Health prediction method and system of evaporation dissolving machine for lyocell fibers

Abstract

The invention discloses a health prediction method and system for an evaporation dissolving machine for lyocell fibers, and belongs to the field of equipment health prediction, and the method comprises the steps: collecting equipment parameters of the evaporation dissolving machine and upstream pulper equipment thereof related to the normal operation of the evaporation dissolving machine, and internal product parameters of the equipment; time sequence data of a normal state and a fault state are included before and after the equipment fault time node of the evaporation dissolving machine; the collected data is preprocessed; and according to the parameter fluctuation degree, the degree of influence on the equipment health state and the severity of the parameters deviating from the normal range, weighting each parameter, obtaining a sample data set based on the time sequence data after each parameter is weighted, inputting the sample data set into an LSTM network model architecture to train a health prediction model, and predicting the health degree of the evaporation dissolving machine. According to the method, the equipment level parameters and the process level parameters of the evaporation dissolving machine are fused, so that the prediction precision and reliability of the equipment health degree are improved.

Description

Technical Field

[0001] The present invention relates to the field of equipment health prediction, and in particular to a health prediction method and system for an evaporation dissolution machine for Lyocell fibers. Background Art

[0002] The core technology of Lyocell fiber manufacturing lies in the preparation of spinning dope. Only when the spinning dope is prepared properly can qualified Lyocell fiber be spun. To prepare high-quality spinning dope, efficient dissolving equipment is required. The evaporator dissolver overcomes the problems of high viscosity of spinning dope, difficult heat transfer, and temperature sensitivity of NMMO solvent. With its large evaporation area and excellent evaporation performance, it can prepare qualified spinning dope in a short time. As the core equipment on the Lyocell fiber production line, the stable operation of the evaporator dissolver is the guarantee of smooth production process. Once a failure occurs, it will not only lead to production stagnation and increase maintenance costs, but may also trigger a chain reaction, causing immeasurable losses to the entire production line.

[0003] Equipment health prediction is to monitor and predict the operation status of equipment in real time through data analysis, model prediction and other technical means, so as to discover potential failure risks in advance and provide a scientific basis for maintenance and replacement. This prediction method not only helps to reduce unexpected downtime and maintenance costs, but also prevents production accidents and improves production efficiency. More importantly, equipment health prediction can optimize the operation of the entire production line, while enabling enterprises to reasonably arrange maintenance plans based on the prediction results to avoid excessive maintenance. Therefore, health prediction has become a key link in ensuring the stable operation of core equipment. It is not only related to the normal operation of the equipment, but also affects the economic benefits of the enterprise.

[0004] The working principle of the evaporation dissolving machine is to make the pulp flow in a film along the wall of the heating tube, evaporate the water therein, dissolve the cellulose, and thus prepare the spinning glue solution. The factors affecting the operating state of the evaporation dissolving machine not only depend on the equipment parameters itself, but are also closely related to the internal product parameters of the equipment. For example, in terms of equipment parameters, the current fluctuation of the stirring motor of the evaporation dissolving machine can reflect the load condition of the stirring motor, and abnormalities may indicate motor overload or failure; the feed pipeline pressure reflects the fluid flow state, and abnormalities may indicate blockage or pressure loss, etc. In terms of the internal product parameters of the equipment, the temperature of the dissolving liquid of the evaporation dissolving machine directly affects the dissolution efficiency of cellulose. Abnormal temperature may cause uneven slurry, which in turn affects the entire process; the pressure of the evaporation chamber affects the evaporation rate of the solvent, and abnormal pressure may cause evaporation failure or vacuum system failure, etc. In addition, as a downstream device of the pulper, the operating state of the evaporation dissolving machine is also affected by the operating conditions of the pulper. For example, the pulp concentration of the pulper will affect the dissolution uniformity, and too high a concentration may cause blockage; the pulp temperature of the pulper directly affects the feed temperature of the evaporator dissolver; the discharge flow rate of the pulper reflects the discharge stability of the pulper, affecting the continuity of subsequent processes, etc.

[0005] However, the current conventional equipment health prediction methods are mostly limited to the equipment level, relying only on the equipment's own operating data (such as vibration, current, pressure, etc.) for health prediction, and lack of process-level health correlation analysis. For example, the Chinese invention patent with publication number CN112381316A discloses a method for predicting the health status of electromechanical equipment based on a hybrid neural network model, and its training parameters are derived only from equipment-level data. However, for the evaporation and dissolution equipment on the Lyocell fiber production line, since its operating status is affected by many factors, it is particularly easy to cause certain parameters to exceed the limit warning due to short-term instantaneous process changes, resulting in false alarms, causing unnecessary downtime, and delaying production progress; or because the influence of the internal product parameters of the equipment is ignored, potential faults cannot be captured and missed reports are generated. For example, assuming that the operating temperature of the evaporation and dissolution equipment exceeds the normal range, the equipment data alone may be judged as a fault, but combined with the process data analysis, it is found that it is a normal phenomenon caused by the current production formula adjustment, and there is no need for false alarm warning. Therefore, due to the limited prediction accuracy, the prediction system that lacks process-level analysis is prone to misjudge normal working conditions as faults, or cannot capture real potential faults.

[0006] In view of the above problems, this scheme specially proposes a health prediction method and system for an evaporation dissolution machine for Lyocell fiber. Summary of the invention

[0007] In view of the problem that the current equipment health prediction is only targeted at the equipment level and lacks fault correlation analysis at the process level, resulting in insufficient prediction accuracy and a high false alarm rate, the present invention proposes a health prediction method and system for an evaporation and dissolution machine for Lyocell fiber. The method of the present invention combines data at the equipment level and the process level to perform fault correlation analysis, thereby improving the prediction accuracy of the prediction system, improving the prediction reliability, optimizing production scheduling, and improving overall production efficiency.

