Steam cabinet alarm control system based on mode recognition

Abstract

The invention discloses a steam cabinet alarm control system based on mode recognition, and belongs to the technical field of computer data processing. The system comprises a data acquisition module, a model construction module, an auxiliary monitoring module, a model optimization module and a mode judgment module. Historical working state data and real-time internal monitoring data of the steam cabinet are comprehensively acquired through the data acquisition module and are preprocessed; the model building module builds a mode recognition model with a preliminary recognition capability based on the preprocessed data; meanwhile, the auxiliary monitoring module continuously monitors and generates detail change characteristic values, and provides the model optimization module to continuously grind a mode recognition model to generate a mode recognition optimization model, so that various information of the operation of the steam cabinet is comprehensively collected and analyzed, various state changes in the operation of the steam cabinet are accurately captured, and the real-time performance of the steam cabinet is improved. Therefore, accurate identification and judgment of the working mode of the steam cabinet are realized, and the problem that intelligent monitoring of the steam cabinet in the prior art is not comprehensive enough is solved.

Description

Technical Field

[0001] The invention relates to the technical field of computer data processing, and in particular to a steam cabinet alarm control system based on pattern recognition. Background Art

[0002] With the booming catering industry and the continuous expansion of industrialized food processing, steam cabinets are widely used as an efficient and stable cooking and processing equipment. From traditional restaurants and restaurants that mass-produce staple foods and steamed dishes to food processing plants that steam various ingredients, steam cabinets are used very frequently. However, early steam cabinets lacked effective monitoring during operation, and safety accidents often occurred due to abnormal steam pressure, dry burning due to lack of water, overheating, etc., which not only damaged the equipment, but also endangered the lives of operators and caused economic losses. In order to ensure production safety and improve equipment stability, steam cabinet alarm control technology came into being, aiming to monitor the key operating parameters of the equipment in real time, timely warn of abnormalities, and ensure the reliable operation of the steam cabinet.

[0003] At present, the alarm control of steam cabinets has made significant progress. On the one hand, sensor technology has become more mature, and high-precision pressure sensors, temperature sensors, and water level sensors are widely integrated into steam cabinet systems, which can accurately collect operating data and control errors within a very small range, providing a solid foundation for accurate alarms. On the other hand, the intelligence level of the control module has been greatly improved. The microprocessor-based control system can quickly process and analyze sensor data. Once the data deviates from the normal threshold, it will immediately issue an alarm through various methods such as sound and light alarms, display pop-up windows, and even remote push messages to the mobile phones of managers. At the same time, some advanced steam cabinets can also automatically take emergency control measures, such as adjusting the steam intake and starting the water replenishment device, to achieve preliminary automated operation and maintenance.

[0004] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0005] In the prior art, traditional steam cabinets usually only have basic heating and steaming functions, and the lack of an intelligent monitoring system exposes many drawbacks in the operation process. On the one hand, due to the lack of intelligent monitoring means, it is difficult to accurately control the dynamic changes inside the steam cabinet, and potential fault hazards are difficult to detect in time. When the steam cabinet encounters an abnormal situation, the operator cannot obtain relevant information in time, which can easily induce food steaming failure, have a negative impact on food quality, and may even cause equipment damage due to long-term abnormal operation, resulting in increased maintenance costs and extended downtime.

[0006] To sum up, there is a problem that the intelligent monitoring of the steam cabinet is not comprehensive enough. Summary of the invention

[0007] In view of the above problems, the present invention is proposed. Therefore, the problem to be solved by the present invention is that in the prior art, it is difficult to accurately control the dynamic changes inside the steam cabinet, and it is also difficult to timely control abnormal conditions, and there is a problem that the intelligent monitoring of the steam cabinet is not comprehensive enough.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: a steam cabinet alarm control system based on pattern recognition, comprising: a data acquisition module, a model building module, an auxiliary monitoring module, a model optimization module, and a pattern judgment module;

