Factory environment monitoring system and method of integrated sensor network

Abstract

The invention relates to the technical field of factory management, in particular to a factory environment monitoring system and method of an integrated sensor network, and the method comprises the steps: collecting multi-dimensional environment data through a sensor in a factory, and designing a data collection strategy of event trigger sampling and adaptive sampling; a sensor self-organizing network protocol is constructed, and automatic discovery and dynamic networking of a sensor network are achieved; creating a link quality adaptive mechanism, and automatically adjusting the transmitting power or switching the channel when the link deteriorates by periodically detecting the link quality index; after feature extraction is carried out on the collected environment data, a machine learning model is constructed, and the environment change trend is predicted; an environment adjusting strategy is constructed, and the environment is adjusted in advance according to the predicted trend abnormity; according to the method, intelligent sensing of the factory environment can be realized, prediction and regulation are integrated, the environment management level is improved, and the production safety and efficiency are guaranteed.

Description

Technical Field

[0001] The present invention relates to the technical field of factory management, and in particular to a factory environment monitoring system and method integrating a sensor network. Background Art

[0002] Most existing factory environmental monitoring systems use wired sensors, which have high deployment costs and poor flexibility, making it difficult to achieve large-scale, multi-dimensional environmental monitoring. At the same time, traditional data collection strategies use fixed time intervals, do not consider the dynamic characteristics of environmental changes, and have low data collection efficiency.

[0003] In addition, the existing sensor network networking methods usually adopt static configuration, which cannot adapt to network topology changes such as node addition and reduction, and the network scalability is poor. In the process of wireless transmission, due to the complex and changeable industrial environment, the quality of wireless links is easily disturbed, resulting in poor data transmission reliability.

[0004] In terms of data processing and analysis, existing methods mainly rely on manual statistics and threshold comparison, lack real-time and intelligence levels, make it difficult to detect environmental change trends in a timely manner, lack early warning and proactive adjustment capabilities for abnormal situations, and may lead to production safety and efficiency problems.

[0005] The existing factory environmental monitoring still has problems such as poor system network adaptability and delayed data analysis. There is an urgent need for a new low-cost, flexible, efficient and intelligent environmental monitoring solution.

[0006] In view of this, the present invention proposes a factory environment monitoring system and method integrating a sensor network. Summary of the invention

[0007] To achieve the above objectives, the present invention provides a factory environment monitoring system and method integrated with a sensor network, and the specific technical solutions are as follows:

[0008] A factory environment monitoring method integrating a sensor network, comprising:

[0009] Collect multi-dimensional data of the factory environment through sensors and build adaptive data collection strategies, including event-triggered sampling strategies and adaptive sampling strategies;

[0010] Build a sensor self-organizing network strategy to automatically discover sensor networks and dynamically form networks;

[0011] Create a link quality adaptive mechanism to periodically detect wireless link signal indicators. When the link quality drops below the threshold, the transmit power is adaptively adjusted or the channel is switched.

[0012] Extract the features of environmental data collected by sensors, build and train machine learning models to predict the trend of environmental data changes, extract features from real-time collected environmental data, and predict the trend of environmental data changes through the trained machine learning models;

[0013] Build a factory environment change trend adjustment strategy to pre-adjust the factory environment when the environmental data change trend predicted by the machine learning model is abnormal.

[0014] Preferably, the sensors used in the factory include temperature sensors, humidity sensors, infrared optical sensors, acoustic sensors and gas sensors;

[0015] Construct an adaptive data acquisition strategy, which includes an event-triggered sampling strategy and an adaptive sampling strategy; define the adjustment step of the sensor sampling time interval as t 0 , the initial sampling interval is t; let the current sampling interval be t i , the current data is d i , the last data is d i-1 ;

[0016] The event-triggered sampling strategy includes: When the sensor sampling is triggered, the sampling data is updated; at the same time, the current sampling interval t i Value reduction t 0 , but not less than the set minimum sampling interval t min ; where δ (g) is the preset threshold of the g-th type of sensor, g∈[1,G], G is the total number of sensor types;

[0017] The adaptive sampling strategy includes: when m consecutive samplings satisfy When the current sampling interval t i The value increases 0 , but not greater than the set maximum sampling interval t max ;

[0018] Continuous m samplings satisfy When the current sampling interval t i Value reduction t 0 , but not less than the set minimum sampling interval t min ; Where m is the preset sensitivity parameter.

[0019] Preferably, a self-organizing network protocol is constructed, and each sensor is regarded as a network node, wherein the network node includes a parent node network composed of a network coordinator and a child node network composed of sensors;

[0020] Assign dynamic addresses to sensors, preset the network address pool in the network coordinator, calculate the network address assigned to each sensor, and perform network address conflict detection on the sensors;

[0021] Create automatic discovery and dynamic networking of sensors. After a new sensor is powered on, it first enters the network listening state and scans multiple channels for a predetermined time. If no network beacon frame is received, the new sensor acts as a network coordinator, creates a new network on an idle channel, and broadcasts beacon frames periodically. If the new sensor receives a network beacon frame, the sensor parses the network parameters in the beacon, joins the existing network, and requests an address assignment.