[0008] In order to achieve the above-mentioned object of the invention, the technical solution of the present invention is as follows: A health prediction method for an evaporation dissolving machine for lyocell fiber, comprising: In addition to collecting the equipment parameters of the evaporation dissolution machine itself, the equipment parameters of the evaporation dissolution machine and the equipment parameters and equipment parameters of the upstream pulping machine of the evaporation dissolution are also added to obtain the time series data of the above parameters including normal state and fault state before and after the time node of the evaporation dissolution machine equipment failure; among them, the equipment parameters of the evaporation dissolution machine itself include the evaporation chamber pressure, the stirring motor current, the feed flow rate and the feed pipeline pressure, and the equipment parameters include the temperature of the dissolved liquid, the solution viscosity and the solvent recovery rate; the equipment parameters of the upstream pulping machine of the evaporation dissolution machine include the discharge flow rate, and the equipment parameters include the pulp concentration and the pulp temperature; Based on the expert scoring method, the collected parameters are weighted from three dimensions: the degree of parameter fluctuation, the degree of influence of the parameters on the health status of the equipment, and the severity of the parameter deviation from the normal range. The sum of the products of each dimension and the corresponding weight of each parameter is calculated to obtain the influence weight of each parameter in health prediction. A sample data set is obtained based on the time series data after weighting various parameters, and is input into the LSTM network model architecture to train the health prediction model to predict the health of the evaporation dissolution machine.

[0009] Furthermore, when collecting data, various parameters are time-synchronized to ensure the timing alignment of different data sources; among them, the parameters of the upstream pulper are compensated in advance according to the material transportation time to match the operating status of the evaporator dissolver.

[0010] Furthermore, it also includes preprocessing the acquired time series data, and the preprocessing includes an outlier elimination step, a missing value filling step and a normalization processing step.

[0011] Furthermore, the outlier elimination step is performed by combining physical constraint method and statistical method. The physical constraint method refers to setting parameter thresholds based on the equipment operating range, and exceeding the range is considered an abnormality; the statistical method refers to using the 3σ principle to eliminate abnormal points based on the distribution of historical data.

[0012] Furthermore, the missing value filling step adopts the K nearest neighbor interpolation method to fill the missing value part by weighted averaging the K neighboring data values ​​near the missing value.

[0013] Furthermore, the parameter fluctuation degree is measured by the normalized standard deviation, and the calculation formula is as follows: ; In the formula, For parameters The normalized standard deviation of For parameters The standard deviation of is the sum of the standard deviations of all parameters.

[0014] Furthermore, the influence of parameters on the health status of equipment is measured by normalized expert score, and the calculation formula is as follows: ; In the formula, For parameters The normalized expert rating of For parameters Expert ratings, For the total score.

[0015] Furthermore, the severity of the deviation of the parameter from the normal range is measured by the physical constraint weight, which is calculated as follows: ; In the formula, Representation parameters The severity of the deviation from the normal range, For parameters The number of times beyond the normal range, For parameters The total number of samples.

[0016] Furthermore, a sample data set is obtained based on the time series data after weighting various parameters, including: determining the sliding window size and step size, converting the time series data into a sliding window of fixed length, and each input sliding window includes a time series feature and a target label.

[0017] Furthermore, the LSTM network model architecture includes an input layer, an LSTM layer, a fully connected layer and an output layer. The input layer is used to accept time series feature data, and its shape is determined by the step size and the total number of features at each step size; the LSTM layer is a single layer with 64 hidden units, which is used to capture long-term dependencies in the time series; the fully connected layer is used to generate prediction results; and the output layer is used to output the fault probability.

[0018] Furthermore, during the training process of the health prediction model, multiple rounds of iterations are used to optimize the model parameters, and overfitting is prevented through training loss convergence.

[0019] Furthermore, the training loss converges, that is, when the change in the loss of multiple consecutive rounds of training is less than the set threshold ϵ, the training is stopped, and the loss function is as follows: ; in, is the real fault label, 0=normal, 1=fault; is the failure probability predicted by the LSTM model; N is the total number of samples.

[0020] Furthermore, LSTM processes the time series feature data through multiple gates and finally calculates the failure probability P at the output layer. fault : P fault =σ(W h ×h T +b h ); Among them, W h and b h is the parameter learned by LSTM training; h T is the hidden state after LSTM processes the last time step; σ(x) is the Sigmoid function.

[0021] Furthermore, the health prediction model calculates the health index HI according to the output failure probability, and provides equipment health assessment and maintenance recommendations based on the HI value.

[0022] Furthermore, the calculation formula of the health index is as follows: HI=1-P fault ; Among them, HI is the health index; P fault The failure probability output by the health prediction model.

[0023] The present invention also proposes a health prediction system for an evaporation dissolution machine for Lyocell fibers, comprising: The data acquisition module is configured to collect equipment parameters of the evaporation dissolution machine and its upstream pulping machine and equipment-internal product parameters, and obtain time series data of the above parameters including normal state and fault state before and after the time node of the evaporation dissolution machine equipment failure; wherein, the equipment parameters collected from the evaporation dissolution machine include evaporation chamber pressure, stirring motor current, feed flow rate and feed pipeline pressure, and the equipment-internal product parameters include dissolved liquid temperature, solution viscosity and solvent recovery rate; the equipment parameters collected from the upstream pulping machine include discharge flow rate, and the equipment-internal product parameters include pulp concentration and pulp temperature; The data processing module is configured to weight the collected parameters based on the expert scoring method from three dimensions: parameter fluctuation degree, parameter impact on equipment health status, and severity of parameter deviation from the normal range, and calculate the sum of the product of each dimension of each parameter and the corresponding weight to obtain the impact weight of each parameter in health prediction; The model prediction module is configured to obtain a sample data set based on the time series data after the weights of various parameters are assigned, and input it into the LSTM network model architecture to train the health prediction model to predict the health of the evaporation dissolution machine.

