A multi-point pressure safety detection method, device and Internet of Things system

Abstract

The present invention discloses a multi-point pressure safety detection method, device and Internet of Things system, wherein the multi-point pressure safety detection Internet of Things system comprises a pressure parameter measurement terminal 1-n, a pressure centralized controller, a pressure monitoring terminal, a pressure cloud platform server and a pressure monitoring mobile phone App, wherein the pressure parameter measurement terminal, the pressure monitoring terminal and the pressure cloud platform server are bidirectionally exchanged through the pressure centralized controller, and the pressure monitoring mobile phone App communicates bidirectionally with the pressure cloud platform server; the pressure parameter measurement terminal collects data of the measured object, and the pressure cloud platform server stores and manages data; through the pressure monitoring terminal and the pressure monitoring mobile phone App, the operating status of the monitored object can be supervised. The present invention can achieve a more accurate assessment of the pressure safety status.

Description

Technical Field

[0001] The present invention relates to pressure detection and safety warning, and in particular to a multi-point pressure safety detection method, device and Internet of Things system. Background Art

[0002] There are high temperature and high pressure conditions in industrial and agricultural production, and real-time monitoring of pressure helps to detect abnormal situations in a timely manner. Taking the pipeline network in the energy and chemical fields as an example, there are pressure measurement points throughout the long and complex pipeline network. However, the working environment of the pressure sensor is complex, and there are problems such as interference. In addition, the accuracy error of the pressure sensor makes it difficult to ensure the accuracy of the detection value. Moreover, the effect of temperature on pressure is not considered during pressure detection. In addition, the current pressure detection and safety warning are mostly based on a simple and crude threshold warning judgment based on a single sensor detection parameter, which is difficult to accurately reflect the overall pressure safety status of the object being measured. Summary of the invention

[0003] Purpose of the invention: The first purpose of the present invention is to provide an accurate and reliable multi-point pressure safety detection method; the second purpose of the present invention is to provide a multi-point pressure safety detection device; the third purpose of the present invention is to provide a multi-point pressure safety detection Internet of Things system.

[0004] Technical solution: The multi-point pressure safety detection method of the present invention comprises:

[0005] Get pressure sensor output 1-n;

[0006] Construct a multi-point pressure detection model:

[0007] The multi-point pressure detection model includes an improved wolf pack algorithm to optimize the pressure fuzzy clusterer, a pressure processing module 1, a pressure processing module 2, a pressure processing module 3, an IWO BiLSTM neural network model 1, an IWO BiLSTM neural network model 2, an IWO BiLSTM neural network model 3, an IWO BiLSTM neural network model 4, an IWO BiLSTM neural network model of triangular fuzzy numbers -ANFIS fuzzy neural network pressure classifier and a TDL beat delay device;

[0008] The pressure sensor outputs 1-n are used as the input of the improved wolf pack algorithm to optimize the pressure fuzzy clusterer. The improved wolf pack algorithm to optimize the pressure fuzzy clusterer outputs three types of pressure sensor outputs, one type of pressure sensor output is used as the input of the pressure processing module 1 and the corresponding input of the BiLSTM neural network model 3 of IWO, another type of pressure sensor output is used as the input of the pressure processing module 2 and the corresponding input of the BiLSTM neural network model 1 of IWO, and another type of pressure sensor output is used as the input of the pressure processing module 3 and the corresponding input of the BiLSTM neural network model 4 of IWO;

[0009] The output of the pressure processing module 1 is used as the corresponding input of the BiLSTM neural network model 2 of IWO and the BiLSTM neural network model 3 of IWO, the output of the pressure processing module 2 is used as the corresponding input of the BiLSTM neural network model 1 of IWO, and the output of the pressure processing module 3 is used as the corresponding input of the BiLSTM neural network model 1 and the BiLSTM neural network model 4;

[0010] The outputs of IWO's BiLSTM neural network model 1 are used as the corresponding inputs of IWO's BiLSTM neural network model 2 and IWO's BiLSTM neural network model 3, respectively. The outputs of IWO's BiLSTM neural network model 3 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model 2 are used as the corresponding inputs of IWO's BiLSTM neural network model 4 and IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers, respectively. The outputs of IWO's BiLSTM neural network model 4 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers are used as the inputs of the TDL beat delay device, and the outputs of the TDL beat delay device are used as the corresponding inputs of IWO's BiLSTM neural network model 2, IWO's BiLSTM neural network model 3, and IWO's BiLSTM neural network model 4, respectively.

[0011] The BiLSTM neural network model of IWO of triangular fuzzy numbers - the x, y and z output by the ANFIS fuzzy neural network pressure classifier constitute a triangular fuzzy number (x, y, z), and pressure safety graded warning is performed based on the triangular fuzzy number (x, y, z).