[0022] Preferably, a link quality evaluation indicator is defined, and a signal strength indicator RSSI is defined a , represents the signal strength received by sensor a; define the signal-to-noise ratio SNR a , represents the signal-to-noise ratio received by sensor a; defines the packet loss rate Represents sensor a to network coordinator a 0 The packet loss rate;

[0023] The link quality is periodically checked. Each network coordinator periodically broadcasts a link detection packet, which includes the node number, transmission power and timestamp. After receiving the link detection packet, the child node sensor of the network coordinator measures the signal strength indicator RSSI and the signal-to-noise ratio SNR, and sends the measurement results together with its own node number and timestamp to the parent node network coordinator. The parent node network coordinator a 0 After receiving the reply from the child node sensor a, calculate the packet loss rate Among them, N sent The parent node network coordinator a 0 The total number of packets sent to child sensor a, N received is the number of packets successfully received by child node sensor a;

[0024] Define link quality thresholds, including: Signal Strength Indicator Threshold RSSI th , signal-to-noise ratio threshold SNR th And the packet loss rate threshold PLR th ; When the signal strength indicator RSSI a <RSSI th When the signal-to-noise ratio SNR a <SNR th When the link signal-to-noise ratio is insufficient, When , the link packet loss rate is too high.

[0025] Preferably, the transmission power between the sensor and the network coordinator is adaptively adjusted, when the parent node network coordinator a 0 When it is detected that the link quality with the child node sensor a does not meet the quality threshold condition, the transmission power is increased;

[0026] Define the transmission power adjustment step ΔP, and set the initial transmission power to P 0 ; When RSSI is met a <RSSI th or SNR a <SNR th When the transmit power is increased by ΔP, if the link quality still does not meet the quality threshold after the transmit power is increased, it will continue to increase until the maximum transmit power P is reached. max ;

[0027] When the child node sensor a detects the connection with the parent node network coordinator a 0 When the link quality does not meet the quality threshold condition and the maximum transmission power has been reached, switch the channel;

[0028] Define candidate channel set C = {c 1 ,c 2 ,...,c K}, where K is the total number of candidate channels; child node sensor a randomly selects a candidate channel c from the candidate channels k , switch to the alternative channel for communication;

[0029] If the link quality still does not meet the quality threshold after switching channels, continue to switch randomly until all candidate channels are switched;

[0030] If all candidate channels cannot meet the quality threshold condition, the child node sensor a sends a signal to the parent node network coordinator a. 0 Sending network reorganization or node reassociation requests;

[0031] Construct a link quality recovery mechanism. When the child node sensor a detects that it is connected to the parent node network coordinator a 0 When the link quality is restored to above the quality threshold, the transmit power is gradually reduced, with each transmit power reduction of ΔP until the initial transmit power P is reached. 0 If the link quality drops again after the transmit power is reduced, the adaptive transmit power adjustment or channel switching process is re-executed.

[0032] Preferably, historical environmental data is obtained, and environmental data characteristics collected by sensors are defined, including: temperature data characteristics, humidity data characteristics, infrared optical data characteristics, acoustic data characteristics, and gas concentration data characteristics;

[0033] The temperature data features include: average temperature Temperature variance and temperature change rate T'; the humidity data characteristics include: average humidity Humidity variance and humidity change rate H'; the infrared optical data features include: average light intensity Light intensity variance and light intensity change rate I'; the acoustic data features include: average sound pressure level Sound pressure level variance And the sound pressure level change rate P'; the gas concentration data characteristics include: average gas concentration Gas concentration variance and the gas concentration change rate C';

[0034] The sliding window method is used to extract the time series features of historical environmental data, a machine learning model is built, and the long short-term memory network LSTM model is used to model and predict the environmental data.

[0035] Preferably, the input of the LSTM model is the extracted environmental data time series feature X=(x 1 ,x 2 ,...,x E ), where E is the number of time windows of the historical environmental feature dataset; the output of the LSTM model is the predicted value of environmental data in the future F time windows in is the predicted value of environmental data in the f-th time window, f∈[1,F];

[0036] The acquired historical environmental feature data is divided into a training set and a test set. The training set is used to train the LSTM model, and the test set is used to evaluate the prediction performance of the LSTM model. During the training process, the mean square error is used as the loss function.

[0037] Use the trained LSTM model to predict the real-time collected environmental data. The specific steps are as follows: Preprocess and extract features of the real-time collected environmental data to obtain the windowed feature vector x of the current time series. t , x t Input into the LSTM model to obtain the predicted value of environmental data in the future F time windows The predicted value The data is compared with the set abnormal trend threshold of the environmental data change trend to determine whether the environmental data change trend is abnormal.

