Automatic batching control method and system for PVC pipeline production

Abstract

The invention provides an automatic batching control method and system for PVC pipeline production, and relates to the technical field of production control, and the method comprises the following steps: obtaining the addition amount of raw materials, and starting a feeding channel, a vibrator and an air pump; weighing the weight of the material in the hopper; training the feed quantity prediction model of the final state of the previous control period to obtain a feed quantity prediction model of the intermediate state of the current control period, and predicting the first power of a vibrator and the second power of an air pump; and closing the feeding channel, training the feeding quantity prediction model of the intermediate state of the current control period, and obtaining the feeding quantity prediction model of the final state of the current control period. According to the invention, when the feeding amount is controlled, the influence of the weight of the material in the hopper on the feeding speed is considered, so that proper power is set for the vibrator and the air pump, the feeding amount prediction model can be continuously updated, the power accuracy of the vibrator and the air pump is improved, and the control precision of the feeding amount of the raw material is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of production control, and in particular to an automatic batching control method and system for PVC pipe production. Background Art

[0002] In related technologies, raw materials for PVC pipe production are mixed using devices such as hoppers, and the raw materials are drawn from the hoppers by equipment such as air pumps. However, this technology has difficulties in precisely controlling the power of these devices, as well as the feed rate, making it difficult to improve the quality of PVC pipes.

[0003] The information disclosed in the background technology section of this application is only intended to deepen the understanding of the general background technology of this application, and should not be regarded as an admission or any form of suggestion that the information constitutes the prior art already known to those skilled in the art. Summary of the Invention

[0004] The present invention provides an automatic batching control method and system for PVC pipe production, which can solve the technical problem in related technologies that it is difficult to accurately control the power of an air pump and thus difficult to accurately control the feed amount.

[0005] According to a first aspect of the present invention, a method for automatic batching control of PVC pipe production is provided, comprising: obtaining, at the start of a current control cycle, the addition amounts of a plurality of raw materials in the current control cycle; opening the feed channels corresponding to the plurality of raw materials, and opening the vibrators and air pumps of each feed channel; weighing the weight of the materials in the hopper corresponding to each raw material at a first preset number of moments before the current control cycle; training a feed amount prediction model of the final state of the previous control cycle based on the rated power of the vibrators and air pumps and the weight of the materials, and obtaining a feed amount prediction model of the intermediate state of the current control cycle; and training a feed amount prediction model of the intermediate state of the current control cycle based on the feed amount prediction model of the intermediate state of the current control cycle after the first preset number of moments in the current control cycle. The weight of the material in the hopper at each moment, and the feed quantity prediction model of the intermediate state of the current control cycle are used to determine the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments in the current control cycle; when the weight reduction of the raw material in the hopper reaches the added amount during the current control cycle, or at the end of the current control cycle, the feed channel is closed; at the end of the current control cycle, the feed quantity prediction model of the intermediate state of the current control cycle is trained according to the weight of the material in the hopper at the first preset number of moments and each moment thereafter, as well as the first power and the second power, to obtain the feed quantity prediction model of the final state of the current control cycle.

[0006] According to the present invention, based on the rated power of the vibrator and the air pump and the material weight, the feed quantity prediction model in the final state of the previous control cycle is trained to obtain the feed quantity prediction model in the intermediate state of the current control cycle, including: inputting the training input vector at the i-th moment, which is composed of the rated power of the vibrator, the rated power of the air pump, and the material weight at the i-th moment, and the training feature vector at the (i - 1)-th moment, into the i-th feature extraction layer of the feed quantity prediction model in the final state of the previous control cycle to obtain the training feature vector at the i-th moment, where i is a positive integer less than the first preset quantity, and when i = 1, the training feature vector at the (i - 1)-th moment is a zero vector; inputting the training feature vector at the i-th moment into the first fully connected layer and the first activation layer of the feed quantity prediction model in the final state of the previous control cycle for processing to obtain the predicted feed quantity between the i-th moment and the (i + 1)-th moment; determining the first loss function of the feed quantity prediction model according to the predicted feed quantity between the i-th moment and the (i + 1)-th moment, the material weight at the i-th moment, and the material weight at the (i + 1)-th moment; training the feed quantity prediction model in the final state of the previous control cycle according to the first loss function to obtain the feed quantity prediction model in the intermediate state of the current control cycle.

[0007] According to the present invention, determining the first loss function of the feed quantity prediction model according to the predicted feed quantity between the i-th moment and the (i + 1)-th moment, the material weight at the i-th moment, and the material weight at the (i + 1)-th moment includes: according to the formula , determining the first loss function of the feed quantity prediction model , where is the predicted feed quantity between the i-th moment and the (i + 1)-th moment, is the material weight at the i-th moment, is the material weight at the (i + 1)-th moment, is the first preset quantity, and both i and are positive integers.

[0008] According to the present invention, based on the material weight in the hopper at each moment after the first preset quantity of moments in the current control cycle and the feed quantity prediction model in the intermediate state of the current control cycle, determining the first power of the vibrator and the second power of the air pump at each moment after the first preset quantity of moments in the current control cycle includes: combining the material weight in the hopper from the 1st moment to the -th moment, the rated power of the vibrator, and the rated power of the air pump to form the first state vector from the 1st moment to the -th moment, and respectively inputting it into each feature extraction layer of the feed quantity prediction model in the intermediate state of the current control cycle to obtain the first feature vector at the -th moment; obtaining the the number of remaining time periods between a moment and the end moment of the current control period; obtain the weight of the added material before the th moment, and obtain a first difference between the weight of the added material and the addition amount of the raw material; form a second state vector of the th moment by combining the number of remaining time periods and the first difference, and process the second state vector through a second fully connected layer to obtain a second feature vector of the th moment; splice the second feature vector of the th moment and the first feature vector of the th moment to obtain a third feature vector of the th moment; input the third feature vector of the th moment into a third fully connected layer and a second activation layer to obtain a first power of the vibrator and a second power of the air pump at the th moment; combine the material weight in the hopper from the (1 + j)th moment to the th moment, the rated power of the vibrator from the (1 + j)th moment to the th moment, the first power of the vibrator from the th moment to the th moment, the rated power of the air pump from the (1 + j)th moment to the th moment, and the second power of the air pump from the th moment to the th moment to form a first state vector from the (1 + j)th moment to the th moment, and input them into each feature extraction level of the feed amount prediction model in the intermediate state of the current control period to obtain a first feature vector of the th moment, where j is an integer greater than or equal to 0, and ​​​​​​​​The first power of the vibrator and the second power of the air pump at a certain moment.

[0009] According to the present invention, the method further includes: inputting the first eigenvector at the th moment into the first fully connected layer and the first activation layer for processing to obtain the predicted feed rate between the th moment and the th moment; according to the material weight in the hopper at the th moment and the predicted feed rate between the th moment and the th moment, obtaining the predicted material weight within the th moment; according to the predicted material weight within the th moment, the first power of the vibrator, and the second power of the air pump, obtaining the first predicted eigenvector at the th moment; according to the predicted feed rate between the th moment and the th moment, obtaining the predicted added material weight before the th moment, and obtaining the first predicted difference between the predicted added material weight and the added amount of the raw material; forming the second predicted state vector at the th moment by combining the number of remaining time periods between the th moment and the end moment of the control period with the first predicted difference, and processing the second state vector through the second fully connected layer to obtain the second predicted eigenvector at the th moment; splicing the first predicted eigenvector at the th moment and the second predicted eigenvector at the th moment to obtain the third predicted eigenvector at the th moment; inputting the third predicted eigenvector at the th moment into the third fully connected layer and the second activation layer to obtain the first predicted power of the vibrator and the second predicted power of the air pump at the th moment; iteratively executing the steps of obtaining the first predicted power and the second predicted power to obtain the first predicted power of the vibrator and the second predicted power of the air pump from the th moment to the Nth moment, and the predicted feed rate between adjacent moments; determining the second loss function corresponding to the th moment according to the predicted feed rate between each adjacent moment in the time period from the th moment to the Nth moment; training the second fully connected layer and the third fully connected layer according to the second loss functions corresponding to multiple moments.