[0024] Furthermore, it also includes a preprocessing module, which is configured to remove outliers, fill in missing values ​​and normalize the acquired time series data.

[0025] Furthermore, the preprocessing module includes an outlier elimination unit, a missing value filling unit and a normalization processing unit. The outlier elimination unit is configured to eliminate outliers by combining a physical constraint method and a statistical method. The physical constraint method refers to setting a parameter threshold based on the operating range of the equipment, and exceeding the range is regarded as an abnormality. The statistical method refers to using the 3σ principle to eliminate abnormal points based on the distribution of historical data; the missing value filling unit is configured to use a K-nearest neighbor interpolation method to fill the missing value part by weighted averaging the K neighboring data values ​​near the missing value; the normalization processing unit is configured to use a normalization formula to scale the data to a uniform range.

[0026] Furthermore, it also includes an early warning and maintenance suggestion module, which is configured to calculate a health index HI according to the failure probability output by the health prediction model, and provide equipment health assessment and maintenance suggestions according to the HI value.

[0027] Furthermore, the model prediction module includes: The model input unit is configured to determine the sliding window size and step size, converting the time series data into a sliding window of fixed length, where each input sliding window includes a time series feature and a target label; A model training unit is configured to optimize model parameters using multiple rounds of iterations and prevent overfitting through training loss convergence; A model output unit is configured to output a failure probability through a health prediction model; A model evaluation unit is configured to evaluate the accuracy of the model using a model evaluation indicator; The model validation unit is configured to use time series cross validation or to test the model performance using the actual operation data of the device.

[0028] In summary, the present invention has the following advantages: 1. The present invention improves the prediction accuracy by integrating the equipment-level parameters and process-level parameters of the evaporation dissolution machine. Based on time series analysis and LSTM network, by collecting the equipment parameters of the evaporation dissolution machine and the equipment-internal product parameters, combined with the equipment parameters of the upstream equipment and the equipment-internal product parameters, a cross-equipment and cross-process data coupling relationship is established. This multi-level data fusion method can more comprehensively reflect the production conditions, so that the prediction model can more accurately identify faults, thereby reducing the false alarm and missed alarm rates and improving the prediction reliability.

[0029] 2. The present invention adopts a method combining standard deviation + expert experience scoring + physical constraints to weight various parameters, ensuring that the importance of key parameters in the prediction model is reasonably reflected.

[0030] 3. The present invention uses the LSTM time series model to improve the accuracy of equipment health prediction. LSTM (Long Short-Term Memory Network) is used to learn time series data to improve the long-term dependency ability of the prediction model. LSTM can capture long-term trends and avoid misjudgments caused by short-term fluctuations.

[0031] 4. The present invention improves the intelligent level of equipment management by establishing a maintenance suggestion mechanism. By calculating the health index (HI), maintenance suggestions under different health conditions are provided: when HI>0.8, it means that the equipment is operating normally and only requires daily monitoring; when HI is in the range of 0.6 to 0.8, it means that the equipment is slightly abnormal and the process parameters can be adjusted; when HI is in the range of 0.4 to 0.6, planned maintenance needs to be arranged; when HI is in the range of 0.2 to 0.4, it means that the equipment is seriously abnormal and it is recommended to check immediately; when HI<0.2, it means that the equipment is faulty and it is recommended to shut down for maintenance or replace parts. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is an implementation roadmap of the present invention. DETAILED DESCRIPTION

[0033] In order to explain the present invention more clearly, the present invention is further described below in conjunction with preferred embodiments and drawings. It should be understood by those skilled in the art that the content described below is illustrative rather than restrictive, and should not be used to limit the scope of protection of the present invention.

[0034] Example 1 As the core equipment in the Lyocell fiber production line, the health of the evaporation dissolution machine is of vital importance. The present invention proposes a health prediction method for the evaporation dissolution machine for Lyocell fiber, such as Figure 1 As shown, the following steps are included: Step 1: Collect equipment parameters and equipment internal product parameters related to the operating status of the evaporation and dissolution machine, and obtain time series data of the above parameters including normal status and fault status before and after the evaporation and dissolution machine equipment failure time node.

[0035] This step specifically includes:

[0036] 1) Collect the equipment parameters of the evaporation dissolution machine and the internal product parameters of the equipment.

[0037] Specifically, the equipment parameters collected for the evaporation dissolution machine include the evaporation chamber pressure, stirring motor current, feed flow rate, and feed pipeline pressure; the equipment intrinsic product parameters collected include the dissolving liquid temperature, solution viscosity, and solvent recovery rate.

[0038] The influence of the above parameters on the operating status of the evaporation dissolution machine is as follows: Evaporation chamber pressure (Pa) - The evaporation chamber pressure affects the evaporation rate. Too low pressure will lead to too high solution concentration, affecting NMMO solvent recovery; too high pressure may cause equipment leakage.

[0039] Stirring motor current (A) - Stirring motor current affects the uniformity of the slurry. Too high may cause motor failure due to increased load, and below the normal range may indicate motor aging or low viscosity.

[0040] Feed flow rate (m³ / h)——The feed flow rate directly affects the stability of the evaporation and dissolution process. Low flow rate may mean that the pipeline is blocked or the pump is operating abnormally, and too high flow rate may cause solution dilution and affect evaporation efficiency.

[0041] Feed pipe pressure (Pa) - Feed pipe pressure reflects the fluid flow state. Abnormality may indicate blockage or pressure loss; it has a more indirect impact on the evaporation dissolution machine, but can be used to monitor abnormalities.

[0042] Solution temperature (°C) - The solution temperature directly affects the efficiency of cellulose dissolution. Abnormal temperature may lead to uneven slurry, which in turn affects the entire process.