[0038] Preferably, the abnormal trend threshold of the environmental data change trend is defined, including the temperature change trend threshold ΔT th , Humidity change trend threshold ΔH th, infrared light intensity change trend threshold ΔI th , Sound pressure level change trend threshold ΔP th And the gas concentration change trend threshold ΔC th ;

[0039] Use the trained LSTM model to predict the environmental data of the next F time windows and obtain the predicted value sequence in They respectively represent the temperature prediction value, humidity prediction value, infrared light intensity prediction value, sound pressure level prediction value and gas concentration prediction value of the f-th time window.

[0040] Preferably, the change trend of the environmental data within the time window is calculated, including: temperature change trend Humidity trend Infrared light intensity change trend Sound pressure level change trend Gas concentration trend

[0041] Anomaly detection for the predicted trend sequence: The predicted temperature change trend is abnormal; if The predicted humidity change trend is abnormal; if The predicted infrared light intensity change trend is abnormal; if The predicted sound pressure level change trend is abnormal; if Then the predicted gas concentration change trend is abnormal;

[0042] When any abnormal value appears in the predicted change trend sequence, the factory environment change trend adjustment strategy is executed to restore the corresponding environmental data change trend to within the abnormal trend threshold;

[0043] After the environment is adjusted, the LSTM model is used to predict future environmental data. If the predicted change trend is still abnormal and the alarm conditions are met, an alarm is issued. The alarm conditions are:

[0044]

[0045] Where I(·) is the indicator function; represents the change trend of environmental data in the predicted f-th time window, X refers to any type of environmental data, ΔX th is the corresponding abnormal trend threshold; α is the abnormal proportion threshold.

[0046] A factory environment monitoring system integrated with a sensor network, which is used in the factory environment monitoring method integrated with a sensor network, comprises: a data acquisition strategy module, an automatic networking module, a network signal adjustment module, an environment data prediction module and an environment adjustment module;

[0047] The data collection strategy module collects multi-dimensional data of the factory environment through sensors and constructs an adaptive data collection strategy, including an event-triggered sampling strategy and an adaptive sampling strategy;

[0048] The automatic networking module is used to build a sensor self-organizing network strategy, adaptively discover sensor networks and dynamically network;

[0049] The network signal adjustment module is used to create a link quality adaptive mechanism, periodically detect wireless link signal indicators, and automatically adjust the transmission power or switch channels when the link quality drops below a threshold;

[0050] The environmental data prediction module is used to extract the characteristics of environmental data collected by sensors, build and train a machine learning model to predict the trend of environmental data changes, extract characteristics from environmental data collected in real time, and predict the trend of environmental data changes through the trained machine learning model;

[0051] The environmental adjustment module is used to construct a factory environment change trend adjustment strategy, and pre-adjust the factory environment when the environmental data change trend predicted by the machine learning model is abnormal.

[0052] Beneficial effects of the present invention: The present invention comprehensively collects factory environmental data, efficiently obtains key information, optimizes data collection strategies, reduces energy consumption, and provides data support for environmental monitoring.

[0053] The present invention simplifies deployment and maintenance, improves network flexibility and scalability, and optimizes sensor network performance by automatically forming a sensor network.

[0054] The present invention ensures data transmission reliability, improves sensor network robustness, and ensures real-time reporting of key data by adaptively adjusting link quality.

[0055] The present invention predicts environmental change trends, discovers anomalies in advance, provides a basis for environmental adjustment decisions, and improves the predictability and initiative of environmental management.

[0056] The present invention automatically adjusts environmental parameters according to predictions, reduces the impact of anomalies on production, improves the level of environmental control, and ensures production safety and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1A flow chart of a factory environment monitoring method with an integrated sensor network provided by the present invention;

[0058] Figure 2 A structural diagram of a factory environment monitoring system with an integrated sensor network provided by the present invention. DETAILED DESCRIPTION

[0059] In order to better understand the present invention, a more detailed description will be made of various aspects of the present invention with reference to the accompanying drawings. It should be understood that these detailed descriptions are only descriptions of exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention in any way. Throughout the specification, the same reference numerals refer to the same elements. The expression "and / or" includes any and all combinations of one or more of the associated listed items.

[0060] It should also be understood that expressions such as "comprises", "including", "having", "includes" and / or "comprising" are open rather than closed expressions in this specification, which indicate the presence of the stated features, elements and / or components, but do not exclude the presence of one or more other features, elements, components and / or combinations thereof. In addition, when expressions such as "at least one of..." appear after a list of listed features, they modify the entire list of features rather than just the individual elements in the list. In addition, when describing embodiments of the present invention, "may" is used to mean "one or more embodiments of the present invention". And, the term "exemplary" is intended to refer to an example or illustration.

[0061] Unless otherwise defined, all words (including engineering terms and scientific and technological terms) used in this article have the same meaning as those commonly understood by ordinary technicians in the field to which the present invention belongs. It should also be understood that unless there is a clear explanation in the present invention, the words defined in the commonly used dictionary should be interpreted as having the same meaning as their meaning in the context of the relevant technology, and should not be interpreted in an idealized or overly formal sense.

[0062] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0063] Example 1

[0064] Reference Figure 1 , which is the first embodiment of the present invention, provides a factory environment monitoring method integrating a sensor network.