[0009] The data acquisition module is used to acquire historical working status data of the steam cabinet, collect internal monitoring data of the steam cabinet in real time, and pre-process the historical working status data and internal monitoring data of the steam cabinet;

[0010] The model building module is used to build a pattern recognition model based on the preprocessed historical working status data;

[0011] The auxiliary monitoring module is used to obtain detail change characteristic values ​​based on the pre-processed internal monitoring data of the steam cabinet, and the detail change characteristic values ​​include: temperature change characteristic values, pressure change characteristic values, water level change characteristic values, steam flow change characteristic values, and power change characteristic values;

[0012] Create an empty data receiving buffer and continuously store the detail change characteristic values ​​through the data receiving buffer;

[0013] When the amount of data accumulated in the data receiving buffer reaches a first threshold, an array corresponding to the detail change characteristic value is extracted, the array corresponding to the detail change characteristic value is sent to the model optimization module, and the data receiving buffer is cleared;

[0014] The model optimization module is used to obtain an array corresponding to the detail change characteristic value, and generate a pattern recognition optimization model according to the array corresponding to the detail change characteristic value;

[0015] The pattern judgment alarm module is used to obtain the current internal monitoring data of the steam cabinet and input the current internal monitoring data of the steam cabinet into the pattern recognition optimization model. When the pattern recognition optimization model outputs that the working mode of the steam cabinet is a normal mode, the internal monitoring data of the steam cabinet is continuously monitored. When the output working mode of the steam cabinet is an abnormal mode, an alarm is triggered.

[0016] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the calculation formula of the temperature change characteristic value is:

[0017]

[0018] in, is the temperature value at time t1, is the temperature value at time t2, Tavg is the average temperature, Trange is the allowable difference of normal temperature fluctuation, e is a natural constant, and π is the circumference of a circle.

[0019] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the calculation formula of the pressure change characteristic value is:

[0020]

[0021] In the formula, is the pressure value at time t1, is the pressure value at time t2, Pavg is the average pressure, and Prange is the allowable difference of normal pressure fluctuation.

[0022] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the calculation formula of the water level change characteristic value is:

[0023]

[0024] In the formula, is the water level value at time t1, is the water level value at time t2, Havg is the average water level, and Hrange is the allowable difference of normal fluctuation of the water level.

[0025] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the calculation formula of the steam flow change characteristic value is:

[0026]

[0027] In the formula, is the steam flow value at time t1, is the steam flow value at time t2, Favg is the average steam flow over a period of time, and Frange is the allowable difference of normal fluctuation of steam flow.

[0028] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the calculation formula of the power change characteristic value is:

[0029]

[0030] In the formula, is the power value at time t1, is the power value at time t2, Wavg is the historical average power, and Wrange is the allowable difference in power fluctuation.

[0031] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the specific process of constructing the pattern recognition model is as follows:

[0032] Extracting normal working state data and abnormal working state data of the steam cabinet from the historical working state data;

[0033] Preprocess normal working status data and abnormal working status data;

[0034] Divide the preprocessed dataset into training set and validation set;

[0035] Use the training set to train the multi-layer perceptron architecture and set the initial learning rate X 101 , momentum coefficient X2 and number of iterations X 301 , the connection weights between neurons in each layer of the model are continuously adjusted through the back-propagation algorithm, so that the loss function value of the model on the validation set is gradually reduced;

[0036] When the validation set loss stops decreasing after N consecutive iterations, the training is stopped and the model parameters at this time are saved as the pattern recognition model.