[0043] Solution viscosity (Pa·s)——Solution viscosity directly affects spinning quality. Too low may indicate insufficient solution concentration, affecting the quality of finished products; too high may cause equipment blockage.

[0044] Solvent (NMMO) Recovery (%) - The solvent concentration affects the final dissolution result. Low recovery may mean a leak or vacuum system failure.

[0045] 2) Collect the equipment parameters of the pulper upstream of the evaporator and dissolver and the internal product parameters of the equipment.

[0046] As a downstream device of the pulper, the operation status of the evaporator dissolver is also affected by the operation status of the pulper. For example, when the viscosity of the pulper slurry is abnormal, it may affect the dissolution efficiency of the evaporator dissolver, causing the viscosity of the solution to fluctuate, and further affecting the subsequent spinning quality. In order to facilitate data collection, no additional collection equipment is added to the feed end of the evaporator dissolver. In this embodiment, the data of the pulper is directly included in the data collection stage, which also includes equipment parameters and equipment internal product parameters.

[0047] Specifically, the equipment parameters collected for the pulper in this solution include the discharge flow rate, and the internal product parameters collected for the equipment include the pulp concentration and the pulp temperature.

[0048] The influence of the above parameters on the operating status of the evaporation dissolution machine is as follows: Discharge flow rate (m³ / h)——The discharge flow rate of the pulper reflects the discharge stability of the pulper and affects the continuity of subsequent processes.

[0049] Slurry concentration (%) - affects dissolution uniformity, too high a concentration may cause blockage.

[0050] Slurry temperature (°C) - affects the feed temperature to the evaporator dissolver.

[0051] Finally, the collected parameters related to the health prediction of the evaporation dissolution machine are shown in Table 1 below.

[0052] Table 1 Collection parameter list

[0053] When collecting data, the timestamp synchronization method is used to uniformly add accurate timestamps to all data to ensure the timing alignment of different data sources. It should be noted that since the change in pulp quality of the pulper will not immediately affect the evaporator and dissolver, but there is a lag of about 10 minutes, the parameters of the pulper should be compensated in advance according to the material transportation time (using data about 10 minutes ahead of the evaporator and dissolver) to match the operating status of the evaporator and dissolver.

[0054] 3) Obtain the time series data of the above parameters including normal state and fault state before and after the failure time node of the evaporation dissolution machine equipment from the long-term historical data.

[0055] The data collection interval is set to 1-5 minutes. Table 2 below gives an exemplary raw time series data table, where all parameter values ​​collected at each time step correspond to a time series.

[0056] Table 2 Original time series data table

[0057] Step 2: Preprocess the collected time series data, including outlier removal, missing value filling and normalization.

[0058] 1) Outlier removal In this embodiment, a combination of physical constraint method and statistical method is used to eliminate outliers. The physical constraint method refers to setting parameter thresholds based on the operating range of the equipment. For example, the pressure in the evaporation chamber of the evaporation dissolution machine should be between 400Pa and 600Pa, and it is considered abnormal if it exceeds the range. The statistical method refers to using the 3σ principle to eliminate outliers based on the distribution of historical data. The specific formula is as follows: ; Among them, μ is the mean, σ is the standard deviation, and xi is the data point.

[0059] 2) Missing value filling The K nearest neighbor interpolation method is used to fill the missing value part by taking the weighted average of the K neighboring data values ​​near the missing value.

[0060] 3) Normalization The purpose of normalizing data is to scale the data to a uniform range, such as [0,1], to eliminate the effects of different dimensions. The normalization formula is as follows: ; In the formula, is the normalized value, is the original value, is the minimum value, is the maximum value.

[0061] The data in the original time series data table are processed by the above normalization formula to obtain the normalized data table shown in Table 3.

[0062] Table 3 Normalized data table

[0063] Step 3: Combining expert knowledge and experience, weight the collected parameters from three dimensions: parameter fluctuation degree, parameter impact on equipment health status, and severity of parameter deviation from the normal range, to further improve the accuracy of the prediction results.

[0064] ① The degree of parameter fluctuation is determined according to the following method: Calculate the standard deviation (σ) of each parameter. The larger the standard deviation, the greater the parameter fluctuation, which may have a stronger impact on the device status.

[0065] To ensure that the sum of the volatility contribution weights of all parameters is 1, the normalized standard deviation is used to measure the parameter volatility intensity. The calculation formula is as follows: ; In the formula, For parameters The normalized standard deviation of is, where For parameters The standard deviation of is the sum of the standard deviations of all parameters.

[0066] ② The degree of influence of the parameters on the health status of the equipment is determined according to the following method: In this embodiment, the influence of the parameters on the health status of the equipment is measured by expert scoring. Combining the expert knowledge and experience, each parameter is ranked and scored (0 to 10 points). The higher the score, the greater the influence of the parameter on the health status of the equipment.

[0067] The importance ranking and scores of the above parameters can be found in Table 4 below.

[0068] Table 4 Parameter importance ranking and score

[0069] It should be noted that although the pulp concentration and discharge flow rate parameters related to the pulper equipment in Table 4 are ranked at the bottom in importance, according to the physical mechanism analysis and the actual production process impact (combined with expert knowledge and experience), the scores set are higher than the parameters ranked higher in importance.

[0070] Based on the above scores, the normalized expert scores of each parameter are obtained, and the calculation formula is as follows: ; In the formula, For parameters Expert ratings, is the total score, For parameters Normalized expert ratings.

[0071] ③ The severity of the deviation of the parameter from the normal range is determined according to the following method: Set the physical constraint weight, taking into account the severity of the parameter deviation from the normal range. The calculation formula is as follows: ; In the formula, For parameters The number of times beyond the normal range, For parameters The total number of samples, Representation parameters The severity of the deviation from the normal range.