[0065] S1: Collect multi-dimensional data of the factory environment through sensors and build adaptive data collection strategies, including event-triggered sampling strategies and adaptive sampling strategies.

[0066] In the factory, sensors including temperature sensors, humidity sensors, infrared optical sensors, acoustic sensors and gas sensors are used; various types of sensors are deployed at corresponding locations as required, and the sensors transmit the collected data through wireless networks.

[0067] The temperature and humidity sensors are high-precision temperature and humidity sensors based on digital interfaces (such as I2C, SPI), which are arranged in different areas of the workshop. The number is N. 1 The infrared optical sensor is a non-cooled infrared sensor, such as the Melexis MLX90614 series, with a detection range of -40 to 125°C; it is arranged near key monitoring equipment, and the number is N 2 The acoustic sensor is a MEMS microphone with a frequency response range of 20Hz to 20kHz and a sensitivity of more than -42dB; it is arranged near the noise source, and the number is N 3 The gas sensor is a MOS or MOX gas sensor, which can detect VOC, CO and NO. 2 ; Arranged near the exhaust port and ventilation port, the number is N 4 indivual.

[0068] Construct an adaptive data acquisition strategy, which includes an event-triggered sampling strategy and an adaptive sampling strategy; define the adjustment step of the sensor sampling time interval as t 0 , the initial sampling interval is t; let the current sampling interval be t i , the current data is d i , the last data is d i-1 .

[0069] The event-triggered sampling strategy includes: When the sensor sampling is triggered, the sampling data is updated; at the same time, the current sampling interval t i Value reduction t 0 , but not less than the set minimum sampling interval t min ; where δ (g) is the preset threshold of the g-th type of sensor, g∈[1,G], and G is the total number of sensor types.

[0070] The adaptive sampling strategy includes: when m consecutive samplings satisfy When the current sampling interval t i The value increases 0 , but not greater than the set maximum sampling interval t max .

[0071] Continuous m samplings satisfy When the current sampling interval t i Value reduction t 0, but not less than the set minimum sampling interval t min ; Where m is the preset sensitivity parameter.

[0072] Step S1 uses sensors to comprehensively collect multi-dimensional data of the factory environment, builds event-triggered sampling strategies and adaptive sampling strategies, and can efficiently obtain factory environment data, reduce energy consumption, and provide a data basis for environmental monitoring and anomaly detection.

[0073] S2: Build a sensor self-organizing network strategy to automatically discover sensor networks and dynamically form networks.

[0074] A self-organizing network protocol is constructed, and each sensor is regarded as a network node. The network node includes a parent node network composed of a network coordinator and a child node network composed of sensors.

[0075] Assign dynamic addresses to sensors and preset network address pools in the network coordinator. The address range is [A min ,A max ], where A min and A max is the minimum and maximum value of a 16-bit network address; when a new sensor joins the network, it sends an address request to the network coordinator, and the address request carries a random number R s After the network coordinator receives the address request, it will s and a random number R stored in itself c Perform XOR operation to get the hash value H s : in, Represents the exclusive OR operator.

[0076] H s For address pool length L = A max -A min +1 modulo, get the address A assigned to the new sensor s : A s =A min +(H s mod L), where mod represents the modulo operation.

[0077] The network coordinator will A s Assign to the new sensor and set R s Stored for subsequent address conflict detection.

[0078] Build the automatic discovery and dynamic networking functions of sensors. After the new sensor is powered on, it first enters the network listening state and scans multiple channels for a predetermined time. If no network beacon frame is received, the new sensor acts as a network coordinator, creates a new network on an idle channel, and broadcasts beacon frames periodically. If the new sensor receives a network beacon frame, the sensor parses the network parameters in the beacon, joins the existing network, and requests address allocation.

[0079] After a new sensor successfully joins the network, it periodically exchanges link quality information with neighboring sensors; define the link quality indicator Q ab , represents the link quality between sensor a and sensor b, and its value range is [0, 1]; Q ab It is calculated by sensor a based on the received signal strength indication (RSSI) and signal-to-noise ratio (SNR) measurements. ab Below the preset threshold Q th When , sensor a actively disconnects from sensor b.

[0080] Step S2 builds a sensor self-organizing network protocol to achieve automatic discovery and dynamic networking of sensor networks, simplify network deployment and maintenance, improve network flexibility and scalability, and introduce link quality evaluation to adaptively optimize network performance.

[0081] S3: Create a link quality adaptive mechanism to periodically detect wireless link signal indicators. When the link quality drops below the threshold, the transmission power is adaptively adjusted or the channel is switched.

[0082] Define link quality evaluation indicators and define signal strength indication (RSSI): RSSI a Indicates the signal strength received by sensor a; the unit is dBm; definition of signal-to-noise ratio (SNR): SNR a Represents the signal-to-noise ratio received by sensor a in dB; defines the packet loss rate (PLR): Represents sensor a to network coordinator a 0 The packet loss rate ranges from [0 to 1].