[0012] Furthermore, the multi-point pressure safety detection method further comprises: obtaining outputs of a plurality of temperature sensors, the temperature sensors being set at positions corresponding to the pressure sensors;

[0013] The pressure processing module includes Adaline neural network model 1, Adaline neural network model 2, VMD variational mode decomposition model 1, VMD variational mode decomposition model 2, IWO's BiLSTM neural network model 5, IWO's BiLSTM neural network model 6, IWO's BiLSTM neural network model 11, IWO's BiLSTM neural network model 12, IWO's BiLSTM neural network model 13, IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22, IWO's BiLSTM neural network model 23, ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 1, ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 and IWO's BiLSTM neural network model of triangular fuzzy numbers-ANFIS fuzzy neural network model;

[0014] The outputs of multiple pressure sensors are used as the corresponding inputs of Adaline neural network model 2. The outputs of Adaline neural network model 2 are used as the inputs of VMD variational mode decomposition model 2, IWO's BiLSTM neural network model 6 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model 6 are used as the corresponding inputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. VMD variational mode decomposition model 2 outputs the high-frequency analysis of pressure. The quantity, intermediate frequency component and low frequency component are respectively used as the input of IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22 and IWO's BiLSTM neural network model 23, the outputs of IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22 and IWO's BiLSTM neural network model 23 are used as the corresponding inputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2, and the outputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers;

[0015] The outputs of multiple temperature sensors corresponding to the multiple pressure sensors are used as the inputs of Adaline neural network model 1. The outputs of Adaline neural network model 1 are used as the inputs of VMD variational mode decomposition model 1, the inputs of IWO's BiLSTM neural network model 5 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. The high-frequency component, mid-frequency component and low-frequency component of the temperature output of VMD variational mode decomposition model 1 are used as the corresponding inputs of IWO's BiLSTM neural network model 11, IWO's BiLSTM neural network model 12 and IWO's BiLSTM neural network model 13 respectively. The output is used as the corresponding input of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO, the output of the BiLSTM neural network model 5 of IWO is used as the corresponding input of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO and the corresponding input of the BiLSTM neural network model 6 of IWO respectively, the output of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO is used as the corresponding input of the BiLSTM neural network model 2 of the ANFIS fuzzy neural network model-IWO, the output of the BiLSTM neural network model of IWO-ANFIS fuzzy neural network model of triangular fuzzy numbers is used as the corresponding input of the BiLSTM neural network model 2 of the ANFIS fuzzy neural network model-IWO and the BiLSTM neural network model 6 of IWO respectively;

[0016] The a, b, and c output by the BiLSTM neural network model of IWO-ANFIS fuzzy neural network model of triangular fuzzy numbers constitute triangular fuzzy numbers (a, b, c) as the predicted values ​​of the outputs of multiple pressure sensors.

[0017] Furthermore, the pressure safety level includes five levels: very safe, safe, normal, unsafe and very unsafe.

[0018] Further,

[0019] Serial number Warning level Triangular fuzzy numbers 1 Very safe (0.0，0.1，0.2) 2 Safety (0.2，0.3，0.4) 3 normal (0.4，0.5，0.6) 4 Unsafe (0.6，0.7，0.8) 5 Very unsafe (0.8，0.9，1.0)

[0020] The pressure safety level is determined according to the above table.

[0021] The multi-point pressure safety detection device of the present invention comprises a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the multi-point pressure safety detection method is implemented.

[0022] The multi-point pressure safety detection Internet of Things system described in the present invention is used to implement the multi-point pressure safety detection method described, including pressure parameter measurement terminals 1-n, a pressure centralized controller, a pressure monitoring terminal, a pressure cloud platform server and a pressure monitoring mobile phone App. The pressure centralized controller is used to realize two-way information exchange between the pressure parameter measurement terminal, the pressure monitoring terminal and the pressure cloud platform server, and the pressure monitoring mobile phone App communicates with the pressure cloud platform server in a two-way manner; the pressure parameter measurement terminal collects data of the object being measured, and the pressure cloud platform server stores and manages data; through the pressure monitoring terminal and the pressure monitoring mobile phone App, the operating status of the monitored object can be supervised.

[0023] Furthermore, the pressure parameter measurement end includes a pressure sensor, a temperature sensor, a flow sensor, a flow velocity sensor, a LoRa communication module and an STM32 microprocessor. Each sensor is connected to the STM32 microprocessor through a corresponding conditioning circuit, and the LoRa communication module is connected to the STM32 microprocessor.

[0024] Furthermore, the pressure centralized controller includes a flow regulating device, a temperature regulating device, a flow rate regulating device, an STM32 microprocessor, a wireless communication module, a LoRa communication module and a USB interface; each regulating device is connected to the STM32 microprocessor through a corresponding D / A conversion circuit; the wireless communication module, the LoRa communication module and the USB interface are connected to the STM32 microprocessor respectively; the pressure centralized controller realizes two-way data interaction with the pressure cloud platform server through the wireless communication module.

[0025] Furthermore, the pressure parameter measurement end and the pressure centralized controller build a self-organizing communication network through the LoRa communication module to achieve two-way communication.

[0026] Furthermore, the pressure monitoring end uses an industrial control computer, which is connected to the pressure centralized controller through a USB interface; the pressure monitoring mobile phone app communicates with the pressure cloud platform server through the 5G network; the main functions of the pressure monitoring end include pressure communication parameter setting, pressure data analysis, pressure data management and multi-point pressure detection model.

[0027] Beneficial effects: Compared with the prior art, the present invention has the following significant advantages:

[0028] (1) The present invention aims at the uncertainty and randomness caused by the pressure sensor accuracy error, interference and measurement anomaly during the parameter measurement process. The pressure processing module converts the parameter value measured by the pressure sensor into a triangular fuzzy number representation of the detection parameter, which effectively solves the fuzziness, dynamics and uncertainty of the pressure sensor measurement parameters and improves the objectivity and credibility of the sensor detection parameters. In addition, the influence of temperature on pressure is taken into account in the design of the pressure processing module.

