Injection molding process data extraction method based on oil pump motor electrical parameter signals

Abstract

The application discloses an injection molding process data extraction method based on an oil pump motor electric parameter signal, comprising the following steps: S1, collecting injection molding machine power curve data and performing data preprocessing; S2, building a BiLSTM-Attenion network and iteratively training the same; S3, constructing a finite state machine based on an injection molding process; S4, fusing the trained BiLSTM-Attenion network and the finite state machine based on the injection molding process, identifying the injection molding process, and generating an identification result curve with a process label; and S5, extracting process data of each process according to the identification result curve with the process label. The application builds a deep learning network BiLSTM-Attention, introduces an injection molding process state machine based on injection molding professional knowledge, fuses the deep learning network and the injection molding process state machine to identify the injection molding process, and finally realizes accurate extraction of injection molding process data.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of injection molding machines, and particularly relates to an injection molding process data extraction method based on an oil pump motor electrical parameter signal. BACKGROUND

[0002] With the transformation and upgrading of digitization and intelligentization, the predictive maintenance and real-time data analysis method based on artificial intelligence play a more important role in the field of injection molding machines. Compared with the traditional internal sensor signal of the injection molding machine, the digital research using the oil pump motor electrical parameter signal is more convenient. While monitoring the energy consumption, the process identification and diagnosis of some common faults can be performed. Due to the characteristics of the step-by-step periodic work of the injection molding machine, the active power of the oil pump motor of the injection molding machine directly reflects the working state and load condition of the hydraulic system, which is closely related to the specific process being performed by the injection molding machine. Therefore, the electrical parameter signal of the oil pump motor of the injection molding machine can be used for process identification of the injection molding machine.

[0003] In the prior art, patent CN118114113A discloses an injection molding process identification method based on an injection molding machine electrical parameter curve. The periodic process signal of the injection molding machine is extracted for data processing, and then the state of each time point is obtained through an LSTM (Long Short Term Neural Network). Then, the process condition is obtained through each time point state. However, the oil pump motor electrical parameter signal is simply processed through the LSTM model, the process identification accuracy is low, and there is an error identification condition, which makes it very difficult to divide the process based on the method. SUMMARY

[0004] The main purpose of the present application is to overcome the shortcomings and deficiencies of the prior art, and to provide an injection molding process data extraction method based on an oil pump motor electrical parameter signal.

[0005] In order to achieve the above purpose, the following technical solutions are adopted in the present application:

[0006] The injection molding process data extraction method based on the oil pump motor electrical parameter signal comprises the following steps:

[0007] S1, collecting injection molding machine power curve data and performing data preprocessing;

[0008] S2, building a BiLSTM-Attenion network and iteratively training the same;

[0009] S3, constructing a finite state machine based on an injection molding process;

[0010] S4, fusing the trained BiLSTM-Attenion network and the finite state machine based on the injection molding process to identify the injection molding process and generate an identification result curve with process labels;

[0011] S5、According to the identification result curve with the process label, the process data of each process is extracted.

[0012] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0013] 1、The present application builds an excellent deep learning model BiLSTM-Attention for injection molding machine process recognition based on oil pump motor electrical parameters, which has higher accuracy than existing models for process recognition; the present application also introduces an injection molding process state machine based on injection molding professional knowledge to determine the authenticity of the deep learning process recognition result and correct the misjudged process, thereby achieving accurate process data recognition; the deep learning network and the injection molding process state machine are combined for injection molding process recognition, and finally the injection molding process data is accurately extracted. BRIEF DESCRIPTION OF DRAWINGS

[0014] Figure 1 is a flowchart of the method of the present application;

[0015] Figure 2 is a schematic diagram of the data preprocessing step in the embodiment;

[0016] Figure 3 is a schematic diagram of the BiLSTM-Attention network structure built in the embodiment;

[0017] Figure 4 is an injection molding process recognition result diagram of the BiLSTM-Attention network trained in the embodiment;

[0018] Figure 5 is a schematic diagram of a finite state machine based on an injection molding process;

[0019] Figure 6 is a schematic diagram of the injection molding process recognition result of a single BiLSTM-Attention network in the embodiment;

[0020] Figure 7 is a schematic diagram of the injection molding process recognition result of the fusion model in the embodiment. DETAILED DESCRIPTION

[0021] The present application will be further described in detail below in conjunction with the embodiments and the accompanying drawings, but the embodiments of the present application are not limited thereto.