[0083] The link quality is periodically checked. Each network coordinator periodically broadcasts a link detection packet, which includes the node number, transmission power and timestamp. After receiving the link detection packet, the child node sensor of the network coordinator measures the signal strength indicator RSSI and the signal-to-noise ratio SNR, and sends the measurement results together with its own node number and timestamp to the parent node network coordinator. The parent node network coordinator a 0 After receiving the reply from the child node sensor a, calculate the packet loss rate Among them, N sentThe parent node network coordinator a 0 The total number of packets sent to child sensor a, N received is the number of packets successfully received by child node sensor a.

[0084] Define link quality thresholds, including: Signal Strength Indicator Threshold RSSI th , signal-to-noise ratio threshold SNR th And the packet loss rate threshold PLR th .

[0085] When the signal strength indicator RSSI a <RSSI th When the signal-to-noise ratio SNR a <SNR th When the packet loss rate is When , the link packet loss rate is too high.

[0086] Adaptively adjust the transmission power between the sensor and the network coordinator. 0 When it is detected that the link quality with the child node sensor a does not meet the quality threshold condition, the transmission power is increased.

[0087] Define the transmission power adjustment step ΔP, in dBm; let the initial transmission power be P 0 ; When RSSI is met a <RSSI th or SNR a <SNR th When the transmit power is increased by ΔP, if the link quality still does not meet the quality threshold after the transmit power is increased, it will continue to increase until the maximum transmit power P is reached. max .

[0088] When the child node sensor a detects the connection with the parent node network coordinator a 0 When the link quality does not meet the quality threshold condition and the maximum transmit power has been reached, the channel is switched.

[0089] Define candidate channel set C = {c 1 ,c 2 ,...,c K}, where K is the total number of candidate channels; child node sensor a randomly selects a candidate channel c from the candidate channels k , switch to the alternative channel for communication.

[0090] If the link quality still does not meet the quality threshold condition after switching channels, random switching will continue until all candidate channels are switched.

[0091] If all candidate channels cannot meet the quality threshold condition, the child node sensor a sends a signal to the parent node network coordinator a. 0 Send a network reorganization or node reassociation request.

[0092] Construct a link quality recovery mechanism. When the child node sensor a detects that it is connected to the parent node network coordinator a 0 When the link quality is restored to above the quality threshold, the transmission power is gradually reduced to reduce energy consumption, and the transmission power is reduced by ΔP each time until the initial transmission power P is reached. 0 If the link quality drops again after the transmit power is reduced, the adaptive transmit power adjustment or channel switching process is re-executed.

[0093] Step S3 creates a link quality adaptive mechanism. By periodically detecting wireless link signal indicators, it determines whether the link quality drops below the threshold, automatically adjusts the transmission power or switches channels, ensures the reliability of data transmission, and improves the robustness of the wireless sensor network.

[0094] S4: Extract the features of environmental data collected by sensors, build and train a machine learning model to predict the changing trend of environmental data, extract features from the environmental data collected in real time, and predict the changing trend of environmental data through the trained machine learning model.

[0095] Acquire historical environmental data and define the environmental data characteristics collected by sensors, including temperature data characteristics, humidity data characteristics, infrared optical data characteristics, acoustic data characteristics, and gas concentration data characteristics.

[0096] The temperature data features include: average temperature Temperature variance and temperature change rate T'; the humidity data characteristics include: average humidity Humidity variance and humidity change rate H'; the infrared optical data features include: average light intensity Light intensity variance and light intensity change rate I'; the acoustic data features include: average sound pressure level Sound pressure level variance And the sound pressure level change rate P'; the gas concentration data characteristics include: average gas concentration Gas concentration variance And the gas concentration change rate C'.

[0097] The acquired historical environmental data is preprocessed, including data cleaning, outlier processing and data normalization.

[0098] Use the sliding window method to extract the time series features of historical environmental data; build a machine learning model and use the long short-term memory network LSTM model to model and predict environmental data.

[0099] The input of the LSTM model is the extracted environmental data time series feature X = (x 1 ,x 2 ,...,x E ), where E is the number of time windows; the output of the LSTM model is the predicted value of environmental data for the next F time windows in is the predicted value of environmental data in the f-th time window, f∈[1,F];

[0100] The acquired historical environmental feature data is divided into a training set and a test set. The training set data is used to train the LSTM model, and the test set data is used to evaluate the prediction performance of the LSTM model. During the training process, the mean square error (MSE) is used as the loss function: Where Y = (y 1 ,y 2 ,...,y F ) is the actual value of the real environmental data.

[0101] Use the trained LSTM model to predict the real-time collected environmental data. The specific steps are as follows: Preprocess and extract features of the real-time collected environmental data to obtain the feature vector x of the current time window t ; x t Input into the LSTM model to obtain the predicted value of environmental data in the future F time windows The predicted value The data is compared with the set abnormal trend threshold of the environmental data change trend to determine whether the environmental data change trend is abnormal.