[0029] (2) The BiLSTM neural network model includes two LSTM neural network models, which can extract the features of the time series of the input pressure information from two directions, thereby improving the prediction accuracy of the input pressure information. The forward and reverse long short-term memory networks extract and process the sequence information of the input pressure information. LSTM can not only extract information from the pressure sequence data like the standard recurrent neural network, but also retain the long-term pressure correlation information from the previous distant steps. The input pressure information has long-term spatial and temporal correlations, and the LSTM neural network model has enough long-term memory to process the spatiotemporal relationship between the input pressure information of the previous distant steps, which improves the accuracy and robustness of processing the input pressure information of the previous distant steps.

[0030] (3) The present invention improves the scientificity and reliability of the pressure warning of the detected object. The BiLSTM neural network model of IWO of triangular fuzzy numbers - ANFIS fuzzy neural network pressure classifier divides the pressure safety level of the detected object into five levels: very safe, safe, normal, unsafe and very unsafe, according to the parameters of the detected object, expert experience and engineering practice, and according to the impact of the pressure size on the detected object, thereby realizing the classification of the pressure safety status of the detected object and improving the scientificity and reliability of the pressure warning of the detected object.

[0031] (4) According to the characteristics of the sample differences of the output parameters of multiple pressure sensors of the detected object, the present invention constructs an improved wolf pack algorithm to optimize the pressure fuzzy clusterer to classify the outputs of the multi-point pressure sensors, fully considers the temporal and spatial characteristics of the output parameters of the multi-point pressure sensors, extracts relatively homogeneous pressure data with similar causes from a large amount of pressure data, and designs multiple corresponding pressure processing modules to predict the sample parameters. This is more targeted and can better reflect the actual situation of the pressure changes of the detected object at any time stage, thereby improving the prediction accuracy of the multi-point pressure. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 is a structural schematic diagram of a multi-point pressure detection model in an embodiment of the present invention;

[0033] Figure 2 is a schematic diagram of the structure of a pressure processing module in an embodiment of the present invention;

[0034] Figure 3 is a structural schematic diagram of a multi-point pressure safety detection Internet of Things system in an embodiment of the present invention;

[0035] Figure 4 is a schematic structural diagram of a pressure parameter measuring end in an embodiment of the present invention;

[0036] Figure 5 is a schematic diagram of the structure of a pressure centralized controller in an embodiment of the present invention;

[0037] Figure 6 It is a schematic diagram of the pressure monitoring terminal software function in an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described below in conjunction with the accompanying drawings.

[0039] The embodiment of the present invention provides a multi-point pressure safety detection method, which specifically includes the following steps:

[0040] (1) obtaining pressure sensor outputs 1-n and multiple temperature sensor outputs, wherein the temperature sensor setting positions correspond to the pressure sensor setting positions, that is, a temperature measuring point is correspondingly set near each pressure measuring point;

[0041] (2) Constructing a multi-point pressure detection model

[0042] like Figure 1 As shown, the multi-point pressure detection model includes an improved wolf pack algorithm to optimize the pressure fuzzy clusterer, a pressure processing module 1, a pressure processing module 2, a pressure processing module 3, an IWO BiLSTM neural network model 1, an IWO BiLSTM neural network model 2, an IWO BiLSTM neural network model 3, an IWO BiLSTM neural network model 4, an IWO BiLSTM neural network model of triangular fuzzy numbers -ANFIS fuzzy neural network pressure classifier and a TDL beat delay device;

[0043] The pressure sensor outputs 1-n are used as the input of the improved wolf pack algorithm to optimize the pressure fuzzy clusterer. The improved wolf pack algorithm to optimize the pressure fuzzy clusterer outputs three types of pressure sensor outputs, one type of pressure sensor output is used as the input of the pressure processing module 1 and the corresponding input of the BiLSTM neural network model 3 of IWO, another type of pressure sensor output is used as the input of the pressure processing module 2 and the corresponding input of the BiLSTM neural network model 1 of IWO, and another type of pressure sensor output is used as the input of the pressure processing module 3 and the corresponding input of the BiLSTM neural network model 4 of IWO;

[0044] The output of the pressure processing module 1 is used as the corresponding input of the BiLSTM neural network model 2 of IWO and the BiLSTM neural network model 3 of IWO, the output of the pressure processing module 2 is used as the corresponding input of the BiLSTM neural network model 1 of IWO, and the output of the pressure processing module 3 is used as the corresponding input of the BiLSTM neural network model 1 and the BiLSTM neural network model 4;

[0045] The outputs of IWO's BiLSTM neural network model 1 are used as the corresponding inputs of IWO's BiLSTM neural network model 2 and IWO's BiLSTM neural network model 3, respectively. The outputs of IWO's BiLSTM neural network model 3 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model 2 are used as the corresponding inputs of IWO's BiLSTM neural network model 4 and IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers, respectively. The outputs of IWO's BiLSTM neural network model 4 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers are used as the inputs of the TDL beat delay device, and the outputs of the TDL beat delay device are used as the corresponding inputs of IWO's BiLSTM neural network model 2, IWO's BiLSTM neural network model 3, and IWO's BiLSTM neural network model 4, respectively.