[0022] EMBODIMENT

[0023] As shown in Figure 1 , the injection molding process data extraction method based on oil pump motor electrical parameter signals of the present application includes the following steps:

[0024] S1, collect injection molding machine power curve data and perform data preprocessing;

[0025] In this embodiment, the data preprocessing is specifically:

[0026] As shown in Figure 2 , the collected injection molding machine power curve is labeled using prior knowledge, which is divided into eight categories according to the working state of the injection molding machine, namely mold closing, mold locking, injection, pressure maintaining, glue melting, mold opening, ejector pin ejection and standby; determine the working state of the injection molding machine at each time point, label the working state of the injection molding machine at each time point of the injection molding machine power curve, then perform Fourier synchronous compression transformation on the time domain power data to obtain the frequency domain feature, and establish a time-frequency fusion feature dataset, which is composed of the time domain and frequency domain of the power data, the time domain signal is represented as:

[0027]

[0028] Where, t represents the center position of the window function, k represents the number of components, A k (t) represents the instantaneous amplitude, represents the instantaneous phase;

[0029] The frequency domain signal after Fourier synchronous compression transformation is represented as:

[0030]

[0031] Where, T(t,ω) represents the frequency distribution of the power signal after Fourier synchronous compression transformation, g (0) represents the window function at time 0, δ is the Dirac function, S(t,f) is the short-time Fourier transform signal of the power signal S(t), and ω(t,f) is the compressed instantaneous frequency;

[0032] Finally, the data is normalized and represented as:

[0033]

[0034] Where, X is the value of the current data, X min is the minimum value of a row of data, X max is the maximum value of a row of data, and X new is the normalized value.

[0035] S2, build BiLSTM-Attenion network and iterate training;

[0036] In this embodiment, as shown in Figure 3 , the BiLSTM-Attenion network specifically includes an input layer, a BiLSTM layer, an attention mechanism layer and an output layer;

[0037] Input layer, to receive raw data, i.e. time-frequency fusion features with dimension 51;

[0038] BiLSTM (Bidirectional Long Short-Term Memory) layer, composed of two directions of LSTM (forward LSTM and reverse LSTM), one processing forward information and one processing reverse information, and then splicing the results;

[0039] Attention mechanism layer, to realize score-based non-normalized linear transformation, followed by normalization by softmax function, specifically:

[0040] First, calculate the attention score, for each hidden state h t , get an unnormalized attention score e t through linear transformation:

[0041]

[0042] where, is the weight matrix, b a is the bias;

[0043] Then normalize e t to get attention weight a t :

[0044]

[0045] The obtained weight is used to measure the attention degree of the model to different parts of the input sequence;

[0046] Output layer, composed of a fully connected layer and a softmax function, to map the results from the attention mechanism layer to the expected output space, and output the classification results through the fully connected layer plus the softmax activation function.

[0047] For the BiLSTM-Attention network, the input is data with time-frequency fusion features with dimension 51, which is then trained through the BiLSTM layer. The number of hidden layers of the BiLSTM network is optimized to 64;

[0048] When the data passes through the BiLSTM layer, for each time step t, there are two hidden states: one from the forward and one from the reverse The two hidden states are connected to form a new hidden state This new hidden state is passed as the output of the BiLSTM layer to the next layer;

[0049] Finally, after the attention mechanism layer and the fully connected layer, the output is 8 categories, and the one with the highest probability is the final classification result.

[0050] The iterative training of the BiLSTM-Attention network comprises:

[0051] The BiLSTM-Attention network is trained by setting the number of training rounds, the training batch, and the training rate, and the network learns the time-frequency features from the injection cycle data (collected power data), and through iterative training, a BiLSTM-Attention network with a relatively ideal recognition accuracy is obtained.

[0052] As shown in Figure 4 , it is an injection process recognition result diagram of the trained BiLSTM-Attention network in this embodiment, and from the diagram, it can be seen that the recognition accuracy of each process is very ideal, and the overall recognition accuracy reaches 98.5%.