[0010] According to the present invention, according to the For each adjacent moment within the time period from the first moment to the Nth moment, determine the predicted feed rate between each adjacent moment, and determine the second loss function corresponding to the th moment, including: according to the formula , determine the second loss function corresponding to the th moment , where is the predicted feed rate between the kth moment and the k + 1th moment, , is the feed rate between the sth moment and the s + 1th moment, , is the added amount of raw material, and both s and k are positive integers.

[0011] According to the present invention, at the end of the current control cycle, based on the material weights in the hopper at each moment after the first preset number of moments and the first power and the second power, train the feed rate prediction model for the intermediate state of the current control cycle to obtain the feed rate prediction model for the final state of the current control cycle, including: based on the material weights in the hopper at each moment after the first preset number of moments, the first power and the second power, and the feed rate prediction model for the intermediate state of the current control cycle, determine the predicted feed rate between each moment after the first preset number of moments; according to the formula , obtain the third loss function of the feed rate prediction model for the intermediate state of the current control cycle , where is the predicted feed rate between the hth moment and the h + 1th moment, is the material weight at the hth moment, is the material weight at the h + 1th moment, , N is the serial number of the end moment of the current control cycle, is the first preset number, and h is a positive integer; according to the third loss function, train the feed rate prediction model for the intermediate state of the current control cycle to obtain the feed rate prediction model for the final state of the current control cycle.

[0012] According to a second aspect of the present invention, there is provided an automatic batching control system for PVC pipe production, comprising: an addition amount module, configured to obtain the addition amounts of various raw materials within the current control period at the start of the current control period; an opening module, configured to open the feeding channels corresponding to the various raw materials, and to turn on the vibrators and air pumps of each feeding channel; a weighing module, configured to weigh the material weights in the hoppers corresponding to the respective raw materials at the first preset number of moments before the current control period; a first training module, configured to train the feeding amount prediction model in the final state of the previous control period according to the rated powers of the vibrators and air pumps and the material weights, to obtain the feeding amount prediction model in the intermediate state of the current control period; a determination module, configured to determine the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments of the current control period according to the material weights in the hoppers at each moment after the first preset number of moments of the current control period and the feeding amount prediction model in the intermediate state of the current control period; a closing module, configured to close the feeding channel when the weight reduction amount in the hopper of the raw material within the current control period reaches the addition amount, or at the end of the current control period; a second training module, configured to train the feeding amount prediction model in the intermediate state of the current control period according to the material weights in the hoppers at the first preset number of moments and each moment after that, and the first power and the second power, to obtain the feeding amount prediction model in the final state of the current control period.

[0013] Technical effects: According to the present invention, when controlling the feeding amount, the influence of the material weight in the hopper on the feeding speed can be considered, so as to set appropriate powers for the vibrator and the air pump. Also, based on data such as the power and material weight obtained within the control period, the feeding amount prediction model can be continuously updated, thereby improving the accuracy of the powers of the vibrator and the air pump, improving the accuracy of the feeding amount prediction model, so as to improve the control accuracy of the feeding amount of raw materials and improve the quality of PVC pipes. When training the feeding amount prediction model in the final state of the previous control period, data at multiple past moments can be fused to accurately predict the predicted feeding amount between the i-th moment and the (i + 1)-th moment. And when setting the first loss function, weights are set according to the number of moments, so as to accurately reflect the influence of the accuracy of the predicted data at this moment on the subsequent prediction accuracy, improve the training accuracy and training efficiency, and improve the accuracy of the feeding amount prediction model. When training the second fully connected layer and the third fully connected layer, after each moment ends, the first predicted power, the second predicted power and the predicted feeding amount at subsequent moments can be solved iteratively, so as to determine the total predicted feeding amount within the control period through the predicted feeding amount, and then reduce the error between the total predicted feeding amount and the added amount of raw materials by setting the second loss function, so as to train the second fully connected layer and the third fully connected layer, thereby continuously improving the accuracy of the second fully connected layer and the third fully connected layer, as well as the prediction accuracy of the first power of the vibrator and the second power of the air pump, and reducing the error between the total predicted feeding amount and the added amount of raw materials. When training the feeding amount prediction model in the intermediate state of the current control period, not only the data within the first preset number of moments is used to train the feeding amount prediction model, but also after the control period ends, the data at multiple moments after the first preset number of moments can be used to continue training the feeding amount prediction model, so as to maximize the use of the measured and predicted data within the control period, improve the training efficiency of the feeding amount prediction model, improve the prediction accuracy, and provide an accurate data basis for predicting the first power of the vibrator and the second power of the air pump.

[0014] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and do not limit the present invention. According to the following detailed description of exemplary embodiments with reference to the accompanying drawings, other features and aspects of the present invention will be clearer. Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other embodiments can be obtained based on these drawings; Figure 1Exemplarily shown is a flowchart of an automatic batching control method for PVC pipe production according to an embodiment of the present invention; Figure 2 Exemplarily shown is a schematic diagram of a feed quantity prediction model according to an embodiment of the present invention; Figure 3 Exemplarily shown is an automatic batching control system for PVC pipe production according to an embodiment of the present invention. Detailed implementation manners

[0016] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are only a part rather than all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0017] The technical solutions of the present invention will be described in detail below with specific embodiments. These specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0018] Figure 1 Exemplarily shown is a flowchart of an automatic batching control method for PVC pipe production according to an embodiment of the present invention. The method includes: Step S1, at the start moment of the current control cycle, obtain the addition amounts of various raw materials within the current control cycle; Step S2, open the feeding channels corresponding to the various raw materials, and turn on the vibrators and air pumps of each feeding channel; Step S3, at the first preset number of moments before the current control cycle, weigh the material weights in the hoppers corresponding to the respective raw materials; Step S4, according to the rated powers of the vibrators and air pumps and the material weights, train the feed quantity prediction model in the final state of the previous control cycle to obtain the feed quantity prediction model in the intermediate state of the current control cycle; Step S5, according to the material weights in the hoppers at each moment after the first preset number of moments of the current control cycle, and the feed quantity prediction model in the intermediate state of the current control cycle, determine the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments of the current control cycle; Step S6, when the weight reduction amount in the hopper of the raw material within the current control cycle reaches the addition amount, or when the current control cycle ends, close the feeding channel; Step S7, at the end of the current control cycle, according to the material weights in the hoppers at the first preset number of moments and each moment after that, and the first power and the second power, train the feed quantity prediction model in the intermediate state of the current control cycle to obtain the feed quantity prediction model in the final state of the current control cycle.