[0046] The BiLSTM neural network model of IWO of triangular fuzzy numbers - the x, y and z output by the ANFIS fuzzy neural network pressure classifier constitute triangular fuzzy numbers (x, y, z). According to the triangular fuzzy numbers (x, y, z), the pressure safety level of the detected object is divided into five levels: very safe, safe, normal, unsafe and very unsafe. The corresponding relationship between the five pressure warning levels and the triangular fuzzy numbers is shown in Table 1.

[0047] Table 1

[0048]

[0049]

[0050] like Figure 2As shown, the pressure processing module includes Adaline neural network model 1, Adaline neural network model 2, VMD variational mode decomposition model 1, VMD variational mode decomposition model 2, IWO's BiLSTM neural network model 5, IWO's BiLSTM neural network model 6, IWO's BiLSTM neural network model 11, IWO's BiLSTM neural network model 12, IWO's BiLSTM neural network model 13, IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22, IWO's BiLSTM neural network model 23, ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 1, ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 and IWO's BiLSTM neural network model of triangular fuzzy number-ANFIS fuzzy neural network model;

[0051] The outputs of multiple pressure sensors are used as the corresponding inputs of Adaline neural network model 2. The outputs of Adaline neural network model 2 are used as the inputs of VMD variational mode decomposition model 2, IWO's BiLSTM neural network model 6 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model 6 are used as the corresponding inputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. VMD variational mode decomposition model 2 outputs the high-frequency analysis of pressure. The quantity, intermediate frequency component and low frequency component are respectively used as the input of IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22 and IWO's BiLSTM neural network model 23, the outputs of IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22 and IWO's BiLSTM neural network model 23 are used as the corresponding inputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2, and the outputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers;

[0052] The outputs of multiple temperature sensors corresponding to the multiple pressure sensors are used as the inputs of Adaline neural network model 1. The outputs of Adaline neural network model 1 are used as the inputs of VMD variational mode decomposition model 1, the inputs of IWO's BiLSTM neural network model 5 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. The high-frequency component, mid-frequency component and low-frequency component of the temperature output of VMD variational mode decomposition model 1 are used as the corresponding inputs of IWO's BiLSTM neural network model 11, IWO's BiLSTM neural network model 12 and IWO's BiLSTM neural network model 13 respectively. The output is used as the corresponding input of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO, the output of the BiLSTM neural network model 5 of IWO is used as the corresponding input of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO and the corresponding input of the BiLSTM neural network model 6 of IWO respectively, the output of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO is used as the corresponding input of the BiLSTM neural network model 2 of the ANFIS fuzzy neural network model-IWO, the output of the BiLSTM neural network model of IWO-ANFIS fuzzy neural network model of triangular fuzzy numbers is used as the corresponding input of the BiLSTM neural network model 2 of the ANFIS fuzzy neural network model-IWO and the BiLSTM neural network model 6 of IWO respectively;

[0053] The a, b, and c output by the BiLSTM neural network model of IWO-ANFIS fuzzy neural network model of triangular fuzzy numbers constitute triangular fuzzy numbers (a, b, c) as the predicted values ​​of the outputs of multiple pressure sensors.

[0054] 1. Pressure treatment module design

[0055] The pressure processing module includes Adaline neural network model, VMD variational mode decomposition model, IWO's BiLSTM neural network model, ANFIS fuzzy neural network model-IWO's BiLSTM neural network model and IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers;

[0056] 1) VMD variational mode decomposition model design

[0057] The VMD variational mode decomposition model is a frequency domain decomposition method that decomposes the input signal into multiple components with narrowband properties in the frequency domain. Assume that the input signal y(t) is composed of k intrinsic mode functions u k (t), the center frequency of each natural mode function is W k (t), this separable convex optimization problem can be well solved by the alternating direction multiplier method, which decomposes the input pressure signal y(t) into k optimal intrinsic mode components (IMF). The original signal y(t) is decomposed by VMD variational mode to obtain k IMF components (X1(t), X2(t), ..., X k (t)). By setting the penalty factor and the initial decomposition number k = 2, the correlation coefficients between different IMF components and the correlation coefficients between the IMF components and the original signal y(t) are calculated, and the decomposition number k is continuously increased to determine the optimal decomposition number.

[0058] 2) IWO’s BiLSTM neural network model design

[0059] The BiLSTM neural network model is a bidirectional long short-term memory network, which is composed of a bidirectional LSTM neural network model. The LSTM neural network model is a type of recurrent neural network. The design of the LSTM neural network model structure is to avoid the long-term dependency problem that occurs in the RNN neural network model. When dealing with pressure time series problems, the LSTM neural network model has stronger time series learning ability and pressure information selection ability, can better coordinate the information distribution in the historical memory unit, and has better memory coordination ability than RNN.

[0060] The LSTM neural network model has three gates, namely the forget gate, input gate, and output gate, which solve the problem that RNN cannot effectively process long-distance information and is prone to gradient disappearance and gradient explosion. The LSTM neural network model can extract the features of the time series of input pressure information from two directions, thereby improving the prediction accuracy of the input pressure information. The forward and reverse long short-term memory networks extract and process the sequence information of the input pressure information, and finally extract the features from H t The predicted value information is output at the output, and the calculation formula is as follows:

[0061]

[0062] in, is the positive sequence information, is the reverse sequence information, M LSTM is the calculation process of LSTM, X t To enter pressure information, is the positive pressure sequence information, is the reverse pressure sequence information, Ht is the output information, w1 is the forward output weight, w2 is the backward output weight, c t Optimize the parameters for bias.