[0037] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the formula of the pattern recognition model is as follows:

[0038] Assume that the multilayer perceptron has L layers, the number of neurons in the input layer is n0, and the corresponding input vector represents the historical working state feature data after preprocessing, the number of neurons in the hidden layer l (1≤l≤L-1) is nl, the number of neurons in the output layer is nL, and the output vector y=(y1,y2), which respectively represents the probability value of the steam cabinet being in normal and abnormal working modes;

[0039] For the hidden layer l, the activation value hl of its neurons is calculated as follows:

[0040] hl=σ(Wlhl- 1 +bl);

[0041] Where Wl is the weight matrix connecting the l-1th layer and the lth layer neurons, with a dimension of nl×nl-1, bl is the bias vector of the lth layer, with a dimension of nl, σ is the ReLU activation function, that is, σ(z)=max(0,z), z refers to the weighted input of the neuron, z=Wlhl- 1 +bl;

[0042] The calculation from the input layer to the hidden layer 1 is: 1 =σ(W 1 x+b 1 );

[0043] After being transmitted through multiple hidden layers, it reaches the output layer. The calculation of the output layer is normalized using the Softmax function. The pattern recognition model obtains the final prediction result yj through the following formula:

[0044]

[0045] Among them, e is a natural constant, zj is the first transfer function, is the output layer weight matrix W L The j-th row transposed vector of , bj is the bias of the neuron corresponding to the output layer, j = 1, 2, z k The k in the output layer is the index used to traverse the neurons in the output layer, and z k is the second transfer function, is the output layer weight matrix W L The row vector transpose corresponding to the connection weight of the k-th output layer neuron, bk is the bias of the output layer corresponding to the k-th output layer neuron, nL is the number of output layer neurons, nL=2, k traverses from 1 to nL.

[0046] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, generating a pattern recognition optimization model according to an array corresponding to the detail change characteristic value includes the following steps:

[0047] After receiving the arrays corresponding to the detail change feature values ​​from the auxiliary monitoring module, these arrays are first standardized to obtain feature value arrays;

[0048] The standardized feature value array is input into the constructed pattern recognition model as the input data of the secondary training, combined with the historical working status data, and the training set and test set are divided again. The secondary training process based on dynamic learning rate adjustment is started, and the learning rate is set to X 102 , every time after X 302 In the iteration, if the accuracy of the pattern recognition model on the test set is improved by less than X4, the learning rate is decayed to X5 times the original value;

[0049] At the same time, the loss function value on the test set is continuously monitored and weighted adjustments are made. When the loss function value no longer decreases after X6 consecutive iterations, the secondary training is stopped to obtain the pattern recognition optimization model.

[0050] As a preferred solution of the steam cabinet alarm control system based on pattern recognition described in the present invention, the specific process of continuously monitoring the loss function value on the test set and performing weighted adjustment is:

[0051] In the secondary training process, when a sample is marked as an abnormal mode, the loss weight Lweighted of the corresponding sample is increased to twice that of the normal sample in the loss function calculation, that is:

[0052] Lweighted = w*Lce;

[0053] Among them, Lce is the original cross entropy loss and w is the weight coefficient.

[0054] The beneficial effects of the present invention are as follows: the data acquisition module of the present invention comprehensively acquires the historical working status data and real-time internal monitoring data of the steam cabinet, and performs preprocessing, while the auxiliary monitoring module continuously monitors and generates characteristic values ​​of detail changes, and the model building module and the optimization module continuously polish the pattern recognition model, thereby comprehensively collecting and analyzing various types of information on the operation of the steam cabinet, accurately capturing various state changes in its operation, and then realizing accurate identification and judgment of the working mode of the steam cabinet, effectively solving the problem of insufficient comprehensiveness of intelligent monitoring of the steam cabinet in the prior art.

[0055] The present invention continuously updates the process of data acquisition, model optimization and pattern judgment in a cycle, that is, continuously acquires new monitoring data for model optimization, and then applies the optimized model to real-time judgment, thereby ensuring that the system can always adapt to various changes in the operation of the steamer, thereby achieving long-term stable operation of the system and effective monitoring of the entire life cycle of the steamer.