[0019] The automatic batching control method for PVC pipe production according to an embodiment of the present invention can, when controlling the feeding amount, consider the influence of the material weight in the hopper on the feeding speed, so as to set appropriate powers for the vibrator and the air pump. It can also continuously update the feeding amount prediction model based on data such as the power and material weight obtained within the control period, thereby improving the accuracy of the powers of the vibrator and the air pump, improving the accuracy of the feeding amount prediction model, improving the control accuracy of the feeding amount of raw materials, and improving the quality of PVC pipes.

[0020] According to an embodiment of the present invention, when producing PVC pipes, a variety of raw materials are required. For example, raw materials such as basic resin, heat stabilizer, and lubricant can be placed in different hoppers respectively, and each hopper can be connected to a feeding channel, and each hopper can weigh the weight of the remaining raw materials in the hopper.

[0021] According to an embodiment of the present invention, in step S1, since the addition sequences and addition amounts of various raw materials are different, multiple control periods can be set, and the addition amounts within each control period can be determined respectively according to the addition sequence of the raw materials. For example, it is necessary to first add raw material A and raw material B, with the addition amounts being 10 kg and 2 kg respectively. In the next process, it is necessary to add raw material C and raw material D, with the addition amounts being 5 kg and 1 kg respectively. Then, raw material A and raw material B can be added in the first control period (that is, the feeding channels connected to the hoppers of raw material A and raw material B are opened), and the addition amounts of raw material A and raw material B are set to 10 kg and 2 kg respectively. Raw material C and raw material D can be added in the second control period (that is, the feeding channels connected to the hoppers of raw material C and raw material D are opened), and the addition amounts of raw material C and raw material D are set to 5 kg and 1 kg respectively. In the example, the duration of each control period is 5 minutes, and each control period includes multiple moments, and the time interval between adjacent moments is 10 seconds. The present invention does not limit the specific values of the duration of the control period and the time interval between adjacent moments.

[0022] According to an embodiment of the present invention, in step S2, as described above, after determining the types and addition amounts of the raw materials to be added in the current control period, the feeding channels corresponding to the various raw materials can be opened. And since the raw materials are usually in powder form, the powdered raw materials may block the feeding channels due to the intermolecular forces between the particles. Therefore, devices such as vibrators and air pumps can be used to assist in feeding. The vibrator can be set on the hopper to vibrate the hopper to break the intermolecular forces between the particles, and help the powdered raw materials fall from the discharge port of the hopper into the feeding channel. The air pump is set in the feeding channel and can suck the raw materials in the feeding channel into the container, so as to be mixed in the container and then proceed to the next processing step.

[0023] According to an embodiment of the present invention, in step S3, after the start of the current control cycle, the vibrator and the air pump can be set to the rated power and maintained for a first preset number (e.g., 10) of moments, and the material weights in the hoppers corresponding to each raw material can be measured. Based on the material weights at each moment, the feed rate between each two moments can be determined, that is, the feed rate between two moments can be obtained by subtracting the material weights at the two moments. Moreover, the influence of the remaining raw material weight in the hopper on the feed rate can be determined when the powers of the vibrator and the air pump remain unchanged. For example, if there is more remaining raw material in the hopper, the pressure at the discharge port of the hopper is greater, which has a positive impact on the feed rate between two moments. That is, the more the remaining raw material in the hopper, the more the feed rate between two moments. Then, this rule can be summarized, and based on this rule, the powers of the vibrator and the air pump in subsequent moments can be adjusted. For example, if it is determined based on this rule that the feed rate between two moments is small and the feeding speed is slow, the powers of the vibrator and the air pump can be increased to improve the feeding speed.

[0024] According to an embodiment of the present invention, in step S4, using data such as the material weights and the rated power at the first preset number of moments at the beginning of each control cycle, the feed rate prediction model in the final state of the previous control cycle can be trained to obtain the feed rate prediction model in the intermediate state of the current control cycle. Using the feed rate prediction model in the intermediate state of the current control cycle, the first power of the vibrator and the second power of the air pump at the remaining moments of the current control cycle can be determined. Also, the material weight in the hopper at the remaining moments can be monitored at all times. At the end of the current control cycle, using the first power, the second power, and the material weight at the remaining moments of the current control cycle to train the feed rate prediction model in the intermediate state of the current control cycle, the feed rate prediction model in the final state of the current control cycle can be obtained. And so on, the above steps can be iteratively executed in the next control cycle. And if the current control cycle is the first control cycle, a pre-trained feed rate prediction model (a feed rate prediction model trained using the data in a control cycle dedicated for training) or an initial state feed rate prediction model (e.g., a model with randomly initialized parameters) can be used.

[0025] Figure 2 Exemplarily shown is a schematic diagram of the feed rate prediction model according to an embodiment of the present invention.

[0026] According to an embodiment of the present invention, in step S4, based on the rated power of the vibrator and the air pump and the material weight, the feed rate prediction model in the final state of the previous control cycle is trained to obtain the feed rate prediction model in the intermediate state of the current control cycle, including: inputting the training input vector at the i-th moment, which is composed of the rated power of the vibrator, the rated power of the air pump, and the material weight at the i-th moment, and the training feature vector at the (i - 1)-th moment, into the i-th feature extraction layer of the feed rate prediction model in the final state of the previous control cycle to obtain the training feature vector at the i-th moment, where i is a positive integer less than the first preset number, and when i = 1, the training feature vector at the (i - 1)-th moment is a zero vector; inputting the training feature vector at the i-th moment into the first fully connected layer and the first activation layer of the feed rate prediction model in the final state of the previous control cycle for processing to obtain the predicted feed rate between the i-th moment and the (i + 1)-th moment; determining the first loss function of the feed rate prediction model according to the predicted feed rate between the i-th moment and the (i + 1)-th moment, the material weight at the i-th moment, and the material weight at the (i + 1)-th moment; and training the feed rate prediction model in the final state of the previous control cycle according to the first loss function to obtain the feed rate prediction model in the intermediate state of the current control cycle.

[0027] According to an embodiment of the present invention, the feed rate prediction model may be a recurrent neural network model or an LSTM model, etc., and may include a plurality of processing blocks. Each processing block may include a feature extraction layer, which may take the current input vector and the feature vector output by the previous processing block as inputs and output the current feature vector. For example, the training input vector at the i-th moment composed of the rated power of the vibrator, the rated power of the air pump, and the material weight at the i-th moment, the output of the (i - 1)-th processing block is the training feature vector at the (i - 1)-th moment. The training input vector at the i-th moment and the training feature vector at the (i - 1)-th moment are used as the inputs of the i-th processing block. After being processed by the i-th block, the training feature vector at the i-th moment is obtained. If i = 1, the training feature vector at the (i - 1)-th moment is a zero vector, or it can also be a randomly initialized vector (i.e., each data in the vector is a random value). The processing block may be a reusable processing block. For example, the feed rate prediction model may only include 1 processing block. When receiving the training input vector at the 1st moment and the training feature vector at the 0th moment (such as a zero vector), this processing block may be used as the 1st processing block (whose feature extraction layer is the 1st feature extraction layer). When receiving the training input vector at the 2nd moment and the training feature vector at the 1st moment, this processing block may be used as the 2nd processing block... When receiving the training input vector at the i-th moment and the training feature vector at the (i - 1)-th moment, this processing block may be used as the i-th processing block. When inputting the training input vector at each moment, it may be directly input to the feature extraction layer of the processing block, or first input to the fully connected layer to obtain a higher-dimensional vector, so as to describe more comprehensive feature information, and then input to the feature extraction layer.