[0072] ④Determine the final weights of each parameter under the influence of the above three dimensions.

[0073] The final weight of each parameter considering the influence of the above three dimensions Determine according to the following formula: ; In the formula, , , , In this embodiment, it is possible to set =0.1, =0.6, =0.3.

[0074] In this step, the method of combining standard deviation + expert experience score + physical constraints is used to assign weights to each parameter, which can ensure that the importance of key parameters in the prediction model is reasonably reflected. When the early data is insufficient, more reliance is placed on expert experience to determine the weight value of each dimension (i.e. , , After data accumulation, parameter weight allocation can be more dependent on data-driven.

[0075] Step 4: Construct a feature sequence based on the time series data after each parameter is weighted to obtain training sample data for the prediction model.

[0076] First, determine the sliding window size and step size, and convert the time series data into a sliding window of fixed length. The sliding window method divides the continuous data into subsequences of fixed length to facilitate the model to learn time dependencies. Each input sliding window includes time series features (solvent temperature, evaporation chamber pressure, etc.) and target labels (normal / faulty), which are obtained from long-term historical data, mainly relying on equipment maintenance records.

[0077] Assume that the original time series data is represented as: ; In which, each data point x t It contains multiple features (solvent temperature, evaporation chamber pressure, etc.), and T represents the acquisition time.

[0078] Assume the sliding window size is w and the stride is s, then Input window It is composed of the following elements: ; Output target value Usually corresponds to the state of the last time step in the window, that is: ; Indicates the state of the last time step in the window. .

[0079] The window size w determines how much past information the model can "see" (for example, w=10 means using the previous 10 time points to predict the future). The step size s determines the interval at which the window slides (for example, s=1 means sliding 1 step each time, with more data overlap; s=5 means sliding 5 steps each time, with less data overlap).

[0080] By constructing a feature sequence in the above manner, training sample data for the prediction model is obtained.

[0081] Step 5: Input the training sample data into the constructed LSTM model architecture for training, use multiple rounds of iterations to optimize the model parameters, and use the early stopping strategy to prevent overfitting.

[0082] <1> Model architecture design Since the original data is a continuous time series that is continuously collected, and LSTM needs to input a fixed-length time window for training. During LSTM training, each sample is a (w, F) matrix, where w represents the window size and F represents the total number of features n at each step.

[0083] The model architecture includes: Input layer: used to accept time series data, with a shape of (step size s, number of features n).

[0084] LSTM layer: A single-layer LSTM with 64 hidden units is used to capture long-term dependencies in time series.

[0085] Fully connected layer: used to generate prediction results.

[0086] Output layer: used to output failure probability.

[0087] <2> Loss Function and Optimizer Classification task: Cross Entropy Loss (Binary Cross-Entropy).

[0088] Loss function: ; in, is the real fault label (0=normal, 1=fault); is the failure probability predicted by LSTM, and N is the total number of samples.

[0089] For regression tasks: mean absolute error (MAE) or mean squared error (MSE).

[0090] Optimizer: Adam optimizer with a learning rate of 0.001.

[0091] The above loss function selected in this embodiment is suitable for the binary classification problem. fault Deviation from the true label When , the loss will become larger, guiding the gradient adjustment.

[0092] <3> Model Training The training sample data obtained in step 4 is input into the constructed model framework, and the training sample data set is divided into a training set and a validation set in a ratio of 80: 20. The model parameters are optimized through multiple rounds of iteration (set to 100 rounds in this embodiment).

[0093] Specifically, in this embodiment, an early stopping strategy is used to prevent overfitting. The stopping training strategy is as follows: The training loss converges, that is, when the training loss changes for multiple consecutive rounds (set to 10 rounds in this embodiment) are less than a certain threshold ϵ (the threshold can be set to 10 in this embodiment) -4 ), then stop training, the formula is as follows: ; in, Represents the loss of this round of training, is the loss of the previous round of training.

[0094] When the training loss tends to be stable, it means that LSTM has learned the optimal parameters and is no longer optimized significantly.

[0095] Step 6: Use evaluation indicators to evaluate the accuracy of the model and verify the model.

[0096] Model evaluation indicators can be precision, recall, F1 score, mean square error (MSE) or mean absolute error (MAE).

[0097] The model validation method can adopt time series cross-validation or use the actual operation data of the equipment to test the model performance.

[0098] The present invention is based on time series analysis and LSTM network. By collecting the equipment parameters of the evaporation dissolution machine and the intrinsic product parameters of the equipment, and combining the equipment parameters of the upstream equipment and the intrinsic product parameters of the equipment, a cross-equipment and cross-process data coupling relationship is established. Through this multi-level data fusion method, the production conditions can be more comprehensively reflected, and the accuracy and reliability of the prediction model can be improved.

[0099] Step 7: Use predictive models to detect abnormal indicators and potential faults Based on the collection parameter list shown in Table 1, the real-time operation data of the evaporation dissolution machine for a period of time is collected, and after being processed in the same way as in the training stage, it is input into the trained prediction model.

[0100] LSTM processes time series data through multiple gates (forget gate, input gate, output gate), and finally calculates the failure probability P at the output layer. fault The last time step of LSTM outputs the hidden state h T , and then calculate the failure probability through the fully connected layer (DenseLayer): P fault =σ(W h ×h T +b h ); Among them, W h and b h is the parameter learned by LSTM training; h T is the hidden state after LSTM processes the last time step; σ(x) is the Sigmoid function: ; In this embodiment, the Sigmoid function is selected as the activation function because the output range of the function is (0, 1), so it is suitable for binary classification problems and represents the probability of equipment failure.

[0101] The health index is calculated based on the failure probability output by the LSTM model. If the health index is lower than the set safety threshold, it is determined that the device has a potential failure. The health index formula is as follows: HI=1-P fault ; Among them, HI is the health index; P fault is the failure probability.