[0056] The present invention standardizes the detail change feature value array in the model optimization module and then inputs it into the constructed model for secondary training, and adopts a dynamic learning rate adjustment strategy and a weighted adjustment loss function, so that the model can keep up with the real-time operating status changes of the steam cabinet and continuously optimize its own performance, thereby achieving the effect of improving the model's sensitivity and accuracy in identifying abnormal working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0058] Figure 1 This is a structural diagram of a steam cabinet alarm control system based on pattern recognition in Example 1. DETAILED DESCRIPTION

[0059] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0060] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0061] The embodiment of the present application solves the problem of insufficient comprehensive intelligent monitoring of steam cabinets in the prior art by providing a steam cabinet alarm control system based on pattern recognition. The data acquisition module comprehensively collects the historical working status data and real-time internal monitoring data of the steam cabinet, and performs preprocessing to ensure the quality and availability of the data. The auxiliary monitoring module then uses a specific formula to calculate the detailed change characteristic values ​​of key parameters such as temperature, pressure, water level, steam flow, and power, and accurately captures the subtle state fluctuations in the operation of the steam cabinet. Finally, the pattern recognition model is built and continuously improved with the help of the model building module and the optimization module, so that it can deeply mine the complex patterns and potential laws in the data. Accurate identification and judgment of the working mode of the steam cabinet is achieved.

[0062] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0063] Example 1

[0064] Reference Figure 1 , which is the first embodiment of the present invention, and provides a steam cabinet alarm control system based on pattern recognition, the system comprising: a data acquisition module, a model building module, an auxiliary monitoring module, a model optimization module, and a pattern judgment module;

[0065] A data acquisition module is used to acquire historical working status data of the steam cabinet, collect internal monitoring data of the steam cabinet in real time, and pre-process the historical working status data and internal monitoring data of the steam cabinet;

[0066] A model building module, used to build a pattern recognition model based on the preprocessed historical working status data;

[0067] An auxiliary monitoring module is used to obtain detail change characteristic values ​​based on the pre-processed internal monitoring data of the steam cabinet, and the detail change characteristic values ​​include: temperature change characteristic values, pressure change characteristic values, water level change characteristic values, steam flow change characteristic values, and power change characteristic values;

[0068] Create an empty data receiving buffer and continuously store the detail change characteristic values ​​through the data receiving buffer;

[0069] When the amount of data accumulated in the data receiving buffer reaches a first threshold, an array corresponding to the detail change characteristic value is extracted, the array corresponding to the detail change characteristic value is sent to the model optimization module, and the data receiving buffer is cleared;

[0070] A model optimization module is used to obtain an array corresponding to the detail change characteristic value, and generate a pattern recognition optimization model according to the array corresponding to the detail change characteristic value;

[0071] The mode judgment alarm module is used to obtain the current internal monitoring data of the steam cabinet and input the current internal monitoring data of the steam cabinet into the pattern recognition optimization model. When the pattern recognition optimization model outputs that the working mode of the steam cabinet is a normal mode, the internal monitoring data of the steam cabinet is continuously monitored. When the output working mode of the steam cabinet is an abnormal mode, an alarm is triggered.

[0072] In this embodiment, the data acquisition module is closely connected with various sensors of the steam cabinet. It not only collects the historical working status data of the steam cabinet in all directions and deeply mines the normal and abnormal operating rules contained therein, but also collects various key monitoring data inside the steam cabinet in real time. The preprocessing includes data cleaning, normalization and other technologies.

[0073] The auxiliary monitoring module captures subtle changes inside the steam cabinet in real time, and transmits the collected detail change feature values ​​to the model optimization module in a timely manner, promoting continuous optimization of the model and continuously improving the accuracy of judgment.

[0074] The mode judgment alarm module quickly and accurately determines the working mode of the steamer based on the output results of the optimized model. Once an abnormal situation is detected, the alarm mechanism is immediately triggered to notify the operator to deal with it in time, thus realizing all-round and intelligent monitoring and management of the steamer. The safety and reliability of the steamer operation are significantly improved, which can not only effectively avoid problems such as food waste and equipment damage caused by malfunctions, but also reduce the high costs and failure risks derived from frequent maintenance.

