediction model are adjusted based on the recovery rate loss value through the back propagation algorithm;

[0051] At the target model update time point, the target recovery rate prediction model is directly replaced by the recovery rate prediction model.

[0052] The construction of the standard database includes the following methods:

[0053] Acquire several ore images, process each ore image through an ore image feature extraction model, and output a corresponding ore image feature vector; perform cluster analysis on all ore image feature vectors to construct several clusters, and select N ore image feature vectors from each cluster as standard ore image feature vectors, where N is generally 3. Through cluster analysis, representative ore images from the acquired ore images can be queried, thereby reducing the number of standard ore image feature vectors in the standard database and facilitating query;

[0054] For each standard ore image feature vector, the corresponding standard equipment control parameter set and optimal recovery rate are obtained through a swarm optimization algorithm; and the standard ore image feature vector, standard equipment control parameter set and optimal recovery rate are mapped one by one and stored in a standard database.

[0055] The standard equipment control parameter set and optimal recovery rate corresponding to the standard ore image feature vector are obtained through the swarm optimization algorithm, which specifically includes the following steps:

[0056] Construct several equipment control parameter simulation individuals, and group all of them into a population set, setting the maximum number of iterations. It should be noted that the storage format of the equipment control parameter simulation individuals is consistent with that of the equipment control parameter set, and also stores the operating parameters of various equipment in the mineral processing process, and each operating parameter is a random number within the corresponding range.

[0057] Calculate the fitness of the equipment control parameter simulation individual. The specific calculation method is to splice the standard ore image feature vector and the equipment control parameter simulation individual and then send them into the target recovery rate prediction model for processing, output the target predicted recovery rate, and record the target predicted recovery rate as the fitness of the equipment control parameter simulation individual. The target predicted recovery rate can represent the recovery rate of the operating parameters corresponding to the equipment control parameter simulation individual for the ore corresponding to the ore image when the beneficiation is performed. The higher the recovery rate, the more the equipment control parameter simulation individual fits the optimal solution.

[0058] The device control parameter simulation individuals in the population set are iteratively updated through a swarm optimization algorithm. In this embodiment, the swarm optimization algorithm adopts a particle swarm optimization algorithm, and the iterative process can be used to find the optimal solution.

[0059] Until the number of iterations reaches the maximum number of iterations, the device control parameter simulation individual with the largest fitness is recorded as the standard device control parameter set, and the fitness corresponding to the standard device control parameter set is recorded as the optimal recovery rate.

[0060] Training the device control parameter output model includes the following steps:

[0061] A number of equipment control parameter training samples are obtained, and the equipment control parameter training samples include control strategy adjustment input data and its corresponding predicted equipment control parameter set. It should be noted that the corresponding predicted equipment control parameter set is an equipment control parameter set that is manually adjusted according to the actual situation corresponding to the control strategy adjustment input data. Adjustment here means that a certain recovery rate target can be achieved, specifically realizing actual mineral processing operations, or computer simulation. All equipment control parameter training samples are combined into a first training set, and the first training set is sent to a parameter-initialized equipment control parameter output model for training, using the control strategy adjustment input data as the input of the equipment control parameter output model, using the predicted equipment control parameter set as the target, calculating a first loss value, and judging whether the first loss value is within a first preset range. The first preset range is determined by the developer. If the first loss value is within the first preset range, output the trained equipment control parameter output model; otherwise, continue to train the equipment control parameter output model through the first training set.

[0062] Training the recovery rate prediction model includes the following steps:

[0063] Acquire several recovery rate prediction training samples, which include ore image feature vectors and equipment control parameter sets, and annotate the recovery rate prediction training samples with recovery rates. It should be noted that the ore image feature vectors and equipment control parameter sets in the recovery rate prediction training samples are actual data, and the annotated recovery rates are also actually recorded manually. All annotated recovery rate prediction training samples are combined into a second training set, and the second training set is sent to a parameter-initialized recovery rate prediction model for training. With the recovery rate as the target, a second loss value is calculated, and it is determined whether the second loss value is within a second preset range. The second preset range is also determined by the developer. If the second loss value is within the second preset range, the trained recovery rate prediction model is output; otherwise, the recovery rate prediction model is continued to be trained with the second training set.