[0102] Assuming the safety threshold is set to 0.3, then: If P fault >0.7 (more than 70%) indicates that the equipment has a high probability of failure and requires maintenance.

[0103] If P fault <0.3 (less than 30%), indicating that the device is in good health.

[0104] If P fault A gradual increase indicates that the equipment is gradually degrading, and HI is gradually decreasing.

[0105] It should be noted that the health index only reflects the current operating status of the equipment and does not indicate a specific fault.

[0106] Step 8. Based on the health index output by the prediction model, obtain the current equipment health and guiding maintenance suggestions. For specific processing methods, please refer to Table 5 below.

[0107] Table 5 Fault warning and maintenance suggestions

[0108] Example 2 The following is a comparative example to illustrate the health prediction method of the lyocell fiber evaporation dissolution machine proposed by the present invention.

[0109] One day, the DCS system monitored the real-time operating parameters of the evaporation dissolution machine over a period of time as shown in the following table: Table 6 Real-time operating parameters

[0110] The physical constraint ranges of various parameters are shown in Table 7 below.

[0111] Table 7 Parameter physical constraints

[0112] According to the physical constraint range of each parameter in Table 7, the device-level data in Table 6 has the following anomalies: The stirring motor current exceeds the upper limit of 40A; The evaporation chamber pressure is close to the upper limit of 600Pa; The power of the stirring motor exceeds the upper limit of 3.5KW; The discharge flow rate of 30 m³ / h exceeds the limit.

[0113] At the same time, DCS issued a warning message, believing that the current mixing system may have an overload problem and recommended shutting down for inspection; the vacuum of the evaporation system was too high and there might be a risk of leakage; the discharge flow was too high and there might be problems with pump delivery.

[0114] At this time, the process level data is analyzed: The slurry concentration is in the normal range of 8 ~ 12%, indicating that the slurry concentration control is stable and is not too thick or too thin.

[0115] The solution viscosity is in the normal range of 4.5 ~ 6.5 Pa·s, indicating that the stirring load is normal.

[0116] The feed flow rate is in the normal range of 40 ~ 50 m³ / h, indicating that the feed is stable.

[0117] The solvent recovery rate is within the normal range of 98~99%, indicating that the solvent recovery system operates normally without obvious leakage or excessive volatilization; the pipeline sealing is good and the solvent loss is controlled within a reasonable range.

[0118] The feed pipeline pressure is in the normal range of 48000 ~ 56000 Pa, indicating that the fluid is transported normally without blockage or leakage.

[0119] The slurry temperature is in the normal range of 55 ~ 65°C, indicating that the temperature is stable and suitable for the evaporation and dissolution process.

[0120] The real-time operating parameters in Table 6 above are input into the LSTM prediction model established in this solution, and the LSTM outputs the failure probability P fault If the equipment health index is 0.18, the equipment health index HI is 0.82. The corresponding maintenance recommendation is to maintain the current operation and monitor key parameters.

[0121] At the same time, according to the maintenance records of the day, the operation and maintenance team arrived at the site to troubleshoot the problem, and found no abnormalities. And 10 minutes after the fault alarm, the current of the stirring motor dropped to 36A, the pressure dropped to 550 Pa, and the stirring power recovered to 3.2kW, all returning to the normal range. By integrating the process-level data analysis, it was confirmed that the short-term fluctuations in the equipment-level parameters were caused by the feed adjustment, and no shutdown for maintenance was required. The correctness of the prediction results of this scheme was verified.

[0122] Example 3 Based on the same inventive concept, this embodiment also proposes a health prediction system for an evaporation dissolution machine for Lyocell fibers, including a data acquisition module, a data processing module and a model prediction module.

[0123] Among them, the data acquisition module is configured to collect equipment parameters and equipment-internal product parameters of the evaporation dissolution machine and its upstream pulper, and obtain time series data of the above parameters including normal state and fault state before and after the time node of equipment failure of the evaporation dissolution machine; among them, the equipment parameters collected from the evaporation dissolution machine include evaporation chamber pressure, stirring motor current, feed flow rate and feed pipeline pressure, and the equipment-internal product parameters include dissolved liquid temperature, solution viscosity and solvent recovery rate; the equipment parameters collected from the upstream pulper include discharge flow rate, and the equipment-internal product parameters include slurry concentration and slurry temperature.

[0124] The data processing module is configured to assign weights to the collected parameters based on the expert scoring method from three dimensions: the degree of parameter fluctuation, the degree of influence of the parameters on the health status of the equipment, and the severity of the parameter deviation from the normal range. The sum of the products of each dimension of each parameter and the corresponding weight is calculated to obtain the influence weight of each parameter in health prediction.

[0125] The model prediction module is configured to obtain a sample data set based on the time series data after the weights of various parameters are assigned, and input it into the LSTM network model architecture to train the health prediction model to predict the health of the evaporation dissolution machine.

[0126] Preferably, in the data processing module, the weight of each parameter is determined according to the following formula: ; In the formula, Representation parameters The degree of fluctuation, Representation parameters The degree of impact on the health of the equipment, Representation parameters The severity of the deviation from the normal range, Representation parameters The final weight of , and They are , , The weight of and , , .

[0127] Among them, the parameter fluctuation degree is measured by normalized standard deviation, and the calculation formula is as follows: ; In the formula, For parameters The standard deviation of is the sum of the standard deviations of all parameters.

[0128] The influence of parameters on the health status of equipment is measured by normalized expert score, and the calculation formula is as follows: ; In the formula, For parameters Expert ratings, For the total score.

[0129] The severity of the deviation of the parameter from the normal range is measured by the physical constraint weight, which is calculated as follows: ; In the formula, For parameters The number of times beyond the normal range, For parameters The total number of samples.