[0075] Furthermore, the calculation formula of the temperature change characteristic value is:

[0076]

[0077] in, is the temperature value at time t1, is the temperature value at time t2, Tavg is the average temperature, Trange is the allowable difference of normal temperature fluctuation, e is a natural constant, and π is the circumference of the circle. For steaming cabinets, which have strict requirements on temperature control, it can detect subtle signs of temperature out of control in advance, effectively preventing poor food steaming or equipment failure caused by abnormal temperature.

[0078] Furthermore, the calculation formula of the pressure change characteristic value is:

[0079]

[0080] In the formula, is the pressure value at time t1, is the pressure value at time t2, Pavg is the average pressure, and Prange is the normal allowable difference of pressure fluctuation. The accurate pressure change characteristic value can ensure timely detection of any abnormal pressure, ensure a uniform and stable steaming process, and avoid food being undercooked or overcooked due to pressure problems.

[0081] Furthermore, the calculation formula of the water level change characteristic value is:

[0082]

[0083] In the formula, is the water level value at time t1, is the water level value at time t2, Havg is the average water level, and Hrange is the normal allowable difference of water level fluctuation. The water level of the steamer is directly related to the amount of steam generated and the steaming time. The accurate water level change characteristic value helps to timely detect problems such as water tank leakage and abnormal water replenishment, ensure the smooth progress of the steaming process, and avoid damage to the equipment due to dry boiling due to lack of water or excessive water overflow.

[0084] Furthermore, the calculation formula of the steam flow rate change characteristic value is:

[0085]

[0086] In the formula, is the steam flow value at time t1, is the steam flow value at time t2, Favg is the average steam flow over a period of time, and Frange is the normal fluctuation allowable difference of the steam flow. Stable steam flow ensures that the food is heated evenly. This characteristic value can immediately detect any abnormality in the steam flow, such as steam generator failure, pipe blockage, etc., to ensure the quality of steaming.

[0087] Furthermore, the calculation formula of the power variation characteristic value is:

[0088]

[0089] In the formula, is the power value at time t1, is the power value at time t2, Wavg is the historical average power, and Wrange is the allowable difference in power fluctuation. Accurate power characteristic values ​​can timely detect problems such as aging of heating elements and unstable power supply voltage, ensuring efficient and stable operation of the steam cabinet and reducing energy consumption and maintenance costs.

[0090] Furthermore, the specific process of constructing the pattern recognition model is as follows:

[0091] Extracting normal working state data and abnormal working state data of the steam cabinet from the historical working state data;

[0092] Preprocess normal working status data and abnormal working status data;

[0093] Divide the preprocessed dataset into training set and validation set;

[0094] Use the training set to train the multi-layer perceptron architecture and set the initial learning rate X 101 , momentum coefficient X2 and number of iterations X 301 , the connection weights between neurons in each layer of the model are continuously adjusted through the back-propagation algorithm, so that the loss function value of the model on the validation set is gradually reduced, including cross entropy loss, mean square error loss, etc.;

[0095] When the validation set loss stops decreasing after N consecutive iterations, the training is stopped and the model parameters at this time are saved as the pattern recognition model.

[0096] Furthermore, the formula of the pattern recognition model is as follows:

[0097] Assume that the multilayer perceptron has L layers, the number of neurons in the input layer is n0, and the corresponding input vector represents the historical working state feature data after preprocessing, the number of neurons in the hidden layer l (1≤l≤L-1) is nl, the number of neurons in the output layer is nL, and the output vector y=(y1,y2), which respectively represents the probability value of the steam cabinet being in normal and abnormal working modes;

[0098] For the hidden layer l, the activation value hl of its neurons is calculated as follows:

[0099] hl=σ(Wlhl- 1 +bl);