[0102] Step S4 extracts the characteristics of environmental data collected by sensors and uses machine learning models to predict the changing trends of environmental data. This can detect abnormal environmental change trends in advance, provide a basis for environmental adjustment decisions, and improve the predictability and initiative of factory environmental management.

[0103] S5: Construct a factory environment change trend adjustment strategy to pre-adjust the factory environment when the environmental data change trend predicted by the machine learning model is abnormal.

[0104] Define abnormal trend thresholds for environmental data change trends, including temperature change trend threshold ΔT th , Humidity change trend threshold ΔH th , infrared light intensity change trend threshold ΔI th, Sound pressure level change trend threshold ΔP th And the gas concentration change trend threshold ΔC th .

[0105] Use the trained LSTM model to predict the environmental data of the next F time windows and obtain the predicted value sequence in They respectively represent the temperature prediction value, humidity prediction value, infrared light intensity prediction value, sound pressure level prediction value and gas concentration prediction value of the f-th time window.

[0106] Calculate the change trend of the predicted value series within the predicted time window: temperature change trend Humidity trend Infrared light intensity change trend Sound pressure level change trend Gas concentration trend Where Δt is the time interval between adjacent prediction time windows.

[0107] Anomaly detection for the predicted trend sequence: The predicted temperature change trend is abnormal; if The predicted humidity change trend is abnormal; if ΔI th , then the predicted infrared light intensity change trend is abnormal; if The predicted sound pressure level change trend is abnormal; if The predicted gas concentration change trend is abnormal.

[0108] When any abnormal value appears in the predicted change trend sequence, the factory environment change trend adjustment strategy is executed to restore the corresponding environmental data change trend to within the abnormal trend threshold.

[0109] When the temperature change trend is abnormal, adjust the factory environment temperature through air-conditioning equipment to restore the temperature change trend to a normal range; when the humidity change trend is abnormal, start the dehumidifier or humidifier to adjust the factory environment humidity to restore the humidity change trend to a normal range; when the infrared light intensity change trend is abnormal, check and repair the infrared light source equipment, adjust the factory environment infrared light intensity to restore its change trend to a normal range; when the sound pressure level change trend is abnormal, check and repair the sound source equipment, adjust the factory environment noise level to restore its change trend to a normal range; when the gas concentration change trend is abnormal, start the ventilation system to adjust the factory environment gas concentration to restore its change trend to a normal range.

[0110] After the environment is adjusted, the LSTM model is used to predict future environmental data. If the predicted change trend is still abnormal and meets the alarm conditions, an alarm is issued to the factory manager, indicating that the abnormal environmental change trend cannot be adjusted automatically and requires manual inspection and processing. The alarm information includes the time when the abnormality occurred, the abnormality type (temperature, humidity, infrared light intensity, sound pressure level or gas concentration change trend) and the degree of abnormality (the magnitude of the predicted change trend value exceeding the abnormal trend threshold). The alarm conditions are:

[0111]

[0112] Where I(·) is the indicator function, which takes 1 when the absolute value of the predicted change trend exceeds the corresponding abnormal trend threshold, otherwise it takes 0; represents the change trend of environmental data in the predicted f-th time window, X refers to any type of environmental data, ΔX th is the corresponding abnormal trend threshold; α is the abnormal proportion threshold.

[0113] Step S5 constructs a factory environment change trend adjustment strategy. According to the abnormal environment change trend predicted by the machine learning model, the environmental parameters are adjusted in time through automated adjustment equipment to restore the environment to normal, reduce the impact of abnormal environment on production, and improve the factory environment control level.

[0114] Example 2

[0115] Reference Figure 2 , which is a second embodiment of the present invention, provides a factory environment monitoring system integrating a sensor network.

[0116] The system comprises: a data acquisition strategy module, an automatic networking module, a network signal adjustment module, an environmental data prediction module and an environmental adjustment module.

[0117] The data collection strategy module collects multi-dimensional data of the factory environment through sensors and constructs an adaptive data collection strategy, including an event-triggered sampling strategy and an adaptive sampling strategy.

[0118] The automatic networking module is used to construct a sensor self-organizing network strategy, automatically discover sensor networks and perform dynamic networking.

[0119] The network signal adjustment module is used to create a link quality adaptive mechanism, periodically detect wireless link signal indicators, and adaptively adjust the transmission power or switch channels when the link quality drops below a threshold.

[0120] The environmental data prediction module is used to extract the characteristics of environmental data collected by sensors, build and train a machine learning model to predict the changing trend of environmental data, extract characteristics of environmental data collected in real time, and predict the changing trend of environmental data through the trained machine learning model.

[0121] The environmental adjustment module is used to construct a factory environment change trend adjustment strategy, and pre-adjust the factory environment when the environmental data change trend predicted by the machine learning model is abnormal.