[0028] According to an embodiment of the present invention, based on the above feed rate prediction model, comprehensive processing can be performed based on the training input vector at the current moment and the training feature vector output at the previous moment to obtain the training feature vector output at the current moment. There is a correlation between the feed rate and the material weight in the hopper, and it is also related to the power of the vibrator and the air pump. Therefore, when the power of the vibrator and the air pump remains unchanged, the feed rate within a period of time changes with the change of the material weight in the hopper. Therefore, the data of past moments can be used as a basis to predict the predicted feed rate for a future time period. For example, the training feature vector at the i-th moment is obtained based on the training input vector at the i-th moment and the training feature vector at the i-1-th moment, and the training feature vector at the i-1-th moment is obtained based on the training input vector at the i-1-th moment and the training feature vector at the i-2-th moment... Therefore, the training feature vector at the i-th moment integrates the data of multiple past moments. Therefore, the training feature vector at the i-th moment can be used to predict the feed rate between the i-th moment and the i + 1-th moment based on the data of multiple past moments. The training feature vector at the i-th moment is input into the first fully connected layer, and the output data of the first fully connected layer is input into the first activation layer, and processed through an activation function (for example, the RELU activation function) to obtain the predicted feed rate between the i-th moment and the i + 1-th moment. That is, the data of multiple past moments can represent how much the feed rate is within the time period between each moment when the vibrator and the air pump are at a certain power (for example, the rated power), so that based on this, the predicted feed rate between the i-th moment and the i + 1-th moment can be determined when the power of the vibrator and the air pump is the rated power.

[0029] According to an embodiment of the present invention, there may be an error in the predicted feed rate between the i-th moment and the i + 1-th moment. The accurate feed rate can be obtained through the material weight at the i-th moment and the material weight at the i + 1-th moment. That is, by subtracting the material weight at the i-th moment from the material weight at the i + 1-th moment, the accurate feed rate can be determined, and the error between the accurate feed rate and the predicted feed rate can be determined, thereby determining the first loss function and training the feed rate prediction model of the final state of the previous control cycle.

[0030] According to an embodiment of the present invention, based on the predicted feed rate between the i-th moment and the i + 1-th moment, the material weight at the i-th moment, and the material weight at the i + 1-th moment, determining the first loss function of the feed rate prediction model includes: determining the first loss function of the feed rate prediction model according to formula (1) , (1) Wherein, is the predicted feed rate between the i-th moment and the i + 1-th moment, is the material weight at the i-th moment, is the material weight at the (i + 1)-th moment, is the first preset quantity, and both i and are positive integers.

[0031] According to an embodiment of the present invention, in formula (1), is the accurate feed rate between the i-th moment and the (i + 1)-th moment. Therefore, is the error between the predicted feed rate and the accurate feed rate between the i-th moment and the (i + 1)-th moment. And, the larger the number of moments i, the closer the moment is to the end moment of the first preset number of moments, and the greater the impact of the accuracy of the predicted feed rate between this moment and the previous moment on the accuracy of subsequent predictions. Therefore, a higher weight can be set. For example, is used as the weight of the error of the predicted feed rate between the i-th moment and the previous moment. The errors of the predicted feed rates between each moment and the previous moment can be weighted and summed to obtain the first loss function of the feed rate prediction model.

[0032] According to an embodiment of the present invention, after obtaining the first loss function, the first loss function can be backpropagated, and the parameters of the feed rate prediction model in the final state of the previous control cycle can be adjusted by the gradient descent method for training to obtain the feed rate prediction model in the intermediate state of the current control cycle. Compared with the feed rate prediction model in the final state of the previous control cycle, the feed rate prediction model in the intermediate state of the current control cycle has more accurate prediction data for the feed rate between adjacent moments.

[0033] In this way, the data of multiple past moments can be fused to accurately predict the predicted feed rate between the i-th moment and the (i + 1)-th moment. And when setting the first loss function, weights are set according to the number of moments, so as to accurately reflect the impact of the accuracy of the prediction data at this moment on the accuracy of subsequent predictions, improve the training accuracy and training efficiency, and improve the accuracy of the feed rate prediction model.

[0034] According to an embodiment of the present invention, in step S5, the feed rate prediction model in the intermediate state of the current control cycle trained above can be used to determine the first power of the vibrator and the second power of the air pump at subsequent moments, so that the feed rate of the raw material can reach the predetermined addition amount within the current control cycle.

[0035] According to an embodiment of the present invention, in step S5, based on the material weights in the hopper at each moment after the first preset number of moments in the current control cycle and the feeding amount prediction model in the intermediate state of the current control cycle, determining the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments in the current control cycle includes: forming a first state vector from the material weight in the hopper from the 1st moment to the th moment, the rated power of the vibrator, and the rated power of the air pump, and respectively inputting it into each feature extraction layer of the feeding amount prediction model in the intermediate state of the current control cycle to obtain the first feature vector at the th moment; obtaining the number of remaining time periods between the th moment and the end moment of the current control cycle; obtaining the added material weight before the th moment, and obtaining the first difference between the added material weight and the addition amount of the raw material; forming a second state vector from the number of remaining time periods and the first difference, and processing the second state vector through a second fully connected layer to obtain the second feature vector at the th moment; splicing the second feature vector at the th moment and the first feature vector at the th moment to obtain the third feature vector at the th moment; inputting the third feature vector at the th moment into a third fully connected layer and a second activation layer to obtain the first power of the vibrator and the second power of the air pump at the th moment; forming a first state vector from the material weight in the hopper from the (1 + j)th moment to the th moment, the rated power of the vibrator from the (1 + j)th moment to the th moment, the first power of the vibrator from the th moment to the th moment, the rated power of the air pump from the (1 + j)th moment to the th moment, the second power of the air pump from the th moment to the th moment, and the second power of the air pump from the th moment to the th moment, and respectively inputting it into each feature extraction layer of the feeding amount prediction model in the intermediate state of the current control cycle to obtain the first feature vector at the th moment, where j is an integer greater than or equal to 0, and , N is the serial number of the end moment of the current control cycle, and N is a positive integer; obtaining the number of remaining time periods between the th moment and the end moment of the current control cycle; obtaining the th moment to the end moment of the current control cycle; obtaining the the weight of the added material before a certain moment, and obtain the first difference between the weight of the added material and the addition amount of the raw material; combine the number of remaining time periods with the first difference to form the second state vector at a certain moment, and process the second state vector through a second fully connected layer to obtain the second feature vector at a certain moment; concatenate the second feature vector at a certain moment and the first feature vector at a certain moment to obtain the third feature vector at a certain moment; input the third feature vector at a certain moment into a third fully connected layer and a second activation layer to obtain the first power of the vibrator and the second power of the air pump at a certain moment.

[0036] According to an embodiment of the present invention, the first state vectors from the 1st moment to the certain moment are obtained in a similar manner to the above training input vectors, that is, by combining the material weight in the hopper, the rated power of the vibrator, and the rated power of the air pump. However, due to certain differences in the parameters of the feed rate prediction model in the intermediate state of the current control cycle and the feed rate prediction model in the final state of the previous control cycle, the feature vectors output at each moment processed by the feed rate prediction model in the intermediate state of the current control cycle are different from the feature vectors output at each moment processed by the feed rate prediction model in the final state of the previous control cycle. The feature vectors output by the processing block can fuse the data from the 1st moment to the certain moment as the first feature vector at a certain moment, which can be used for subsequent prediction.