[0100] Where Wl is the weight matrix connecting the l-1th layer and the lth layer neurons, with a dimension of nl×nl-1, bl is the bias vector of the lth layer, with a dimension of nl, σ is the ReLU activation function, that is, σ(z)=max(0,z), z refers to the weighted input of the neuron, z=Wlhl- 1 +bl;

[0101] The calculation from the input layer to the hidden layer 1 is: 1 =σ(W 1 x+b 1 );

[0102] After being transmitted through multiple hidden layers, it reaches the output layer. The calculation of the output layer is normalized using the Softmax function. The pattern recognition model obtains the final prediction result yj through the following formula:

[0103]

[0104] Among them, e is a natural constant, zj is the first transfer function, is the output layer weight matrix W L The j-th row transposed vector of , bj is the bias of the neuron corresponding to the output layer, j = 1, 2, z k The k in the output layer is the index used to traverse the neurons in the output layer, and z k is the second transfer function, is the output layer weight matrix W L The row vector transpose corresponding to the connection weight of the k-th output layer neuron, bk is the bias of the output layer corresponding to the k-th output layer neuron, nL is the number of output layer neurons, nL=2, k traverses from 1 to nL.

[0105] Among them, the input layer receives the preprocessed historical data feature vector x and the first layer weight matrix W 1 Do the dot product and add the bias vector b 1 , and then use the ReLU activation function to get the activation value h of hidden layer 1 1 The subsequent hidden layer l repeats this process, using the previous layer activation value hl- 1 Calculated with the weight matrix Wl and the bias vector bl and activated by ReLU to obtain hl. After passing through multiple hidden layers, it finally reaches the output layer. First, according to h L-1 and the output layer weight matrix W L , the bias vector bj calculates the transfer function zj, and then uses the Softmax function to convert these transfer function values ​​into the probability value yj that the steam cabinet is in normal or abnormal mode, j = 1, 2, where j = 1 corresponds to the normal mode probability and j = 2 corresponds to the abnormal mode probability.

[0106] It needs to be explained that the ReLU activation function gives neurons nonlinear processing capabilities, can extract complex features, and solve linear inseparability problems; the Softmax function normalizes the output of the output layer neurons, intuitively presents the probability of the steamer being in normal or abnormal mode, and facilitates decision-making.

[0107] Further, generating a pattern recognition optimization model according to an array corresponding to the detail change characteristic value includes the following steps:

[0108] After receiving the arrays corresponding to the detail change feature values ​​from the auxiliary monitoring module, these arrays are first standardized to obtain feature value arrays;

[0109] The standardized feature value array is input into the constructed pattern recognition model as the input data of the secondary training, combined with the historical working status data, and the training set and test set are divided again. The secondary training process based on dynamic learning rate adjustment is started, and the learning rate is set to X 102 , every time after X 302 In the iteration, if the accuracy of the pattern recognition model on the test set is improved by less than X4, the learning rate is decayed to X5 times the original value;

[0110] At the same time, the loss function value on the test set is continuously monitored and weighted adjustments are made. When the loss function value does not decrease for X6 consecutive iterations, the secondary training is stopped to obtain the pattern recognition optimization model.

[0111] The standardized detail change feature value array is mixed with the previous historical working status data in a certain ratio (such as 50%:50%) and randomly divided into training set and test set. 102 Start the second training, every time you complete X 302 In the iteration, the model accuracy is evaluated on the test set. If the improvement is less than X4 (such as 0.5%), the new learning rate = original learning rate * X5 (such as 0.9), which prompts the model to fine-tune parameters with smaller steps and continue to optimize.

[0112] Furthermore, the specific process of continuously monitoring the loss function value on the test set and making weighted adjustments is as follows:

[0113] In the secondary training process, when a sample is marked as an abnormal mode, the loss weight Lweighted of the corresponding sample is increased to twice that of the normal sample in the loss function calculation, that is:

[0114] Lweighted = w*Lce;

[0115] Among them, Lce is the original cross entropy loss and w is the weight coefficient.