[0122] In addition, the parts of the above technical solutions provided in the embodiments of the present invention that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0123] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A factory environment monitoring method integrating a sensor network, characterized in that: include: Collect multi-dimensional data of the factory environment through sensors and build adaptive data collection strategies, including event-triggered sampling strategies and adaptive sampling strategies; Build a sensor self-organizing network strategy to automatically discover sensor networks and dynamically form networks; Create a link quality adaptive mechanism to periodically detect wireless link signal indicators. When the link quality drops below the threshold, the transmit power is adaptively adjusted or the channel is switched. Extract the features of environmental data collected by sensors, build and train machine learning models to predict the trend of environmental data changes, extract features from real-time collected environmental data, and predict the trend of environmental data changes through the trained machine learning models; Build a factory environment change trend adjustment strategy to pre-adjust the factory environment when the environmental data change trend predicted by the machine learning model is abnormal.

2. A factory environment monitoring method with integrated sensor network according to claim 1, characterized in that: In the factory, sensors include temperature sensors, humidity sensors, infrared optical sensors, acoustic sensors, and gas sensors; Construct an adaptive data acquisition strategy, which includes an event-triggered sampling strategy and an adaptive sampling strategy; define the adjustment step of the sensor sampling time interval as t0, and the initial sampling interval as t; set the current sampling interval as t i , the current data is d i , the last data is d i-1 ; The event-triggered sampling strategy includes: When the sensor sampling is triggered, the sampling data is updated; at the same time, the current sampling interval t i The value is reduced by t0, but not less than the set minimum sampling interval t min ; where δ (g) is the preset threshold of the g-th type of sensor, g∈[1,G], G is the total number of sensor types; The adaptive sampling strategy includes: when m consecutive samplings satisfy When the current sampling interval t i The value increases by t0, but is not greater than the set maximum sampling interval t max ; Continuous m samplings satisfy When the current sampling interval t i The value is reduced by t0, but not less than the set minimum sampling interval t min ; Where m is the preset sensitivity parameter.

3. A factory environment monitoring method with integrated sensor network according to claim 2, characterized in that: Construct a self-organizing network protocol, and regard each sensor as a network node, wherein the network node includes a parent node network composed of a network coordinator and a child node network composed of sensors; Assign dynamic addresses to sensors, preset the network address pool in the network coordinator, calculate the network address assigned to each sensor, and perform network address conflict detection on the sensors; Create automatic discovery and dynamic networking of sensors. After a new sensor is powered on, it first enters the network listening state and scans multiple channels for a predetermined time. If no network beacon frame is received, the new sensor acts as a network coordinator, creates a new network on an idle channel, and broadcasts beacon frames periodically. If the new sensor receives a network beacon frame, the sensor parses the network parameters in the beacon, joins the existing network, and requests an address assignment.

4. A factory environment monitoring method with integrated sensor network according to claim 3, characterized in that: Define link quality evaluation indicators and define signal strength indication RSSI a , represents the signal strength received by sensor a; define the signal-to-noise ratio SNR a , represents the signal-to-noise ratio received by sensor a; Defining Packet Loss Rate represents the packet loss rate from sensor a to network coordinator a0; Periodically check the link quality. Each network coordinator periodically broadcasts a link detection packet, which includes the node number, transmission power and timestamp. After receiving the link detection packet, the child node sensor of the network coordinator measures the signal strength indication RSSI and signal-to-noise ratio SNR, and sends the measurement result together with its own node number and timestamp to the parent node network coordinator; After the parent node network coordinator a0 receives the reply from the child node sensor a, it calculates the packet loss rate Among them, N sent is the total number of packets sent from the parent node network coordinator a0 to the child node sensor a, N received is the number of packets successfully received by child node sensor a; Define link quality thresholds, including: Signal Strength Indicator Threshold RSSI th , signal-to-noise ratio threshold SNR th And the packet loss rate threshold PLR th ; When the signal strength indicator RSSI a <RSSI th When the signal-to-noise ratio SNR a <SNR th When the packet loss rate is When , the link packet loss rate is too high.

5. The factory environment monitoring method of integrated sensor network according to claim 4, characterized in that: Adaptively adjust the transmission power between the sensor and the network coordinator. When the parent node network coordinator a0 detects that the link quality with the child node sensor a does not meet the quality threshold condition, increase the transmission power; Define the transmission power adjustment step ΔP, set the initial transmission power to P0; when the RSSI a <RSSI th or SNR a <SNR th When the transmit power is increased by ΔP, if the link quality still does not meet the quality threshold after the transmit power is increased, it will continue to increase until the maximum transmit power P is reached. max ; When the child node sensor a detects that the link quality with the parent node network coordinator a0 does not meet the quality threshold condition and has reached the maximum transmission power, it switches the channel; Define the candidate channel set C = {c1, c2, ..., c K }, where K is the total number of candidate channels; child node sensor a randomly selects a candidate channel c from the candidate channels k , switch to the alternative channel for communication; If the link quality still does not meet the quality threshold after switching channels, continue to switch randomly until all candidate channels are switched; If all candidate channels cannot meet the quality threshold condition, the child node sensor a sends a network reorganization or node reassociation request to the parent node network coordinator a0; A link quality recovery mechanism is constructed. When the child node sensor a detects that the link quality with the parent node network coordinator a0 has recovered to above the quality threshold, the transmit power is gradually reduced, and the transmit power is reduced by ΔP each time until the initial transmit power P0 is reached; if the link quality drops again after reducing the transmit power, the adaptive transmit power adjustment or channel switching process is re-executed.