[0053] S3, constructing a finite state machine based on the injection process;

[0054] The finite state machine (FSM, finite state machine) is a mathematical model for describing the conditions of a finite number of states of an object and the transitions and actions between these states, which is composed of a set of states, an initial state, inputs, and transition rules based on inputs and existing states. The state machine is deterministic and is widely used in computer science and engineering applications. Since the injection cycle action of the injection molding machine is fixed, the basic workflow is mold closing, mold locking, injection, pressure maintaining, glue melting, mold opening, ejector pin ejection, and therefore a finite state machine based on the injection process can be constructed.

[0055] In this embodiment, specifically:

[0056] According to the injection cycle of the injection molding machine, a finite state machine based on the injection process is constructed, including mold closing, mold locking, injection, pressure maintaining, glue melting, mold opening, ejector pin ejection, and standby state;

[0057] According to the state transition relationship, a finite state machine based on the injection process is programmed using python language, and the input is the next state of the injection molding machine, and the output is whether to perform state transition. As shown in Figure 5 , it is a finite state machine diagram constructed in this embodiment.

[0058] As shown in Figure 5 , the state transition relationship is specifically:

[0059] The initial state is set to the standby state;

[0060] The condition for the initial state to enter the mold closing state is that the injection molding machine is switched from standby to mold closing, and remains in the mold closing state for three time steps, each time step being 0.1s; otherwise, the original state is maintained;

[0061] The condition for the mold closing state to enter the mold locking state is that only the mold closing state can be changed to the mold locking state, and the mold locking state is kept for three time steps; otherwise, the original state is kept, and the previous state of the mold closing state is saved as the mold locking state, which is a necessary condition for the mold closing state to change to the mold opening state;

[0062] The condition for the mold locking state to change to the standby state is that the mold locking state is changed to the standby state after the injection molding machine is locked, and the standby state is kept for three time steps; otherwise, the original state is kept, and the previous state of the standby state is saved as the mold locking state, which is a necessary condition for the standby state to change to the injection state;

[0063] The condition for the standby state to change to the injection state is that only the standby state can be changed to the injection state, and the previous state of the standby state is the mold locking state, and the injection state is kept for three time steps; otherwise, the original state is kept.

[0064] The condition for the injection state to change to the pressure maintaining state is that only the injection state can be changed to the pressure maintaining state, and the pressure maintaining state is kept for three time steps; otherwise, the original state is kept.

[0065] The condition for the pressure maintaining state to change to the melting glue state is that only the pressure maintaining state can be changed to the melting glue state, and the melting glue state is kept for three time steps; otherwise, the original state is kept.

[0066] The condition for the melting glue state to change to the standby state is that the melting glue state is changed to the standby state after the melting glue state is completed, and the standby state is kept for three time steps; otherwise, the original state is kept, and the previous state of the standby state is saved as the melting glue state, which is a necessary condition for the standby state to change to the mold opening state.

[0067] The condition for the standby state to change to the mold opening state is that only the standby state can be changed to the mold opening state, and the previous state of the standby state is the melting glue state, and the mold opening state is kept for three time steps; otherwise, the original state is kept.

[0068] The condition for the mold opening state to change to the standby state is that the mold opening state is kept for three time steps; otherwise, the original state is kept.

[0069] The condition for the standby state to change to the ejector pin ejection state is that only the standby state can be changed to the ejector pin ejection state, and the previous state of the standby state is the mold opening state.

[0070] S4, the BiLSTM-Attenion network trained and fused and the finite state machine based on the injection molding process identify the injection molding process, and generate an identification result curve with a process label; in this embodiment, specifically:

[0071] The BiLSTM-Attenion network trained and fused and the finite state machine based on the injection molding process input the process identification result output by the BiLSTM-Attenion network into the finite state machine based on the injection molding process, and obtain an identification result curve with a process label through the determination of the finite state machine.