[0116] In this embodiment, when calculating the loss function each time, for the optimized pattern recognition optimization model, if the real label of the sample indicates an abnormal mode (such as the temperature of the steam cabinet is out of control, the pressure is abnormal, etc.), the weight coefficient of the sample in the cross entropy loss calculation is set to 2, that is, the loss value becomes twice the original, so that the pattern recognition optimization model gives priority to reducing the loss of such samples when adjusting parameters, so as to more accurately capture the abnormal feature mode and improve the ability to distinguish abnormalities. Normal samples are kept and participate in the training of the pattern recognition optimization model according to the conventional cross entropy loss calculation to maintain the balance of overall data learning.

[0117] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0118] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0119] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0120] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0121] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0122] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A steam cabinet alarm control system based on pattern recognition, characterized in that: include: Data acquisition module, model building module, auxiliary monitoring module, model optimization module, mode judgment module; The data acquisition module is used to acquire historical working status data of the steam cabinet, collect internal monitoring data of the steam cabinet in real time, and pre-process the historical working status data and internal monitoring data of the steam cabinet; The model building module is used to build a pattern recognition model based on the preprocessed historical working status data; The auxiliary monitoring module is used to obtain detail change characteristic values ​​based on the pre-processed internal monitoring data of the steam cabinet, and the detail change characteristic values ​​include: temperature change characteristic values, pressure change characteristic values, water level change characteristic values, steam flow change characteristic values, and power change characteristic values; Create an empty data receiving buffer and continuously store the detail change characteristic values ​​through the data receiving buffer; When the amount of data accumulated in the data receiving buffer reaches a first threshold, an array corresponding to the detail change characteristic value is extracted, the array corresponding to the detail change characteristic value is sent to the model optimization module, and the data receiving buffer is cleared; The model optimization module is used to obtain an array corresponding to the detail change characteristic value, and generate a pattern recognition optimization model according to the array corresponding to the detail change characteristic value; The pattern judgment alarm module is used to obtain the current internal monitoring data of the steam cabinet and input the current internal monitoring data of the steam cabinet into the pattern recognition optimization model. When the pattern recognition optimization model outputs that the working mode of the steam cabinet is a normal mode, the internal monitoring data of the steam cabinet is continuously monitored. When the output working mode of the steam cabinet is an abnormal mode, an alarm is triggered.

2. A steam cabinet alarm control system based on pattern recognition as claimed in claim 1, characterized in that: The calculation formula of the temperature change characteristic value is: in, is the temperature value at time t1, is the temperature value at time t2, Tavg is the average temperature, Trange is the allowable difference of normal temperature fluctuation, e is a natural constant, and π is the circumference of a circle.

3. A steam cabinet alarm control system based on pattern recognition as claimed in claim 1, characterized in that: The calculation formula of the pressure change characteristic value is: In the formula, is the pressure value at time t1, is the pressure value at time t2, Pavg is the average pressure, Prange is the allowable difference of normal pressure fluctuation, e is a natural constant, and π is the pi.

4. A steam cabinet alarm control system based on pattern recognition as claimed in claim 1, characterized in that: The calculation formula of the water level change characteristic value is: In the formula, is the water level value at time t1, is the water level value at time t2, Havg is the average water level, Hrange is the allowable difference of normal fluctuation of water level, e is a natural constant, and π is the pi.

5. A steam cabinet alarm control system based on pattern recognition as claimed in claim 1, characterized in that: The calculation formula for the characteristic value of steam flow change is: In the formula, is the steam flow value at time t1, is the steam flow value at time t2, Favg is the average steam flow over a period of time, Frange is the allowable difference of normal fluctuation of steam flow, e is a natural constant, and π is the ratio of pi.

6. A steam cabinet alarm control system based on pattern recognition as claimed in claim 1, characterized in that: The calculation formula of power variation characteristic value is: In the formula, is the power value at time t1, is the power value at time t2, Wavg is the historical average power, Wrange is the allowable difference in power fluctuation, e is a natural constant, and π is the circumference of a circle.