6. A factory environment monitoring method with integrated sensor network according to claim 5, characterized in that: Obtain historical environmental data and define the environmental data characteristics collected by sensors, including: temperature data characteristics, humidity data characteristics, infrared optical data characteristics, acoustic data characteristics, and gas concentration data characteristics; The temperature data features include: average temperature Temperature variance and temperature change rate T'; the humidity data characteristics include: average humidity Humidity variance and humidity change rate H'; the infrared optical data features include: average light intensity Light intensity variance and light intensity change rate I'; the acoustic data features include: average sound pressure level Sound pressure level variance And the sound pressure level change rate P'; the gas concentration data characteristics include: average gas concentration Gas concentration variance and the gas concentration change rate C'; The sliding window method is used to extract the time series features of historical environmental data, a machine learning model is built, and the long short-term memory network LSTM model is used to model and predict the environmental data.

7. A factory environment monitoring method with integrated sensor network according to claim 6, characterized in that: The input of the LSTM model is the extracted environmental data time series feature X = (x1, x2, ..., x E ), where E is the number of time windows of the historical environmental feature dataset; the output of the LSTM model is the predicted value of environmental data in the future F time windows in is the predicted value of environmental data in the f-th time window, f∈[1,F]; The acquired historical environmental feature data is divided into a training set and a test set. The training set is used to train the LSTM model, and the test set is used to evaluate the prediction performance of the LSTM model. During the training process, the mean square error is used as the loss function. Use the trained LSTM model to predict the real-time collected environmental data. The specific steps are as follows: Preprocess and extract features of the real-time collected environmental data to obtain the windowed feature vector x of the current time series. t , x t Input into the LSTM model to obtain the predicted value of environmental data in the future F time windows The predicted value The data is compared with the set abnormal trend threshold of the environmental data change trend to determine whether the environmental data change trend is abnormal.

8. The factory environment monitoring method of integrated sensor network according to claim 7, characterized in that: Define abnormal trend thresholds for environmental data change trends, including temperature change trend threshold ΔT th , Humidity change trend threshold ΔH th , infrared light intensity change trend threshold ΔI th , Sound pressure level change trend threshold ΔP th And the gas concentration change trend threshold ΔC th ; Use the trained LSTM model to predict the environmental data of the next F time windows and obtain the predicted value sequence in They respectively represent the temperature prediction value, humidity prediction value, infrared light intensity prediction value, sound pressure level prediction value and gas concentration prediction value of the f-th time window.

9. A factory environment monitoring method with integrated sensor network according to claim 8, characterized in that: Calculate the changing trend of environmental data within the time window, including: temperature changing trend Humidity trend Infrared light intensity change trend Sound pressure level change trend Gas concentration trend Anomaly detection for the predicted trend sequence: The predicted temperature change trend is abnormal; if The predicted humidity change trend is abnormal; if ΔI th , then the predicted infrared light intensity change trend is abnormal; if The predicted sound pressure level change trend is abnormal; if Then the predicted gas concentration change trend is abnormal; When any anomaly occurs in the predicted change trend sequence, the factory environment change trend adjustment strategy is executed to restore the corresponding environmental data change trend to within the abnormal trend threshold; After the environment is adjusted, the LSTM model is used to predict future environmental data. If the predicted change trend is still abnormal and the alarm conditions are met, an alarm is issued. The alarm conditions are: Where I(·) is the indicator function; represents the change trend of environmental data in the predicted f-th time window, X refers to any type of environmental data, ΔX th is the corresponding abnormal trend threshold; α is the abnormal proportion threshold.

10. A factory environment monitoring system integrated with a sensor network, used to implement a factory environment monitoring method integrated with a sensor network as claimed in any one of claims 1 to 9, characterized in that: include: Data collection strategy module, automatic networking module, network signal adjustment module, environmental data prediction module and environmental adjustment module; The data collection strategy module collects multi-dimensional data of the factory environment through sensors and constructs an adaptive data collection strategy, including an event-triggered sampling strategy and an adaptive sampling strategy; The automatic networking module is used to build a sensor self-organizing network strategy, automatically discover sensor networks and perform dynamic networking; The network signal adjustment module is used to create a link quality adaptive mechanism, periodically detect wireless link signal indicators, and adaptively adjust the transmission power or switch channels when the link quality drops below a threshold; The environmental data prediction module is used to extract the characteristics of environmental data collected by sensors, build and train a machine learning model to predict the trend of environmental data changes, extract characteristics from environmental data collected in real time, and predict the trend of environmental data changes through the trained machine learning model; The environmental adjustment module is used to construct a factory environment change trend adjustment strategy, and pre-adjust the factory environment when the environmental data change trend predicted by the machine learning model is abnormal.

Images