[0063] IWO (Invasive Weed Optimization) mainly simulates the growth and reproduction of weeds, and has the characteristics of good robustness and outstanding adaptability. Compared with genetic algorithms or PSO algorithms, IWO has obvious advantages in dealing with multi-peak function problems. The BiLSTM neural network model of the IWO optimization algorithm can be divided into 5 steps: 1) Population initialization: Within the limited range of the solution space, m0 weeds are randomly scattered. The positions of these m0 weeds are a set of solutions to the problem function to be solved. All seed positions are evaluated to determine their respective fitness values. 2) Growth and reproduction: Sort the seed fitness values ​​determined by the evaluation in the previous step, and then simulate the weed reproduction operation. The number of seeds generated by the parent weeds is linearly related to the fitness value of the parent. 3) Spatial diffusion: With the parent as the axis mean, the offspring individuals are diffused in the solution space in a normal distribution. 4) Competitive elimination: After completing the reproduction and diffusion of the first two steps, the parents and offspring are arranged together, and eliminated according to the fitness value, leaving the individuals with the highest fitness value to meet the population limit. 5) Repeat steps 2) to 4) and continue iterating until the maximum number of iterations is reached or the optimal solution that meets the conditions appears.

[0064] 3) Adaline neural network model design

[0065] The adaptive linear unit of the Adaline neural network model is one of the early neural network models. The input signal of this model is X(K)=[x0(K),x1(K),…,x n (K)] T , n is the dimension of the input signal, and each set of input signals corresponds to a set of weight vectors, which are expressed as W(K)=[k0(K),k1(K),…,k n (K)], when x0(K) is equal to negative 1, it is the bias value of the Adaline neural network model that determines the excitation or inhibition state of the neuron. The network output can be defined according to the input vector and weight vector of the Adaline neural network model as:

[0066]

[0067] There is a special input in the Adaline neural network model, namely the ideal response output d(K). It is sent to the Adaline neural network model, and then compared with the network output Y(K). The difference is sent to the learning algorithm mechanism to adjust the weight vector until the optimal weight vector is obtained. Y(K) and d(K) tend to be consistent. The adjustment process of the weight vector is the learning process of the network. The learning algorithm is the core part of the learning process. The weight optimization search algorithm of the Adaline neural network model adopts the LMS algorithm least squares method. The Adaline neural network model outputs the linear value of the detected parameter.

[0068] 4) ANFIS fuzzy neural network model - IWO's BiLSTM neural network model design

[0069] ANFIS fuzzy neural network model - IWO's BiLSTM neural network model is a series connection of ANFIS fuzzy neural network model and IWO's BiLSTM neural network model. The output of ANFIS fuzzy neural network model is used as the input of IWO's BiLSTM neural network model. ANFIS fuzzy neural network model organically combines neural network and fuzzy control, which can not only give play to the advantages of both but also make up for their respective shortcomings. The fuzzy membership function and fuzzy rules of ANFIS fuzzy neural network model are obtained by learning a large amount of known input pressure data. The biggest feature of ANFIS fuzzy neural network model is the data-based modeling method, rather than arbitrary given based on experience or intuition. This is especially important for systems whose characteristics are not fully understood or whose characteristics are very complex.

[0070] ANFIS fuzzy neural network model is a modeling method based on neural network and fuzzy logic. Its core idea is to integrate the rules of fuzzy reasoning and fuzzy sets into neural network. The mathematical model of ANFIS fuzzy neural network model is as follows:

[0071] output=ω1f1(x)+ω2f2(x)+…+ω n f n (x)(3)

[0072] Among them, f i (x) is the corresponding membership function of the input variable, ω i is the weight coefficient, and output is the output variable.

[0073] The membership function and weight coefficients are trained and optimized by back propagation algorithms and other methods, so that the ANFIS fuzzy neural network model can achieve the goal of adaptive learning and inference. The training process of the ANFIS fuzzy neural network model can be divided into two stages: forward propagation and backward propagation. In the forward propagation stage, the input data is fuzzified by the membership function, and then the fuzzy output is obtained by reasoning through fuzzy rules. In the backward propagation stage, the weights and parameters of the network are adjusted according to the difference between the target output and the actual output by error back propagation, thereby optimizing the performance of the model. The working mode of the ANFIS fuzzy neural network model is similar to that of the feedforward-back propagation type adaptive neural network. The ANFIS fuzzy neural network model uses FIS parameters as connection weights for encoding and optimization. In the forward process, it identifies the defuzzified consequent parameters by the least squares method, while in the reverse process, it uses the gradient descent method to adjust the antecedent parameters, adopts the Gaussian function as the membership function of the input information, and selects the linear function as the output membership function. The ANFIS fuzzy neural network model reduces the dimension of the search space in the gradient method and can also improve the convergence speed of the parameters of the ANFIS adaptive neural reasoning model.