[0064] Example 2, an intelligent control system for a mineral processing process, such as Figure 1 Shown, including:

[0065] The data acquisition module is used to acquire ore images and equipment control parameter sets at the monitoring time point. The ore images are images of the ore taken during the beneficiation operation. The equipment control parameter sets include the operating parameters of various equipment in the beneficiation process, such as the speed of the grinding mill, the amount of reagent added, and the stirring speed of the flotation tank. These equipment control parameters will affect the recovery rate of the beneficiation operation. The recovery rate of the beneficiation operation refers to the proportion of valuable metals extracted from the original ore.

[0066] A control strategy adjustment input data construction module is used to feed the ore image into the ore image feature extraction model for processing. The ore image feature extraction model is established based on the pre-trained YoLov7 model. The pre-training process uses the ImageNet dataset to output the ore image feature vector. The output of the penultimate layer of the pre-trained YoLov7 model is selected as the ore image feature vector. The ore image feature vector can represent the grade of the ore. For example, for the copper ore targeted in this embodiment, the ore image feature vector represents the copper content in the copper ore. The ore image feature vector and the equipment control parameter set are spliced ​​to construct the control strategy adjustment input data.

[0067] The first screening module is used to input the control strategy adjustment input data into the target device control parameter output model for processing and output the target device control parameter set; select standard ore image feature vectors one by one from the standard database, and the standard database includes standard ore image feature vectors, standard device control parameter sets and optimal recovery rates; for each selected standard ore image feature vector, calculate the similarity between the standard ore image feature vector and the ore image feature vector, and determine whether the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold; the similarity calculation adopts the cosine similarity algorithm; the similarity threshold is determined by the developer and is generally set to 0.85; if the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold, it means that the standard ore image feature vector and the ore image feature vector are relatively similar, and the corresponding standard device control parameter set can provide a reference for the intelligent control of the equipment, and the standard device control parameter set corresponding to the standard ore image feature vector is added to the to-be-selected set; if the similarity between the standard ore image feature vector and the ore image feature vector is not higher than the similarity threshold, no operation is performed;

[0068] The second screening module is used to traverse the candidate set, and for each standard equipment control parameter set selected in the candidate set, concatenate the ore image feature vector with the standard equipment control parameter set to construct recovery rate prediction data, and then send the recovery rate prediction data to the target recovery rate prediction model for processing, and output the target predicted recovery rate; until all standard equipment control parameter sets in the candidate set are selected, the highest target predicted recovery rate is recorded as the marked recovery rate, and the standard equipment control parameter set corresponding to the marked recovery rate is recorded as the marked equipment control parameter set. The marked recovery rate represents the direction in which the equipment control parameter set should be adjusted with the assistance of the standard equipment control parameter set;

[0069] The equipment control module is used to input the control strategy adjustment input data into the equipment control parameter output model after parameter adjustment for processing, output the predicted equipment control parameter set, and control the equipment in the mineral processing operation through the predicted equipment control parameter set. The specific control is achieved through the PID algorithm;

[0070] A target recovery rate prediction model updating module is used to update the target recovery rate prediction model;

[0071] The device control parameter output model adjustment module calculates the real-time loss value based on the target device control parameter set and the marked device control parameter set, and adjusts the parameters of the device control parameter output model through the back propagation algorithm based on the real-time loss value;

[0072] The target device control parameter output model updating module is used to update the target device control parameter output model.

[0073] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail is prior art known to those skilled in the art.