[0037] According to an embodiment of the present invention, the number of remaining time periods between a certain moment and the end moment of the current control cycle can be obtained. For example, if the serial number of the last moment of the current control cycle is N, then the number of remaining time periods between a certain moment and the end moment of the current control cycle is ; The weight of the added material before a certain moment can be obtained, that is, by summing the feed rates between every two adjacent moments before a certain moment to obtain the weight of the added material. The first difference between the weight of the added material and the addition amount of the raw material can also be obtained, that is, by subtracting the weight of the added material from the addition amount to obtain the first difference. ; The weight of the added material before a certain moment can be obtained, that is, by summing the feed rates between every two adjacent moments before a certain moment to obtain the weight of the added material. The first difference between the weight of the added material and the addition amount of the raw material can also be obtained, that is, by subtracting the weight of the added material from the addition amount to obtain the first difference. ; The weight of the added material before a certain moment can be obtained, that is, by summing the feed rates between every two adjacent moments before a certain moment to obtain the weight of the added material. The first difference between the weight of the added material and the addition amount of the raw material can also be obtained, that is, by subtracting the weight of the added material from the addition amount to obtain the first difference. ; The weight of the added material before a certain moment can be obtained, that is, by summing the feed rates between every two adjacent moments before a certain moment to obtain the weight of the added material. The first difference between the weight of the added material and the addition amount of the raw material can also be obtained, that is, by subtracting the weight of the added material from the addition amount to obtain the first difference. ; The weight of the added material before a certain moment can be obtained, that is, by summing the feed rates between every two adjacent moments before a certain moment to obtain the weight of the added material. The first difference between the weight of the added material and the addition amount of the raw material can also be obtained, that is, by subtracting the weight of the added material from the addition amount to obtain the first difference.

[0038] According to an embodiment of the present invention, combine the number of remaining time periods with the first difference to form the the second state vector at a moment, and the dimension of the second state vector can be increased through the second fully connected layer, so as to obtain a second feature vector with a higher dimension, so as to more comprehensively describe the state at this moment, that is, describe the remaining feeding demand and the remaining feeding duration, so as to provide a data basis for predicting the first power of the vibrator and the second power of the air pump at subsequent moments.

[0039] According to an embodiment of the present invention, the second feature vector at a moment and the first feature vector at a moment can be spliced to obtain a third feature vector at a moment, wherein the first feature vector at a moment can be used to describe the material weight in the hopper at the moment, and the predicted feeding amount in the next time period (that is, from the moment to the moment) when the power of the vibrator and the air pump is the rated power. The second feature vector is used to describe the remaining feeding demand and the remaining feeding duration. Therefore, the first feature vector can be used as a reference and the second feature vector can be used as a description of the feeding demand to determine how to set the power of the vibrator and the air pump in the subsequent time period so that the total feeding amount within the remaining duration reaches the feeding demand. Therefore, the third feature vector obtained by splicing the first feature vector and the second feature vector can be used for solving, and the third feature vector is processed through the third fully connected layer and the second activation layer to obtain the first power of the vibrator and the second power of the air pump at a moment. That is, the third fully connected layer and the second activation layer can determine the predicted feeding amount between the moment and the moment based on the first feature vector, so as to predict the material weight in the hopper at the

[0040] moment, and further predict how to set the first power of the vibrator and the second power of the air pump in the subsequent time period. According to an embodiment of the present invention, at the moment, the above-predicted first power of the vibrator and the second power of the air pump are used for setting, and the material weight in the hopper at the moment can be actually measured, so as to form the first state vector at a moment. The feeding amount prediction model may include processing blocks. Therefore, the first state vectors from the second moment to the moment can be sequentially input into each processing block, and finally the first feature vector at a moment is obtained. Generally, the The first eigenvector at a moment. Similarly to the above, the number of remaining time periods between the [n]th moment and the end moment of the current control period can be determined, as well as the weight of the added material before the [n]th moment and the first difference between the weight of the added material and the addition amount of the raw material. The number of remaining time periods and the first difference are combined to form the second state vector at the [n]th moment, and then the second eigenvector is obtained and concatenated with the first eigenvector to obtain the third eigenvector at the [n]th moment. Furthermore, the first power of the vibrator and the second power of the air pump at the [n]th moment are obtained through the third fully connected layer and the second activation layer. And so on, after each moment, the first power of the vibrator and the second power of the air pump at the next moment can be obtained, so as to obtain the first power of the vibrator and the second power of the air pump at each moment within the current control period, and make the total feed amount within the current control period reach the addition amount of the raw material. The number of remaining time periods between the [n]th moment and the end moment of the current control period, and the weight of the added material before the [n]th moment and the first difference between the weight of the added material and the addition amount of the raw material, and the number of remaining time periods and the first difference are combined to form the second state vector at the [n]th moment, and then the second eigenvector is obtained and concatenated with the first eigenvector to obtain the third eigenvector at the [n]th moment. Furthermore, the first power of the vibrator and the second power of the air pump at the [n]th moment are obtained through the third fully connected layer and the second activation layer. And so on, after each moment, the first power of the vibrator and the second power of the air pump at the next moment can be obtained, so as to obtain the first power of the vibrator and the second power of the air pump at each moment within the current control period, and make the total feed amount within the current control period reach the addition amount of the raw material.

[0041] According to an embodiment of the present invention, the above-mentioned second fully connected layer, third fully connected layer and second activation layer can be trained. The method further includes: inputting the first eigenvector at the [n]th moment into the first fully connected layer and the first activation layer for processing to obtain the predicted feed amount between the [n]th moment and the [m]th moment; obtaining the predicted material weight within the [n]th moment according to the material weight in the hopper at the [n]th moment and the predicted feed amount between the [m]th moment; obtaining the first predicted eigenvector at the [n]th moment according to the predicted material weight within the [n]th moment, the first power of the vibrator and the second power of the air pump; obtaining the predicted weight of the added material before the [n]th moment according to the predicted feed amount between the [n]th moment and the [m]th moment, and obtaining the first predicted difference between the predicted weight of the added material and the addition amount of the raw material; combining the number of remaining time periods between the [n]th moment and the end moment of the control period with the first predicted difference to form the second predicted state vector at the [n]th moment, and processing the second state vector through the second fully connected layer to obtain the second predicted eigenvector at the [n]th moment; concatenating the first predicted eigenvector at the [n]th moment with the second predicted eigenvector at the [n]th moment to obtain the predicted feed amount between the [n]th moment and the [m]th moment to obtain the predicted weight of the added material before the [n]th moment, and obtaining the first predicted difference between the predicted weight of the added material and the addition amount of the raw material; combining the number of remaining time periods between the [n]th moment and the end moment of the control period with the first predicted difference to form the second predicted state vector at the [n]th moment, and processing the second state vector through the second fully connected layer to obtain the second predicted eigenvector at the [n]th moment; concatenating the first predicted eigenvector at the [n]th moment with the second predicted eigenvector at the [n]th moment to obtain the third predicted eigenvector at the [n]th moment, and then The third predicted feature vector at a certain moment; input the third predicted feature vector at the th moment into the third fully connected layer and the second activation layer to obtain the first predicted power of the vibrator and the second predicted power of the air pump at the th moment; iteratively execute the steps of obtaining the first predicted power and the second predicted power to obtain the first predicted power of the vibrator and the second predicted power of the air pump from the th moment to the Nth moment, as well as the predicted feed quantity between adjacent moments; according to the predicted feed quantity between each adjacent moment in the time period from the th moment to the Nth moment, determine the second loss function corresponding to the th moment; train the second fully connected layer and the third fully connected layer according to the second loss functions corresponding to multiple moments.