[0074] 5) BiLSTM neural network model of IWO for triangular fuzzy numbers - ANFIS fuzzy neural network model design

[0075] The IWO BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers is a series connection of the IWO BiLSTM neural network model and the ANFIS fuzzy neural network model. The a, b and c output by the ANFIS fuzzy neural network model constitute a triangular fuzzy number (a, b, c). The design process of the IWO BiLSTM neural network model and the ANFIS fuzzy neural network model refers to the relevant design method of the embodiment of the present invention.

[0076] 2. Multi-point pressure detection model design

[0077] The multi-point pressure detection model includes an improved wolf pack algorithm to optimize the pressure fuzzy clusterer, a pressure processing module, an IWO BiLSTM neural network model, an IWO BiLSTM neural network model of triangular fuzzy numbers -ANFIS fuzzy neural network pressure classifier and a TDL beat delayer;

[0078] 1) Improved wolf pack algorithm to optimize the design of pressure fuzzy clusterer

[0079] As an emerging group intelligent optimization algorithm, the wolf pack algorithm has the advantages of few sensitive parameters, fast convergence speed and high precision. However, in actual situations, the wolf pack algorithm still has disadvantages such as "premature", complex calculation and many parameters. Therefore, an improved optimization algorithm is proposed based on the wolf pack algorithm, which mainly has three aspects: ① In order to make the initial solution better distributed in the solution space, the population initialization is improved; ② The adaptive step size factor is introduced to reduce the algorithm's sensitivity to parameters and improve the algorithm's convergence speed; ③ The perturbation strategy and interaction strategy are introduced to enable the algorithm to better perform global optimization and jump out of the local optimal solution.

[0080] The improved wolf pack algorithm can reduce the sensitivity of parameters and obtain the approximate solution of the global optimum. The steps of the improved wolf pack algorithm are as follows: Step 1: Initialize the number of artificial wolves, the maximum number of iterations, the maximum number of wanderings, the number of callings, and update the proportional factor. Use the best point set method to initialize the population, determine the fitness value of each artificial wolf, and sort and take the minimum position as the position of the head wolf. Step 2: Select the artificial wolf with the smaller fitness after sorting as the scout wolf, and perform wandering according to the updated wandering step length to find new prey and find the best prey from the prey until the position of the head wolf is updated or the maximum number of wanderings is reached. Step 3. Step 3: The head wolf initiates a call to summon the fierce wolf to its position, and selects the minimum fitness position among the current fierce wolf, the position after the call, and the position after the call as the new fierce wolf position. If the fitness is less than that of the head wolf, the head wolf position is updated, and the calling behavior is initiated with the current head wolf position until the fierce wolf reaches the maximum number of calls or the head wolf is updated. Step 4 is executed. Step 4: Calculate the new position of the artificial wolf participating in the siege behavior according to the adaptive siege step length. If the fitness value is less than the alpha wolf, update the alpha wolf position. Step 5: Sort all current artificial wolves, eliminate some artificial wolves with poor fitness values, and update the alpha wolf status if its fitness value is less than the current alpha wolf. This embodiment of the present invention cancels the initial judgment distance and uses the alpha wolf update or the maximum number of summons as the condition for entering the siege behavior.

[0081] The design process of fuzzy clusterer: Suppose a finite set X = {x1, x2, ... x n} is a set of n samples, C is the number of predetermined categories, m i (i=1,2,…c) is the center of each cluster, c represents the number of cluster centers; μ j (x i ) is the membership of the i-th sample to the j-th class, and the clustering criterion function is defined by the membership function:

[0082]

[0083] Among them, J is the current number of clusters, M is the fuzzy C partition matrix of X, V is the cluster center set of X, and the result of the fuzzy clusterer is to obtain M and V that minimize the clustering criterion function; ||x i -m j || is x i to m j b is the fuzzy weighted power exponent, which is a parameter that controls the fuzziness of the clustering results.

[0084] In the fuzzy C-means clustering method, the sum of the membership of the samples to each cluster is required to be 1, that is:

[0085]

[0086] The fuzzy clustering algorithm can be completed according to the following iterative steps: A. Set the number of clusters c and parameter b, the algorithm termination threshold ε, the number of iterations t = 1, and the maximum number of iterations allowed is t max ; B. Initialize each cluster center m i ; C. Calculate the membership function based on the current cluster center; D. Update the cluster centers of each type with the current membership function; E. Select a suitable matrix norm, if ||V(t+1)-V(t)||≤ε or t≥t max , stop the calculation; otherwise, t=t+1, return to step C.

[0087] The design process of the pressure processing module, the BiLSTM neural network model of IWO, the BiLSTM neural network model of IWO of triangular fuzzy numbers-ANFIS fuzzy neural network pressure classifier refers to the relevant design method of the embodiment of the present invention.

[0088] The embodiment of the present invention also provides a multi-point pressure safety detection device, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the multi-point pressure safety detection method described in the embodiment of the present invention is implemented. The multi-point pressure safety detection device has all the advantages of the multi-point pressure safety detection method.

[0089] like Figure 3As shown, an embodiment of the present invention also provides a multi-point pressure safety detection Internet of Things system, including pressure parameter measurement terminals 1-n, a pressure centralized controller, a pressure monitoring terminal, a pressure cloud platform server and a pressure monitoring mobile phone App. The pressure centralized controller is used to realize two-way information exchange between the pressure parameter measurement terminal, the pressure monitoring terminal and the pressure cloud platform server, and the pressure monitoring mobile phone App and the pressure cloud platform server realize two-way communication through the 5G network; the pressure parameter measurement terminal collects data of the object under test, and the pressure cloud platform server stores and manages the collected data; the management personnel can supervise the operating status of the object under test through the pressure monitoring terminal and the pressure monitoring mobile phone App.