[0072] like Figure 6 The image shows the identification results of a power data segment processed by a BiLSTM-Attenion network. The orange curve represents the active power curve of the injection molding machine's motor, and the blue curve represents the process identification results. 0-7 represent the eight identified states of the injection molding machine. It can be seen that certain abnormal points appear during the melt molding stage, indicating misdetection by the BiLSTM-Attenion network. The classification results are input into a finite state machine based on the injection molding process. The finite state machine corrects the erroneous process identifications, resulting in the corrected identification results, as shown below. Figure 7 As shown in the figure, the injection molding process identification results are obtained through the fusion model. It can be seen from the figure that the identification results are completely consistent with the operating state of the injection molding machine, which is beneficial for the extraction of injection molding process data.

[0073] S5. Based on the recognition result curve with process labels, extract the process data for each process; in this embodiment, specifically:

[0074] Based on the identification result curves with process labels, the active power curves of each process are extracted according to different process labels. Then, based on the active power curves of each process, the start and end times, process duration, and operating status of each process are obtained. This data can provide data resources for subsequent injection molding machine operating status monitoring and fault diagnosis.

[0075] The start and end times of the process are obtained based on the time nodes of state transitions. The process duration is the holding time of each state. The operating status of the process is monitored based on the power curve.

[0076] It should also be noted that, in this specification, terms such as "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0077] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for extracting injection molding process data based on electrical parameter signals of an oil pump motor, characterized in that The method comprises the following steps: S1, collecting injection molding machine power curve data and performing data preprocessing; S2, building a BiLSTM-Attenion network and iteratively training the same; S3, constructing a finite state machine based on the injection molding process, specifically: According to the injection molding cycle of the injection molding machine, a finite state machine based on the injection molding process is constructed, including mold closing, mold locking, injection, pressure maintaining, glue melting, mold opening, ejector pin ejection, and standby state; According to the state transition relationship, a finite state machine based on the injection molding process is programmed using python language, with the input being the next state of the injection molding machine and the output being whether to perform state transition; S4, fusing the trained BiLSTM-Attenion network and the finite state machine based on the injection molding process to identify the injection molding process and generate an identification result curve with process labels, specifically: Fuse the trained BiLSTM-Attenion network and the finite state machine based on the injection molding process, input the process identification result output by the BiLSTM-Attenion network into the finite state machine based on the injection molding process, correct the incorrect process identification through the judgment of the finite state machine, and obtain the corrected identification result; S5, extracting process data of each process according to the identification result curve with process labels.

2. The injection molding process data extraction method based on an electrical parameter signal of an oil pump motor according to claim 1, characterized by, In step S1, the data preprocessing is specifically: Use prior knowledge to label the collected injection molding machine power curve, divide it into eight states according to the working state of the injection molding machine, namely mold closing, mold locking, injection, pressure maintaining, glue melting, mold opening, ejector pin ejection, and standby state; determine the working state of the injection molding machine at each time point, label the working state of the injection molding machine at each time point of the injection molding machine power curve, then perform Fourier synchronous compression transformation on the time domain power data to obtain the frequency domain features, establish a time-frequency fusion feature dataset, and finally normalize the data and represent it as: where X is the value of the current data, X min is the minimum value of a row of data, X max is the maximum value of a row of data, X new is the normalized value.

3. The injection molding process data extraction method based on motor electrical parameter signals of an oil pump as defined in claim 2, characterized by, The time-frequency fusion feature dataset is composed of the time domain and frequency domain of the power data, the time domain signal is represented as: where t represents the center position of the window function, k represents the number of components, A k (t) represents the instantaneous amplitude, represents the instantaneous phase; The frequency domain signal after Fourier synchronous compression transformation is represented as: where T(t, ω) represents the frequency distribution of the power signal after Fourier synchro-compression transform, g (0) denotes the window function at time 0, δ is the Dirac function, S(t, f) is the short-time Fourier transform signal of the power signal S(t), and ω(t, f) is the compressed instantaneous frequency.