Claims

1. An intelligent control method for a mineral processing process, characterized in that: include: Obtain ore images and equipment control parameter sets at monitoring time points; The ore image is sent to the ore image feature extraction model for processing, and the ore image feature vector is output. The ore image feature vector and the equipment control parameter set are spliced ​​together to build a control strategy to adjust the input data; The control strategy adjustment input data is sent to the target device control parameter output model for processing, and the target device control parameter set is output; the standard ore image feature vectors are selected one by one from the standard database, and the standard database includes the standard ore image feature vector, the standard device control parameter set and the optimal recovery rate; for each selected standard ore image feature vector, the similarity between the standard ore image feature vector and the ore image feature vector is calculated, and it is determined whether the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold; if the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold, the standard device control parameter set corresponding to the standard ore image feature vector is added to the to-be-selected set; if the similarity between the standard ore image feature vector and the ore image feature vector is not higher than the similarity threshold, no operation is performed; Traversing the candidate set, for each standard equipment control parameter set selected in the candidate set, concatenating the ore image feature vector with the standard equipment control parameter set to construct recovery rate prediction data, then feeding the recovery rate prediction data into the target recovery rate prediction model for processing, and outputting the target predicted recovery rate; until all standard equipment control parameter sets in the candidate set are selected, recording the highest target predicted recovery rate as the marked recovery rate, and recording the standard equipment control parameter set corresponding to the marked recovery rate as the marked equipment control parameter set; and also including updating the target recovery rate prediction model; Calculate the real-time loss value based on the target device control parameter set and the marked device control parameter set, and adjust the parameters of the device control parameter output model through the back propagation algorithm based on the real-time loss value. Send the control strategy adjustment input data to the device control parameter output model after parameter adjustment for processing, and output the predicted device control parameter set. Use the predicted device control parameter set to control the equipment in the mineral processing operation; At the target model update time point, the target device control parameter output model is directly replaced by the device control parameter output model.

2. The intelligent control method for a mineral processing process according to claim 1, characterized in that: The target recovery rate prediction model is updated, specifically including the following steps: At the monitoring time point, the ore image feature vector is spliced ​​with the target equipment control parameter set and then fed into the recovery rate prediction model for processing, outputting the predicted recovery rate; the recovery rate loss value is calculated based on the marked recovery rate and the predicted recovery rate, and the parameters of the recovery rate prediction model are adjusted based on the recovery rate loss value through the back propagation algorithm; At the target model update time point, the target recovery rate prediction model is directly replaced by the recovery rate prediction model.

3. The intelligent control method for a mineral processing process according to claim 2, characterized in that: The construction of the standard database includes the following methods: Acquire several ore images, process each ore image through an ore image feature extraction model, and output a corresponding ore image feature vector; perform cluster analysis on all ore image feature vectors to construct several clusters, and select N ore image feature vectors from each cluster as standard ore image feature vectors; For each standard ore image feature vector, the corresponding standard equipment control parameter set and optimal recovery rate are obtained through a swarm optimization algorithm; and the standard ore image feature vector, standard equipment control parameter set and optimal recovery rate are mapped one by one and stored in a standard database.

4. The intelligent control method for a mineral processing process according to claim 3, characterized in that: The standard equipment control parameter set and optimal recovery rate corresponding to the standard ore image feature vector are obtained through the swarm optimization algorithm, which specifically includes the following steps: Construct several device control parameter simulation individuals, and form a population set of all device control parameter simulation individuals, and set the maximum number of iterations; Calculate the fitness of the device control parameter simulation individual. The specific calculation method is to splice the standard ore image feature vector and the device control parameter simulation individual, and then send them into the target recovery rate prediction model for processing, output the target predicted recovery rate, and record the target predicted recovery rate as the fitness of the device control parameter simulation individual. The control parameters of the equipment in the population set are simulated and updated iteratively through the group optimization algorithm; Until the number of iterations reaches the maximum number of iterations, the device control parameter simulation individual with the largest fitness is recorded as the standard device control parameter set, and the fitness corresponding to the standard device control parameter set is recorded as the optimal recovery rate.