[0130] Furthermore, it also includes a preprocessing module, which is configured to remove outliers, fill in missing values ​​and normalize the acquired time series data.

[0131] Specifically, the preprocessing module includes an outlier elimination unit, a missing value filling unit and a normalization processing unit. The outlier elimination unit is configured to eliminate outliers by combining a physical constraint method and a statistical method. The physical constraint method refers to setting a parameter threshold based on the operating range of the equipment, and exceeding the range is regarded as an abnormality. The statistical method refers to using the 3σ principle to eliminate abnormal points based on the distribution of historical data; the missing value filling unit is configured to use a K-nearest neighbor interpolation method to fill the missing value part by weighted averaging the K neighboring data values ​​near the missing value; the normalization processing unit is configured to use a normalization formula to scale the data to a uniform range.

[0132] Furthermore, the model prediction module includes a model input unit, a model training unit, a model output unit, a model evaluation unit and a model verification unit.

[0133] The model input unit is configured to determine a sliding window size and a step size, and convert the time series data into a sliding window of fixed length, wherein each input sliding window includes a time series feature and a target label.

[0134] The model training unit is configured to optimize the model parameters through multiple rounds of iterations and prevent overfitting through training loss convergence. Training loss convergence means that training is stopped when the change in loss over multiple consecutive rounds of training is less than the threshold ϵ. The loss function is as follows: ; in, is the real fault label, 0=normal, 1=fault; is the failure probability predicted by the LSTM model; N is the total number of samples.

[0135] The model output unit is configured to output the failure probability P through the health prediction model fault , the calculation formula of failure probability is as follows: ; in, and are the parameters learned by LSTM training; is the hidden state after LSTM processes the last time step; σ(x) is the Sigmoid function.

[0136] The model evaluation unit is configured to evaluate the accuracy of the model using model evaluation indicators such as precision, recall, F1 score, mean square error (MSE) or mean absolute error (MAE); and to test the model performance using time series cross validation or actual device operation data.

[0137] The model validation unit is configured to adopt time series cross validation, or to test the model performance using the actual operation data of the device.

[0138] Furthermore, it also includes an early warning and maintenance suggestion module, which is configured to output the failure probability P according to the health prediction model. fault Calculate the health index HI and provide equipment health assessment and maintenance recommendations based on the HI value.

[0139] Specifically, the early warning and maintenance suggestion module automatically outputs the failure probability according to the LSTM model to calculate the health index. The calculation formula is: HI=1-P fault ; Among them, HI: health index; P fault : Failure probability. If the health index is lower than the set safety threshold, it is determined that the device has a potential failure.

[0140] The functions of the above functional modules correspond to the contents of the corresponding steps in Example 1, and will not be repeated for any unfinished description.

[0141] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any simple modification or equivalent change made to the above embodiment based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A health prediction method for an evaporation dissolution machine for lyocell fiber, characterized in that: include: In addition to collecting the equipment parameters of the evaporation dissolution machine itself, the equipment parameters of the evaporation dissolution machine and the equipment parameters and equipment parameters of the upstream pulping machine of the evaporation dissolution are also added to obtain the time series data of the above parameters including normal state and fault state before and after the time node of the evaporation dissolution machine equipment failure; among them, the equipment parameters of the evaporation dissolution machine itself include the evaporation chamber pressure, the stirring motor current, the feed flow rate and the feed pipeline pressure, and the equipment parameters include the temperature of the dissolved liquid, the solution viscosity and the solvent recovery rate; the equipment parameters of the upstream pulping machine of the evaporation dissolution machine include the discharge flow rate, and the equipment parameters include the pulp concentration and the pulp temperature; Based on the expert scoring method, the collected parameters are weighted from three dimensions: the degree of parameter fluctuation, the degree of influence of the parameters on the health status of the equipment, and the severity of the parameters deviating from the normal range. The sum of the products of each dimension and the corresponding weight of each parameter is calculated to obtain the influence weight of each parameter in health prediction. A sample data set is obtained based on the time series data after weighting various parameters, and is input into the LSTM network model architecture to train the health prediction model to predict the health of the evaporation dissolution machine.

2. The health prediction method of a lyocell fiber evaporation dissolution machine according to claim 1, characterized in that: When collecting data, various parameters are time-synchronized to ensure the timing alignment of different data sources; among them, the parameters of the upstream pulper are compensated in advance according to the material transportation time to match the operating status of the evaporator dissolver.

3. The health prediction method of a lyocell fiber evaporation dissolution machine according to claim 1, characterized in that: The method also includes preprocessing the acquired time series data, wherein the preprocessing includes an outlier elimination step, a missing value filling step, and a normalization processing step.

4. The health prediction method of a lyocell fiber evaporation dissolution machine as claimed in claim 3, characterized in that: The outlier elimination step is performed by combining a physical constraint method and a statistical method. The physical constraint method refers to setting a parameter threshold based on the equipment operating range, and exceeding the range is considered an abnormality; the statistical method refers to eliminating abnormal points based on the distribution of historical data using the 3σ principle.

5. The health prediction method of a lyocell fiber evaporation dissolution machine as claimed in claim 3, characterized in that: The missing value filling step adopts the K nearest neighbor interpolation method to fill the missing value part by weighted averaging the K neighboring data values ​​near the missing value.

6. The health prediction method of a lyocell fiber evaporation dissolution machine according to claim 1, characterized in that: The parameter fluctuation degree of the parameter is measured by the normalized standard deviation, and the calculation formula is as follows: ； In the formula, For parameters The normalized standard deviation of For parameters The standard deviation of is the sum of the standard deviations of all parameters.