[0042] According to an embodiment of the present invention, the above training steps can be performed after each moment, so as to gradually improve the accuracy of the second fully connected layer and the third fully connected layer, and also improve the accuracy of determining the first power of the vibrator and the second power of the air pump.

[0043] According to an embodiment of the present invention, by processing the first feature vector at the th moment through the first fully connected layer and the first activation layer, the predicted feed quantity between the th moment and the th moment can be obtained. Subtract the predicted feed quantity between the th moment and the th moment from the material weight in the hopper at the th moment to obtain the predicted material weight within the th moment, so that the predicted material weight within the th moment, the first power of the vibrator, and the second power of the air pump can be combined to obtain the first predicted feature vector at the th moment.

[0044] According to an embodiment of the present invention, add the total feed quantity before the th moment (the difference between the material weight in the hopper at the start moment of the control cycle and the material weight in the hopper at the th moment) to the predicted feed quantity between the th moment and the th moment to obtain the predicted added material weight before the th moment, so as to subtract the predicted added material weight from the added amount of the raw material to obtain the first predicted difference. Combine the number of remaining time periods between the th moment and the end moment of the control cycle with the first predicted difference to form the the second predicted state vector at a moment, and process the second state vector through a second fully connected layer to obtain the second predicted feature vector at the th moment, and then splice it with the first predicted feature vector at the th moment to obtain the third predicted feature vector at the th moment. Process the third predicted feature vector at the th moment through a third fully connected layer and a second activation layer to obtain the first predicted power of the vibrator and the second predicted power of the air pump at the th moment.

[0045] According to an embodiment of the present invention, the above process can be iteratively executed, so as to perform backward prediction. For example, input the first predicted feature vector at the th moment into the first fully connected layer and the first activation layer for processing to obtain the predicted feed quantity between the th moment and the th moment; subtract the predicted feed quantity between the th moment from the predicted material weight within the th moment to the th moment to obtain the predicted material weight within the th moment; compose the predicted material weight within the th moment, the first predicted power of the vibrator and the second predicted power of the air pump into the first predicted feature vector at the th moment; sum the total feed quantity before the th moment, the predicted feed quantity between the th moment and the th moment, and the predicted feed quantity between the th moment and the th moment to obtain the predicted added material weight before the th moment, and obtain the first predicted difference between the predicted added material weight and the added amount of the raw material; compose the number of remaining time periods between the th moment and the end moment of the control cycle and the first predicted difference into the second predicted state vector at the th moment, and process the second state vector through a second fully connected layer to obtain the second predicted feature vector at the th moment, and then splice it with the first predicted feature vector at the th moment to obtain the third predicted feature vector at the th moment, and through the third fully connected layer and the second activation layer, obtain the first predicted power of the vibrator and the second predicted power of the air pump at the th moment, and so on, the first predicted power of the vibrator and the second predicted power of the air pump at the The first predicted power of the vibrator and the second predicted power of the air pump from the first moment to the Nth moment, as well as the predicted feed amount between adjacent moments. That is, at the th moment, the first predicted power and the second predicted power at subsequent moments, as well as the predicted material weight at subsequent moments (instead of obtaining actual data after the control cycle ends) can be obtained through the above derivation method. Thus, the second loss function can be calculated after each moment, and training can be performed after each moment to continuously train the second fully connected layer and the third fully connected layer within the control cycle, improving the accuracy of determining the first power and the second power at subsequent moments.

[0046] According to an embodiment of the present invention, based on the predicted feed amount between each adjacent moment in the time period from the th moment to the Nth moment, determine the second loss function corresponding to the th moment, including: determining the second loss function corresponding to the th moment according to formula (2) , (2) Wherein, is the predicted feed amount between the kth moment and the k + 1th moment, , is the feed amount between the sth moment and the s + 1th moment, , is the added amount of raw materials, and both s and k are positive integers.

[0047] According to an embodiment of the present invention, is the total feed amount from the start moment of the control cycle to the th moment, is the sum of the predicted feed amounts between each adjacent moment in the time period from the th moment to the Nth moment. The sum of the two is the predicted total feed amount within the control cycle. is the difference between the added amount of raw materials and the predicted total feed amount within the control cycle, which can be used as the second loss function corresponding to the th moment. The second fully connected layer and the third fully connected layer are trained through this second loss function. For example, the second fully connected layer and the third fully connected layer are trained by backpropagating the second loss function. And the above training can be performed after each moment, thereby improving the accuracy of the second fully connected layer and the third fully connected layer, continuously improving the accuracy of the predicted first power of the vibrator and the second power of the air pump, so that the total feed amount of raw materials within the control cycle is as close as possible to the added amount of raw materials.

[0048] In this way, after each moment ends, the first predicted power, the second predicted power, and the predicted feed rate for subsequent moments can be solved iteratively. Thus, the total predicted feed amount within the control period can be determined based on the predicted feed rate, and then the error between the total predicted feed amount and the added amount of raw materials can be reduced by setting the second loss function, so as to train the second fully connected layer and the third fully connected layer, thereby continuously improving the accuracy of the second fully connected layer and the third fully connected layer, as well as the prediction accuracy of the first power of the vibrator and the second power of the air pump, and reducing the error between the total predicted feed amount and the added amount of raw materials.

[0049] According to an embodiment of the present invention, in step S6, before the end of the current control period, if the weight reduction in the hopper of the raw materials reaches the added amount, that is, the total feed amount of the raw materials within the current control period reaches the added amount, the feed channel can be closed, or the feed channel can be closed at the end of the current control period. And if the total feed amount of the raw materials does not reach the added amount at the end of the current control period, a prompt message can be generated to prompt the staff to replenish the materials. Moreover, the aforementioned feed rate prediction model, the second fully connected layer, and the third fully connected layer can be continuously trained in subsequent control periods, thereby continuously improving the model accuracy, making the total feed amount of the raw materials within one control period as close as possible to the added amount, and the duration of the process in which the total feed amount of the raw materials reaches the added amount as close as possible to the duration of the control period.

[0050] According to an embodiment of the present invention, in step S7, at the end of the current control period, based on the material weights in the hopper at the first preset number of moments and each subsequent moment, as well as the first power and the second power, train the feed rate prediction model in the intermediate state of the current control period to obtain the feed rate prediction model in the final state of the current control period, including: determining the predicted feed rate between each of the moments after the first preset number of moments based on the material weights in the hopper at each of the moments after the first preset number of moments, the first power, the second power, and the feed rate prediction model in the intermediate state of the current control period; obtaining the third loss function of the feed rate prediction model in the intermediate state of the current control period according to formula (3) , (3) wherein, is the predicted feed rate between the h-th moment and the (h + 1)-th moment, is the material weight at the h-th moment, is the material weight at the (h + 1)-th moment, , N is the serial number of the end moment of the current control period, is the first preset quantity, and h is a positive integer; according to the third loss function, the feed quantity prediction model in the intermediate state of the current control period is trained to obtain the feed quantity prediction model in the final state of the current control period.