5. The intelligent control method for mineral processing according to claim 4, characterized in that: Training the device control parameter output model includes the following steps: Acquire several equipment control parameter training samples, which include control strategy adjustment input data and its corresponding predicted equipment control parameter set; combine all equipment control parameter training samples into a first training set, and send the first training set to a parameter-initialized equipment control parameter output model for training, using the control strategy adjustment input data as the input of the equipment control parameter output model and the predicted equipment control parameter set as the target, calculate a first loss value, and determine whether the first loss value is within a first preset range. If the first loss value is within the first preset range, output the trained equipment control parameter output model; otherwise, continue to train the equipment control parameter output model using the first training set.

6. The intelligent control method for mineral processing according to claim 5, characterized in that: Training the recovery rate prediction model includes the following steps: A plurality of recovery prediction training samples are obtained, the recovery prediction training samples including ore image feature vectors and equipment control parameter sets, and the recovery prediction training samples are labeled by recovery rate; all labeled recovery prediction training samples are combined into a second training set, and the second training set is sent to a parameter-initialized recovery prediction model for training, with the recovery rate as a target, a second loss value is calculated, and it is determined whether the second loss value is within a second preset range; if the second loss value is within the second preset range, the trained recovery prediction model is output; otherwise, the recovery prediction model is continued to be trained using the second training set.

7. The intelligent control method for mineral processing according to claim 6, characterized in that: The similarity calculation uses the cosine similarity algorithm.

8. An intelligent control system for a mineral processing process, characterized in that: The system applies an intelligent control method for a mineral processing process according to any one of claims 1 to 7, comprising: Data acquisition module, used to obtain ore images and equipment control parameter sets at the monitoring time point; A control strategy adjustment input data construction module is used to send the ore image to the ore image feature extraction model for processing, output the ore image feature vector, splice the ore image feature vector and the equipment control parameter set, and construct the control strategy adjustment input data; a first screening module, configured to input control strategy adjustment input data into a target device control parameter output model for processing, and output a target device control parameter set; select standard ore image feature vectors one by one from a standard database, wherein the standard database includes standard ore image feature vectors, standard device control parameter sets, and optimal recovery rates; calculate, for each selected standard ore image feature vector, the similarity between the standard ore image feature vector and the ore image feature vector; and determine whether the similarity between the standard ore image feature vector and the ore image feature vector is higher than a similarity threshold; if the similarity between the standard ore image feature vector and the ore image feature vector is higher than the similarity threshold, add the standard device control parameter set corresponding to the standard ore image feature vector to a to-be-selected set; and if the similarity between the standard ore image feature vector and the ore image feature vector is not higher than the similarity threshold, perform no operation; The second screening module is used to traverse the candidate set, and for each standard equipment control parameter set selected in the candidate set, concatenate the ore image feature vector with the standard equipment control parameter set to construct recovery rate prediction data, and then feed the recovery rate prediction data into the target recovery rate prediction model for processing, and output the target predicted recovery rate; until all standard equipment control parameter sets in the candidate set are selected, the highest target predicted recovery rate is recorded as the marked recovery rate, and the standard equipment control parameter set corresponding to the marked recovery rate is recorded as the marked equipment control parameter set; The equipment control module is used to input the control strategy adjustment input data into the equipment control parameter output model after parameter adjustment for processing, output the predicted equipment control parameter set, and control the equipment in the mineral processing operation through the predicted equipment control parameter set; A target recovery rate prediction model updating module is used to update the target recovery rate prediction model; The device control parameter output model adjustment module calculates the real-time loss value based on the target device control parameter set and the marked device control parameter set, and adjusts the parameters of the device control parameter output model through the back propagation algorithm based on the real-time loss value; The target device control parameter output model updating module is used to update the target device control parameter output model.

Images