7. The health prediction method of a lyocell fiber evaporation dissolution machine according to claim 1, characterized in that: The influence of parameters on the health status of equipment is measured by normalized expert score, and the calculation formula is as follows: ； In the formula, For parameters The normalized expert rating of For parameters Expert ratings, For the total score.

8. The health prediction method of a lyocell fiber evaporation dissolution machine according to claim 1, characterized in that: The severity of the deviation of the parameter from the normal range is measured by the physical constraint weight, which is calculated as follows: ； In the formula, For parameters The severity of the deviation from the normal range, For parameters The number of times beyond the normal range, For parameters The total number of samples.

9. The health prediction method of a lyocell fiber evaporation dissolution machine according to claim 1, characterized in that: A sample data set is obtained based on the time series data after the parameters are weighted, including: determining the sliding window size and step size, converting the time series data into a sliding window of fixed length, and each input sliding window includes time series features and target labels.

10. The health prediction method of the lyocell fiber evaporation dissolution machine according to claim 9, characterized in that: The LSTM network model architecture includes an input layer, an LSTM layer, a fully connected layer and an output layer. The input layer is used to receive time series feature data, and its shape is determined by the step size and the total number of features at each step size; the LSTM layer is a single layer with 64 hidden units, which is used to capture long-term dependencies in the time series; the fully connected layer is used to generate prediction results; and the output layer is used to output the failure probability.

11. The health prediction method of the lyocell fiber evaporation dissolution machine according to claim 10, characterized in that: LSTM processes time series feature data through multiple gates and finally calculates the failure probability P at the output layer. fault : P fault =σ(W h ×h T +b h ); Among them, W h and b h is the parameter learned by LSTM through training; h T is the hidden state after LSTM processes the last time step; σ(x) is the Sigmoid function.

12. The health prediction method of the lyocell fiber evaporation dissolution machine according to claim 1, characterized in that: During the training process of the health prediction model, multiple rounds of iterations are used to optimize the model parameters, and overfitting is prevented through training loss convergence.

13. The health prediction method of the lyocell fiber evaporation dissolution machine according to claim 11, characterized in that: The training loss converges, that is, when the change in the loss of multiple consecutive rounds of training is less than the set threshold ϵ, the training is stopped. The loss function is as follows: ； in, is the real fault label, 0=normal, 1=fault; is the failure probability predicted by the LSTM model; N is the total number of samples.

14. The health prediction method of the lyocell fiber evaporation dissolution machine according to claim 11, characterized in that: Also includes: The health prediction model calculates the health index HI based on the output failure probability and provides equipment health assessment and maintenance recommendations based on the HI value.

15. The health prediction method of the lyocell fiber evaporation dissolution machine according to claim 14, characterized in that: The calculation formula of health index is as follows: HI=1-P fault ; Among them, HI is the health index; P fault The failure probability output by the health prediction model.

16. A health prediction system for an evaporation dissolution machine for lyocell fiber, characterized in that: include: The data acquisition module is configured to collect equipment parameters of the evaporation dissolution machine and its upstream pulping machine and equipment-internal product parameters, and obtain time series data of the above parameters including normal state and fault state before and after the time node of the evaporation dissolution machine equipment failure; wherein, the equipment parameters collected from the evaporation dissolution machine include evaporation chamber pressure, stirring motor current, feed flow rate and feed pipeline pressure, and the equipment-internal product parameters include dissolved liquid temperature, solution viscosity and solvent recovery rate; the equipment parameters collected from the upstream pulping machine include discharge flow rate, and the equipment-internal product parameters include pulp concentration and pulp temperature; The data processing module is configured to weight the collected parameters based on the expert scoring method from three dimensions: parameter fluctuation degree, parameter impact on equipment health status, and severity of parameter deviation from the normal range, and calculate the sum of the products of each dimension of each parameter and the corresponding weight as the impact weight of each parameter in health prediction; The model prediction module is configured to obtain a sample data set based on the time series data after the weights of various parameters are assigned, and input it into the LSTM network model architecture to train the health prediction model to predict the health of the evaporation dissolution machine.

17. The health prediction system of the lyocell fiber evaporation dissolution machine according to claim 16, characterized in that: It also includes a preprocessing module, which is configured to perform outlier removal, missing value filling and normalization processing on the acquired time series data.

18. The health prediction system of the lyocell fiber evaporation dissolution machine according to claim 17, characterized in that: The preprocessing module includes an outlier elimination unit, a missing value filling unit and a normalization processing unit. The outlier elimination unit is configured to eliminate outliers by combining a physical constraint method and a statistical method. The physical constraint method refers to setting a parameter threshold based on the equipment operation range, and exceeding the range is regarded as an abnormality. The statistical method refers to using the 3σ principle to eliminate abnormal points based on the distribution of historical data; the missing value filling unit is configured to use a K-nearest neighbor interpolation method to fill the missing value part by weighted averaging the K neighboring data values ​​near the missing value; the normalization processing unit is configured to scale the data to a uniform range using a normalization formula.

19. The health prediction system of the lyocell fiber evaporation dissolution machine according to claim 16, characterized in that: It also includes an early warning and maintenance suggestion module, which is configured to calculate a health index HI according to the failure probability output by the health prediction model, and provide equipment health assessment and maintenance suggestions according to the HI value.

20. The health prediction system of the lyocell fiber evaporation dissolution machine according to claim 16, characterized in that: The model prediction module includes: The model input unit is configured to determine the sliding window size and step size, converting the time series data into a sliding window of fixed length, where each input sliding window includes a time series feature and a target label; A model training unit is configured to optimize model parameters using multiple rounds of iterations and prevent overfitting through training loss convergence; A model output unit is configured to output a failure probability through a health prediction model; A model evaluation unit is configured to evaluate the accuracy of the model using a model evaluation indicator; The model validation unit is configured to use time series cross validation or to test the model performance using the actual operation data of the device.

Images