[0051] According to an embodiment of the present invention, as described above, in the above steps, the first power and the second power at each moment after obtaining the first preset quantity are obtained, and the material weight in the hopper can be actually weighed at each moment. Therefore, an input vector at each moment can be formed based on the first power, the second power, and the material weight at each moment, combined with the output vector at the previous moment, and jointly input into the feed quantity prediction model in the intermediate state of the current control period to obtain the output vector at that moment, and processed through the first fully connected layer and the first activation layer to obtain the predicted feed quantity within the time period between that moment and the next moment. Further, since the above current control period has ended, therefore, the actual material weight in the hopper at all moments within the current control period can be obtained, and thus the actual material weights at adjacent moments can be subtracted to obtain the measured feed quantity. In formula (3), is the measured feed quantity between the h-th moment and the (h + 1)-th moment, is the predicted feed quantity between the h-th moment and the (h + 1)-th moment, and the error between the two is . By summing the errors at each moment, the third loss function can be obtained, and then the third loss function is backpropagated, and the parameters of the feed quantity prediction model in the intermediate state of the current control period are adjusted in the direction of reducing the third loss function to obtain the feed quantity prediction model in the final state of the current control period. Thus, it can be used in the next control period.

[0052] In this way, not only the data within the first preset number of moments is used to train the feed quantity prediction model, but also after the control period ends, the data of multiple moments after the first preset number of moments can be used to continue training the feed quantity prediction model, so as to maximize the use of the measured and predicted data within the control period, improve the training efficiency of the feed quantity prediction model, improve the prediction accuracy, and provide an accurate data basis for predicting the first power of the vibrator and the second power of the air pump.

[0053] The automatic batching control method for PVC pipe production according to an embodiment of the present invention can, when controlling the feeding amount, consider the influence of the material weight in the hopper on the feeding speed, so as to set appropriate powers for the vibrator and the air pump. It can also continuously update the feeding amount prediction model based on data such as the power and material weight obtained within the control period, thereby improving the accuracy of the powers of the vibrator and the air pump, enhancing the accuracy of the feeding amount prediction model, improving the control accuracy of the feeding amount of raw materials, and improving the quality of PVC pipes. When training the feeding amount prediction model in the final state of the previous control period, data at multiple past moments can be fused to accurately predict the predicted feeding amount between the i-th moment and the (i + 1)-th moment. When setting the first loss function, weights are set according to the number of moments, so as to accurately reflect the influence of the accuracy of the predicted data at this moment on the subsequent prediction accuracy, improve the training accuracy and training efficiency, and improve the accuracy of the feeding amount prediction model. When training the second fully connected layer and the third fully connected layer, after each moment ends, the first predicted power, the second predicted power, and the predicted feeding amount at subsequent moments can be solved iteratively, so as to determine the total predicted feeding amount within the control period through the predicted feeding amount. Furthermore, the error between the total predicted feeding amount and the added amount of raw materials can be reduced by setting the second loss function, so as to train the second fully connected layer and the third fully connected layer, thereby continuously improving the accuracy of the second fully connected layer and the third fully connected layer, as well as the prediction accuracy of the first power of the vibrator and the second power of the air pump, and reducing the error between the total predicted feeding amount and the added amount of raw materials. When training the feeding amount prediction model in the intermediate state of the current control period, not only the data within the first preset number of moments is used to train the feeding amount prediction model, but also the data at multiple moments after the first preset number of moments can be used to continue training the feeding amount prediction model after the control period ends, so as to maximize the utilization of the measured and predicted data within the control period, improve the training efficiency of the feeding amount prediction model, enhance the prediction accuracy, and provide an accurate data basis for predicting the first power of the vibrator and the second power of the air pump.

[0054] Figure 3An automatic batching control system for PVC pipe production according to an embodiment of the present invention is exemplarily shown. The system includes: an addition amount module, configured to obtain the addition amounts of various raw materials in the current control cycle at the start moment of the current control cycle; an opening module, configured to open the feeding channels corresponding to the various raw materials, and turn on the vibrators and air pumps of each feeding channel; a weighing module, configured to weigh the material weights in the hoppers corresponding to the respective raw materials at the first preset number of moments before the current control cycle; a first training module, configured to train the feeding amount prediction model in the final state of the previous control cycle according to the rated powers of the vibrators and air pumps and the material weights, to obtain the feeding amount prediction model in the intermediate state of the current control cycle; a determination module, configured to determine the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments in the current control cycle according to the material weights in the hoppers at each moment after the first preset number of moments in the current control cycle and the feeding amount prediction model in the intermediate state of the current control cycle; a closing module, configured to close the feeding channel when the weight reduction amount in the hopper of the raw material in the current control cycle reaches the addition amount, or at the end of the current control cycle; a second training module, configured to train the feeding amount prediction model in the intermediate state of the current control cycle according to the material weights in the hoppers at the first preset number of moments and each moment after that, and the first power and the second power, to obtain the feeding amount prediction model in the final state of the current control cycle.

[0055] The present invention can be a method, an apparatus, a system, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for performing various aspects of the present invention.

[0056] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the drawings are only examples and do not limit the present invention. The objectives of the present invention have been fully and effectively achieved. The functions and structural principles of the present invention have been shown and described in the embodiments, and the embodiments of the present invention can be deformed or modified in any way without departing from the principles.

Claims

1. An automatic batching control method for PVC pipe production, characterized in that, Including: At the beginning of the current control cycle, obtain the addition amounts of various raw materials within the current control cycle; Open the feeding channels corresponding to the various raw materials, and turn on the vibrators and air pumps of each feeding channel; at the first preset number of moments before the current control cycle, weigh the material weights in the hoppers corresponding to the respective raw materials; according to the rated powers of the vibrators and air pumps and the material weights, train the feeding amount prediction model in the final state of the previous control cycle to obtain the feeding amount prediction model in the intermediate state of the current control cycle; according to the material weights in the hoppers at each moment after the first preset number of moments in the current control cycle and the feeding amount prediction model in the intermediate state of the current control cycle, determine the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments in the current control cycle; when the weight reduction amount in the hopper of the raw material within the current control cycle reaches the addition amount, or when the current control cycle ends, close the feeding channel; at the end of the current control cycle, according to the material weights in the hoppers at the first preset number of moments and each moment after that, and the first power and the second power, train the feeding amount prediction model in the intermediate state of the current control cycle to obtain the feeding amount prediction model in the final state of the current control cycle.

2. The automatic batching control method for PVC pipe production according to claim 1, characterized in that, Training the feeding amount prediction model in the final state of the previous control cycle according to the rated powers of the vibrators and air pumps and the material weights to obtain the feeding amount prediction model in the intermediate state of the current control cycle includes: inputting the training input vector at the i-th moment composed of the rated power of the vibrator, the rated power of the air pump, and the material weight at the i-th moment, and the training feature vector at the (i - 1)-th moment into the i-th feature extraction layer of the feeding amount prediction model in the final state of the previous control cycle to obtain the training feature vector at the i-th moment, where i is a positive integer less than the first preset number, and when i = 1, the training feature vector at the (i - 1)-th moment is a zero vector; inputting the training feature vector at the i-th moment into the first fully connected layer and the first activation layer of the feeding amount prediction model in the final state of the previous control cycle for processing to obtain the predicted feeding amount between the i-th moment and the (i + 1)-th moment; determining the first loss function of the feeding amount prediction model according to the predicted feeding amount between the i-th moment and the (i + 1)-th moment, the material weight at the i-th moment, and the material weight at the (i + 1)-th moment; training the feeding amount prediction model in the final state of the previous control cycle according to the first loss function to obtain the feeding amount prediction model in the intermediate state of the current control cycle.