[0090] like Figure 4 As shown in the figure, the pressure parameter measurement end is the parameter sensing terminal of the object under test, including pressure sensor, temperature sensor, flow sensor, flow velocity sensor, LoRa communication module and STM32 microprocessor. Each sensor is connected to the STM32 microprocessor through the corresponding conditioning circuit, and the LoRa communication module is connected to the STM32 microprocessor. The software of the pressure parameter measurement end mainly realizes the communication of the LoRa module and the acquisition and control of the parameters of the object under test. The software adopts C language programming, which has a high degree of compatibility, greatly improves the work efficiency of software design and development, and enhances the reliability, readability and portability of the program code.

[0091] like Figure 5 As shown, the pressure centralized controller includes a flow regulating device, a temperature regulating device, a flow rate regulating device, an STM32 microprocessor, a wireless communication module, a LoRa communication module and a USB interface; each regulating device is connected to the STM32 microprocessor through a corresponding D / A conversion circuit to adjust the medium flow, medium temperature and medium flow rate that affect the pressure of the detected object; the wireless communication module, the LoRa communication module and the USB interface are connected to the STM32 microprocessor respectively; the pressure centralized controller realizes two-way data interaction with the pressure cloud platform server through the wireless communication module. The detection parameters of the pressure parameter measurement end are uploaded to the pressure cloud platform server through the pressure centralized controller for storage and management, so as to realize the monitoring of the detected object. The pressure parameter measurement end and the pressure centralized controller build a self-organizing communication network through the LoRa communication module to realize two-way communication.

[0092] The pressure monitoring end uses an industrial control computer and is connected to the pressure centralized controller through a USB interface; the main functions of the pressure monitoring end include pressure communication parameter setting, pressure data analysis, pressure data management and multi-point pressure detection model, such as Figure 6 shown.

Claims

1. A multi-point pressure safety detection method, characterized in that: include: Get pressure sensor output 1-n; Construct a multi-point pressure detection model: The multi-point pressure detection model includes an improved wolf pack algorithm to optimize the pressure fuzzy clusterer, a pressure processing module 1, a pressure processing module 2, a pressure processing module 3, an IWO BiLSTM neural network model 1, an IWO BiLSTM neural network model 2, an IWO BiLSTM neural network model 3, an IWO BiLSTM neural network model 4, an IWO BiLSTM neural network model of triangular fuzzy numbers -ANFIS fuzzy neural network pressure classifier and a TDL beat delay device; The pressure sensor outputs 1-n are used as the input of the improved wolf pack algorithm to optimize the pressure fuzzy clusterer. The improved wolf pack algorithm to optimize the pressure fuzzy clusterer outputs three types of pressure sensor outputs, one type of pressure sensor output is used as the input of the pressure processing module 1 and the corresponding input of the BiLSTM neural network model 3 of IWO, another type of pressure sensor output is used as the input of the pressure processing module 2 and the corresponding input of the BiLSTM neural network model 1 of IWO, and another type of pressure sensor output is used as the input of the pressure processing module 3 and the corresponding input of the BiLSTM neural network model 4 of IWO; The output of the pressure processing module 1 is used as the corresponding input of the BiLSTM neural network model 2 of IWO and the BiLSTM neural network model 3 of IWO, the output of the pressure processing module 2 is used as the corresponding input of the BiLSTM neural network model 1 of IWO, and the output of the pressure processing module 3 is used as the corresponding input of the BiLSTM neural network model 1 and the BiLSTM neural network model 4; The outputs of IWO's BiLSTM neural network model 1 are used as the corresponding inputs of IWO's BiLSTM neural network model 2 and IWO's BiLSTM neural network model 3, respectively. The outputs of IWO's BiLSTM neural network model 3 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model 2 are used as the corresponding inputs of IWO's BiLSTM neural network model 4 and IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers, respectively. The outputs of IWO's BiLSTM neural network model 4 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network pressure classifier for triangular fuzzy numbers are used as the inputs of the TDL beat delay device, and the outputs of the TDL beat delay device are used as the corresponding inputs of IWO's BiLSTM neural network model 2, IWO's BiLSTM neural network model 3, and IWO's BiLSTM neural network model 4, respectively. The BiLSTM neural network model of IWO of triangular fuzzy numbers - the x, y and z output by the ANFIS fuzzy neural network pressure classifier constitute a triangular fuzzy number (x, y, z), and the pressure safety classification warning is carried out according to the triangular fuzzy number (x, y, z); The multi-point pressure safety detection method further includes: obtaining outputs of multiple temperature sensors, the temperature sensors being set at positions corresponding to the pressure sensors; The three pressure processing modules have the same structure, and the pressure processing modules include Adaline neural network model 1, Adaline neural network model 2, VMD variational mode decomposition model 1, VMD variational mode decomposition model 2, IWO's BiLSTM neural network model 5, IWO's BiLSTM neural network model 6, IWO's BiLSTM neural network model 11, IWO's BiLSTM neural network model 12, IWO's BiLSTM neural network model 13, IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22, IWO's BiLSTM neural network model 23, ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 1, ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 and IWO's BiLSTM neural network model of triangular fuzzy number-ANFIS fuzzy neural network model; The outputs of multiple pressure sensors are used as the corresponding inputs of Adaline neural network model 2. The outputs of Adaline neural network model 2 are used as the inputs of VMD variational mode decomposition model 2, IWO's BiLSTM neural network model 6 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. The outputs of IWO's BiLSTM neural network model 6 are used as the corresponding inputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. VMD variational mode decomposition model 2 outputs the high-frequency analysis of pressure. The quantity, intermediate frequency component and low frequency component are respectively used as the input of IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22 and IWO's BiLSTM neural network model 23, the outputs of IWO's BiLSTM neural network model 21, IWO's BiLSTM neural network model 22 and IWO's BiLSTM neural network model 23 are used as the corresponding inputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2, and the outputs of ANFIS fuzzy neural network model-IWO's BiLSTM neural network model 2 are used as the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers; The outputs of multiple temperature sensors corresponding to the multiple pressure sensors are used as the inputs of Adaline neural network model 1. The outputs of Adaline neural network model 1 are used as the inputs of VMD variational mode decomposition model 1, the inputs of IWO's BiLSTM neural network model 5 and the corresponding inputs of IWO's BiLSTM neural network model-ANFIS fuzzy neural network model of triangular fuzzy numbers. The high-frequency component, mid-frequency component and low-frequency component of the temperature output of VMD variational mode decomposition model 1 are used as the corresponding inputs of IWO's BiLSTM neural network model 11, IWO's BiLSTM neural network model 12 and IWO's BiLSTM neural network model 13 respectively. The output is used as the corresponding input of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO, the output of the BiLSTM neural network model 5 of IWO is used as the corresponding input of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO and the corresponding input of the BiLSTM neural network model 6 of IWO respectively, the output of the BiLSTM neural network model 1 of the ANFIS fuzzy neural network model-IWO is used as the corresponding input of the BiLSTM neural network model 2 of the ANFIS fuzzy neural network model-IWO, the output of the BiLSTM neural network model of IWO-ANFIS fuzzy neural network model of triangular fuzzy numbers is used as the corresponding input of the BiLSTM neural network model 2 of the ANFIS fuzzy neural network model-IWO and the BiLSTM neural network model 6 of IWO respectively; The a, b, and c output by the BiLSTM neural network model of IWO-ANFIS fuzzy neural network model of triangular fuzzy numbers constitute triangular fuzzy numbers (a, b, c) as the predicted values ​​of the outputs of multiple pressure sensors.