7. A steam cabinet alarm control system based on pattern recognition as claimed in claim 1, characterized in that: The specific process of constructing the pattern recognition model is as follows: Extracting normal working state data and abnormal working state data of the steam cabinet from the historical working state data; Preprocess normal working status data and abnormal working status data; Divide the preprocessed dataset into training set and validation set; Use the training set to train the multi-layer perceptron architecture and set the initial learning rate X 101 , momentum coefficient X2 and number of iterations X 301 , the connection weights between neurons in each layer of the model are continuously adjusted through the back-propagation algorithm, so that the loss function value of the model on the validation set is gradually reduced; When the validation set loss stops decreasing after N consecutive iterations, the training is stopped and the model parameters at this time are saved as the pattern recognition model.

8. The steam cabinet alarm control system based on pattern recognition as claimed in claim 1, characterized in that: The formula of the pattern recognition model is as follows: Assume that the multilayer perceptron has L layers, the number of neurons in the input layer is n0, and the corresponding input vector represents the historical working state feature data after preprocessing, the number of neurons in the hidden layer l (1≤l≤L-1) is nl, the number of neurons in the output layer is nL, and the output vector y=(y1,y2), which respectively represents the probability value of the steam cabinet being in normal and abnormal working modes; For the hidden layer l, the activation value hl of its neurons is calculated as follows: hl=σ(Wlhl- 1 +bl); Where Wl is the weight matrix connecting the l-1th layer and the lth layer neurons, with a dimension of nl×nl-1, bl is the bias vector of the lth layer, with a dimension of nl, σ is the ReLU activation function, that is, σ(z)=max(0,z), z refers to the weighted input of the neuron, z=Wlhl- 1 +bl; The calculation from the input layer to the hidden layer 1 is: 1 =σ(W 1 x+b 1 ); After being transmitted through multiple hidden layers, it reaches the output layer. The calculation of the output layer is normalized using the Softmax function. The pattern recognition model obtains the final prediction result yj through the following formula: Among them, e is a natural constant, zj is the first transfer function, is the output layer weight matrix W L The j-th row transposed vector of , bj is the bias of the neuron corresponding to the output layer, j = 1, 2, z k The k in the output layer is the index used to traverse the neurons in the output layer, and z k is the second transfer function, is the output layer weight matrix W L The row vector transpose corresponding to the connection weight of the k-th output layer neuron, bk is the bias of the output layer corresponding to the k-th output layer neuron, nL is the number of output layer neurons, nL=2, k traverses from 1 to nL.

9. A steam cabinet alarm control system based on pattern recognition as claimed in claim 7, characterized in that: Generating a pattern recognition optimization model according to an array corresponding to the detail variation eigenvalues ​​includes the following steps: After receiving the arrays corresponding to the detail change feature values ​​from the auxiliary monitoring module, these arrays are first standardized to obtain feature value arrays; The standardized feature value array is input into the constructed pattern recognition model as the input data of the secondary training, combined with the historical working status data, and the training set and test set are divided again. The secondary training process based on dynamic learning rate adjustment is started, and the learning rate is set to X 102 , every time after X 302 In the iteration, if the accuracy of the pattern recognition model on the test set is improved by less than X4, the learning rate is decayed to X5 times the original value; At the same time, the loss function value on the test set is continuously monitored and weighted adjustments are made. When the loss function value no longer decreases after X6 consecutive iterations, the secondary training is stopped to obtain the pattern recognition optimization model.

10. A steam cabinet alarm control system based on pattern recognition as claimed in claim 8, characterized in that: The specific process of continuously monitoring the loss function value on the test set and making weighted adjustments is: In the secondary training process, when a sample is marked as an abnormal mode, the loss weight Lweighted of the corresponding sample is increased to twice that of the normal sample in the loss function calculation, that is: Lweighted = w*Lce; Among them, Lce is the original cross entropy loss and w is the weight coefficient.

Images