3. A method for automatically controlling the batching of PVC pipe production according to claim 2, characterized in that, Determine the first loss function of the feed rate prediction model according to the predicted feed rate between the \(i\)-th moment and the \((i + 1)\)-th moment, the material weight at the \(i\)-th moment, and the material weight at the \((i + 1)\)-th moment, including: according to the formula , determine the first loss function of the feed rate prediction model , where is the predicted feed rate between the \(i\)-th moment and the \((i + 1)\)-th moment, is the material weight at the \(i\)-th moment, is the material weight at the \((i + 1)\)-th moment, is the first preset quantity, and both \(i\) and are positive integers.

4. A method for automatically controlling the batching of PVC pipe production according to claim 2, characterized in that, Based on the material weights in the hopper at each moment after the first preset number of moments in the current control period, and the feed rate prediction model in the intermediate state of the current control period, determine the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments in the current control period, including: From the material weight in the hopper at the 1st moment to the th moment, the rated power of the vibrator, and the rated power of the air pump, form the first state vector at the 1st moment to the th moment, and input them into each feature extraction layer of the feed rate prediction model in the intermediate state of the current control period to obtain the first feature vector at the th moment; Obtain the number of remaining time periods between the th moment and the end moment of the current control period; Obtain the weight of the added material before the th moment, and obtain the first difference between the weight of the added material and the addition amount of the raw material; Combine the number of remaining time periods and the first difference to form the second state vector at the th moment, and process the second state vector through the second fully connected layer to obtain the second feature vector at the th moment; Concatenate the second feature vector at the th moment and the first feature vector at the th moment to obtain the third feature vector at the th moment; Input the third feature vector at the th moment into the third fully connected layer and the second activation layer to obtain the first power of the vibrator and the second power of the air pump at the th moment; From the material weight in the hopper at the (1 + j)th moment to the th moment, the rated power of the vibrator from the (1 + j)th moment to the th moment, the first power of the vibrator from the th moment to the th moment, the rated power of the air pump from the (1 + j)th moment to the th moment, and the second power of the air pump from the th moment to the th moment, form the first state vector at the (1 + j)th moment to the th moment, and input them into each feature extraction layer of the feed rate prediction model in the intermediate state of the current control period to obtain the first feature vector at the th moment, where j is an integer greater than or equal to 0, and , N is the serial number of the end moment of the current control period, and N is a positive integer; Obtain the number of remaining time periods between the th moment and the end moment of the current control period; Obtain the the weight of the added material before a certain moment, and obtain the first difference between the weight of the added material and the added amount of the raw material; combine the number of remaining time periods with the first difference to form the second state vector at a certain moment, and process the second state vector through the second fully connected layer to obtain the second feature vector at a certain moment; splice the second feature vector at a certain moment and the first feature vector at a certain moment to obtain the third feature vector at a certain moment; input the third feature vector at a certain moment into the third fully connected layer and the second activation layer to obtain the first power of the vibrator and the second power of the air pump at a certain moment.

5. The automatic batching control method for PVC pipe production according to claim 4, characterized in that, The method further comprises: The first feature vector at the moment is input into the first fully connected layer and the first activation layer for processing, and the first feature vector is obtained. From moment to The predicted feed volume between the moments; The weight of the material in the hopper at the first moment is From moment to The predicted feed amount between the moments is obtained The predicted material weight within a moment; The predicted material weight, the first power of the vibrator and the second power of the air pump at the moment are obtained. The first predicted feature vector at the moment; From moment to The predicted feed amount between moments is obtained The predicted weight of the added material before the moment, and obtain the first predicted difference between the predicted weight of the added material and the amount of raw material added; The number of remaining time periods between the moment and the end of the control cycle and the first prediction difference constitute the first The second predicted state vector at the moment is processed by the second fully connected layer to obtain the second state vector The second predicted feature vector at the moment The first predicted feature vector at the moment The second prediction feature vectors of the moment are spliced to obtain the The third predicted feature vector at the moment; The third predicted feature vector at the moment is input into the third fully connected layer and the second activation layer to obtain the The first predicted power of the vibrator and the second predicted power of the air pump at a moment; iteratively execute the steps of obtaining the first predicted power and the second predicted power to obtain the first predicted power and the second predicted power. The first predicted power of the vibrator and the second predicted power of the air pump from the first moment to the Nth moment, as well as the predicted feed amount between adjacent moments; The predicted feed amount between each adjacent moment in the time period from moment 1 to moment N is used to determine the The second loss function corresponding to the multiple moments is used to train the second fully connected layer and the third fully connected layer according to the second loss functions corresponding to the multiple moments.

6. The automatic batching control method for PVC pipe production according to claim 5, characterized in that, According to the predicted feed rate between each adjacent moment in the time period from the -th moment to the N-th moment, determine the second loss function corresponding to the -th moment, including: according to the formula , determine the second loss function corresponding to the -th moment, where is the predicted feed rate between the k-th moment and the k + 1-th moment, , is the feed rate between the s-th moment and the s + 1-th moment, , is the addition amount of the raw material, and both s and k are positive integers.

7. A method for automatically controlling the batching in the production of PVC pipes according to claim 1, characterized in that, At the end of the current control cycle, based on the material weights in the hopper at the first preset number of moments and each subsequent moment, as well as the first power and the second power, train the feed rate prediction model for the intermediate state of the current control cycle to obtain the feed rate prediction model for the final state of the current control cycle, including: determining the predicted feed rates between each moment after the first preset number of moments according to the material weights in the hopper at each moment after the first preset number of moments, the first power, the second power, and the feed rate prediction model for the intermediate state of the current control cycle; obtaining the third loss function of the feed rate prediction model for the intermediate state of the current control cycle according to the formula , where the third loss function of the feed rate prediction model for the intermediate state of the current control cycle is obtained , where is the predicted feed rate between the h-th moment and the (h + 1)-th moment, is the material weight at the h-th moment, is the material weight at the (h + 1)-th moment, , N is the serial number of the end moment of the current control cycle, is the first preset number, and h is a positive integer; train the feed rate prediction model for the intermediate state of the current control cycle according to the third loss function to obtain the feed rate prediction model for the final state of the current control cycle.

8. An automatic batching control system for PVC pipe production, characterized in that, Including: An addition amount module, configured to obtain the addition amounts of various raw materials within the current control cycle at the beginning of the current control cycle; An opening module, configured to open the feeding channels corresponding to the various raw materials and turn on the vibrators and air pumps of each feeding channel; A weighing module, configured to weigh the material weights in the hoppers corresponding to the respective raw materials at the first preset number of moments before the current control cycle; The first training module is used to train the feed quantity prediction model in the final state of the previous control cycle based on the rated power of the vibrator and the air pump and the material weight, so as to obtain the feed quantity prediction model in the intermediate state of the current control cycle; the determination module is used to determine the first power of the vibrator and the second power of the air pump at each moment after the first preset number of moments in the current control cycle according to the material weight in the hopper at each moment after the first preset number of moments in the current control cycle and the feed quantity prediction model in the intermediate state of the current control cycle; the closing module is used to close the feed channel when the weight reduction amount of the raw materials in the hopper in the current control cycle reaches the addition amount, or at the end of the current control cycle. The second training module is used to train the feed quantity prediction model in the intermediate state of the current control cycle based on the material weight in the hopper at the first preset number of moments and each moment after that, as well as the first power and the second power, at the end of the current control cycle, so as to obtain the feed quantity prediction model in the final state of the current control cycle.

Images