2. The multi-point pressure safety detection method according to claim 1, characterized in that: The pressure safety levels include very safe, safe, normal, unsafe and very unsafe.

3. The multi-point pressure safety detection method according to claim 2, characterized in that: The pressure safety level is determined according to the above table.

4. A multi-point pressure safety detection device, comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the computer program is executed by a processor, the multi-point pressure safety detection method according to any one of claims 1 to 3 is implemented.

5. A multi-point pressure safety detection Internet of Things system, used to implement the multi-point pressure safety detection method according to any one of claims 1 to 3, characterized in that: It includes pressure parameter measurement terminals 1-n, a pressure centralized controller, a pressure monitoring terminal, a pressure cloud platform server and a pressure monitoring mobile phone App. The pressure centralized controller is used to realize two-way information exchange between the pressure parameter measurement terminal, the pressure monitoring terminal and the pressure cloud platform server, and the pressure monitoring mobile phone App communicates with the pressure cloud platform server in a two-way manner. The pressure parameter measurement terminal collects data of the measured object, and the pressure cloud platform server stores and manages data. Through the pressure monitoring terminal and the pressure monitoring mobile phone App, the operating status of the monitored object can be supervised.

6. The multi-point pressure safety detection Internet of Things system according to claim 5 is characterized in that: The pressure parameter measurement end includes a pressure sensor, a temperature sensor, a flow sensor, a flow velocity sensor, a LoRa communication module and an STM32 microprocessor. Each sensor is connected to the STM32 microprocessor through a corresponding conditioning circuit, and the LoRa communication module is connected to the STM32 microprocessor.

7. The multi-point pressure safety detection Internet of Things system according to claim 6 is characterized in that: The pressure centralized controller includes a flow regulating device, a temperature regulating device, a flow rate regulating device, an STM32 microprocessor, a wireless communication module, a LoRa communication module and a USB interface; each regulating device is connected to the STM32 microprocessor through a corresponding D / A conversion circuit; the wireless communication module, the LoRa communication module and the USB interface are connected to the STM32 microprocessor respectively; the pressure centralized controller realizes two-way data interaction with the pressure cloud platform server through the wireless communication module.

8. The multi-point pressure safety detection Internet of Things system according to claim 7 is characterized in that: The pressure parameter measurement end and the pressure centralized controller build a self-organizing communication network through the LoRa communication module to achieve two-way communication.

9. The multi-point pressure safety detection Internet of Things system according to claim 7, characterized in that: The pressure monitoring end uses an industrial control computer and is connected to the pressure centralized controller through a USB interface; the pressure monitoring mobile phone app communicates with the pressure cloud platform server through the 5G network; the main functions of the pressure monitoring end include pressure communication parameter setting, pressure data analysis, pressure data management and multi-point pressure detection model.

Images