4. The injection molding process data extraction method based on motor electrical parameter signals of an oil pump according to claim 1, characterized by, In step S2, the BiLSTM-Attenion network specifically includes an input layer, a BiLSTM layer, an attention mechanism layer, and an output layer; The input layer receives the original data, i.e., the time-frequency fusion feature data with a dimension of 51; The BiLSTM layer is composed of two directions of LSTM, forward LSTM and reverse LSTM, one processing forward information and the other processing reverse information; The attention mechanism layer realizes a non-normalized linear transformation based on the score, and then normalizes it through a softmax function, specifically: First, the attention score is computed for each hidden state h t by a linear transformation to get an unnormalized attention score e t : wherein, is a weight matrix, b a is a bias, t represents a time step; Then normalize e t Get attention weight a t : The obtained weight is used to measure the attention degree of the model to different parts of the input sequence; The output layer is composed of a fully connected layer and a softmax function, which maps the results from the attention mechanism layer to the expected output space, and outputs the classification results through the fully connected layer plus the softmax activation function.

5. The injection molding process data extraction method based on motor electrical parameter signals of an oil pump according to claim 4, characterized by, For the BiLSTM-Attenion network, the input is time-frequency fusion feature data with a dimension of 51, which is then trained through the BiLSTM layer, and the number of hidden layers of the BiLSTM layer is 64; When data passes through the BiLSTM layer, for each time step t, there are two hidden states: one from the forward LSTM and one from the previous LSTM. and from the reverse LSTM Two hidden states are joined together to form a new hidden state. The new hidden state is passed as the output of the BiLSTM layer to the next layer; Finally, the attention mechanism layer and the full connection layer output the results of 8 categories, and the maximum probability is the final classification result.

6. The injection molding process data extraction method based on motor electrical parameter signals of an oil pump according to claim 5, characterized by, In step S2, the iterative training of the BiLSTM-Attention network includes: The number of training rounds, the training batch, and the training rate are set, and the BiLSTM-Attention network is trained. The network learns the time-frequency features from the injection molding cycle data, and through iterative training, a BiLSTM-Attention network that reaches the preset recognition accuracy is obtained.

7. The injection molding process data extraction method based on motor parameter signals of an oil pump as defined in claim 1, characterized by, The state transition relationship is as follows: The initial state is set to the standby state; The condition for the initial state to enter the mold closing state is that the injection molding machine changes from standby to mold closing, and stays in the mold closing state for three time steps, each of which is 0.1s; otherwise, the original state is maintained; The condition for the mold closing state to enter the mold locking state is that it can only change from mold closing to mold locking and stay in the mold locking state for three time steps; otherwise, the original state is maintained; The condition for the mold locking state to transfer to the standby state is that the injection molding machine completes mold locking and enters the standby state, and stays in the standby state for three time steps; otherwise, the original state is maintained, and the previous state of the standby state before this condition is saved as mold locking, which is a necessary condition for changing from standby to injection state; The condition for transferring to the injection state is that it can only change from standby to injection, and stay in the injection state for three time steps to enter the injection state; otherwise, the original state is maintained; The condition for the injection state to enter the pressure maintaining state is that it can only change from injection to pressure maintaining, and stay in the pressure maintaining state for three time steps; otherwise, the original state is maintained; The condition for the pressure maintaining state to enter the glue melting state is that it can only change from pressure maintaining to glue melting, and stay in the glue melting state for three time steps; otherwise, the original state is maintained; The condition for completing glue melting to enter the standby state is that it changes from glue melting to the standby state, and stays in the standby state for three time steps; otherwise, the original state is maintained, and the previous state of the standby state before this condition is saved as glue melting, which is a necessary condition for changing from standby to mold opening state; The condition for transferring to the mold opening state is that it can only change from standby to mold opening, and stay in the mold opening state for three time steps to enter the mold opening state; otherwise, the original state is maintained; The condition for the mold opening state to transfer to the standby state is that it stays in the standby state for three time steps; otherwise, the original state is maintained; The condition for entering the ejector pin ejection state is that it can only change from standby to mold opening, and the previous state of the standby state is mold opening.

8. The injection molding process data extraction method based on motor electrical parameter signals of an oil pump according to claim 7, characterized by, Step S5 is as follows: According to the obtained recognition result curve with process labels, the active power curve of each process is extracted according to different process labels, and then the start and end time, process time length, and running state of each process are obtained according to the active power curve of each process; The start and end time of the process is obtained according to the time node of state transition, the process time length is the holding time of each state, and the running state of the process is obtained by monitoring the power curve.

Images