Water quality monitoring station hardware polling and monitoring method combined with Internet of Things

Abstract

The invention discloses a water quality monitoring station hardware inspection and monitoring method combined with the Internet of Things, and relates to the technical field of hardware detection. According to the water quality monitoring station hardware routing inspection and monitoring method combined with the Internet of Things, hardware operation data of a plurality of time points of a set water quality monitoring station are continuously acquired and uploaded through the Internet of Things, operation robust characteristic values of all the time points are analyzed, and monitoring and early warning are carried out; analyzing a hardware operation robust evolution characteristic value based on the hardware operation robust characteristic value of each time point; the method comprises the following steps: acquiring a hardware appearance image, extracting an appearance damage aggregation characteristic value by combining a pre-trained appearance damage identification model, fusing the appearance damage aggregation characteristic value with an operation robust evolution characteristic, and finally analyzing a hardware inspection fault hidden danger characteristic value. Therefore, unified evaluation of the operation state and the appearance damage is realized, the accuracy of the inspection result is improved, and continuous and stable operation of the monitoring station is effectively guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hardware detection, in particular to a water quality monitoring station hardware inspection and monitoring method combined with the Internet of Things. BACKGROUND

[0002] The water quality monitoring station is an important part of the water environment monitoring system, usually arranged in key areas such as rivers, lakes and reservoirs, for realizing real-time acquisition and remote transmission of water quality indicators. The existing water quality monitoring station generally includes a sampling pump, a power supply device and other hardware. With the development of Internet of Things technology, the water quality monitoring station gradually has the ability of remote data acquisition and wireless transmission, and can continuously monitor without manual intervention. However, these hardware are prone to mechanical wear and tear, electrical fluctuations and other factors during long-term operation, and their operation stability directly affects the reliability of the overall work of the monitoring station.

[0003] Among them, the limitations of the prior art at least include the following problems: the prior art lacks unified evaluation of the running state and appearance damage, and it is difficult to fully reflect the overall health status of the hardware, so it is difficult to identify potential hidden dangers in time. At the same time, the hardware of the water quality monitoring station not only has to withstand the influence of complex working conditions, but also is affected by external environment, for example, the mud and particles carried in the water body may cause surface abrasion, the humid or corrosive environment may cause local liquid penetration and rust, and the chemical composition in the water may form scale deposition, thereby it is difficult to form an overall and forward-looking judgment, and the risk of maintenance lag and equipment failure is increased. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a water quality monitoring station hardware inspection and monitoring method combined with the Internet of Things, which solves the problem that the prior art lacks unified evaluation of the running state and appearance damage, and it is difficult to accurately identify hidden dangers, resulting in maintenance lag risk.

[0005] In order to achieve the above object, the application realizes the above object by the following technical scheme: the hardware inspection and monitoring method of the water quality monitoring station combined with the Internet of Things, comprising the following steps: continuously acquiring hardware operation data of a plurality of time points of a set water quality monitoring station, and uploading based on the Internet of Things; based on the hardware operation data of each time point of the set water quality monitoring station after uploading, analyzing the hardware operation robustness characteristic value of the corresponding time point, and monitoring and warning the corresponding time point of the set water quality monitoring station; based on the hardware operation robustness characteristic value of each time point of the set water quality monitoring station, analyzing the hardware operation robustness evolution characteristic value of the set water quality monitoring station; acquiring hardware appearance image data of the set water quality monitoring station, combining a pre-trained appearance damage recognition model, analyzing the hardware appearance damage aggregation characteristic value of the set water quality monitoring station, and combining the hardware operation robustness evolution characteristic value, analyzing the hardware inspection fault hidden danger characteristic value of the set water quality monitoring station; based on the hardware inspection fault hidden danger characteristic value, the hardware of the set water quality monitoring station is maintained.

[0006] Further, the hardware operation data includes an electro-hydraulic imbalance index, a magnetic flux deviation rate value, a vibration difference factor, a winding temperature difference factor, a water power cavitation risk value, and a pump cavity water power load value. The specific steps of analyzing the hardware operation robustness characteristic value of each time point of the set water quality monitoring station are as follows: based on the hardware operation data of each time point of the set water quality monitoring station, analyzing the hardware operation abnormal characteristic set of the corresponding time point, including mechanical and electrical load response characteristic value and pump cavity hydraulic bearing characteristic value; acquiring the communication transmission stability characteristic value of each time point of the set water quality monitoring station, and combining the hardware operation abnormal characteristic set to analyze the hardware operation robustness characteristic value of the corresponding time point.

[0007] Further, the specific steps of analyzing the hardware operation abnormal characteristic set of each time point of the set water quality monitoring station are as follows: based on the magnetic flux deviation rate value, the vibration difference factor, and the winding temperature difference factor of each time point of the set water quality monitoring station, analyzing the mechanical and electrical load response characteristic value of the corresponding time point; based on the electro-hydraulic imbalance index, the water power cavitation risk value, and the pump cavity water power load value of each time point of the set water quality monitoring station, analyzing the pump cavity hydraulic bearing characteristic value of the corresponding time point.

[0008] Further, the specific steps of acquiring the communication transmission stability characteristic value of each time point of the set water quality monitoring station are as follows: acquiring the hardware transmission network state data of each time point of the set water quality monitoring station, and performing standardization processing; based on the hardware transmission network state data of each time point of the set water quality monitoring station after standardization processing, analyzing the communication transmission stability characteristic value of the corresponding time point.

[0009] Further, the specific steps of analyzing the hardware operation robust evolution characteristic value of the set water quality monitoring station are as follows: based on the hardware operation robust characteristic value of each time point of the set water quality monitoring station, analyzing the operation robust characteristic difference value of several groups of adjacent time points of the set water quality monitoring station; based on the operation robust characteristic difference value of several groups of adjacent time points of the set water quality monitoring station, analyzing the hardware operation robust evolution characteristic value of the set water quality monitoring station.

[0010] Further, the hardware appearance image data is specifically pixel value and two-dimensional coordinates of each pixel point in the hardware appearance image, and the appearance damage identification model comprises an input layer, a damage extraction layer and a damage output layer.

[0011] Further, the specific steps of analyzing the hardware appearance damage aggregation characteristic value of the set water quality monitoring station are as follows: inputting the hardware appearance image data of the set water quality monitoring station into the pre-trained appearance damage identification model, analyzing the appearance damage feature set of the set water quality monitoring station, including cavitation pit texture characteristic value, sealing liquid infiltration erosion characteristic value and fouling deposition structure characteristic value; based on the appearance damage feature set of the set water quality monitoring station, analyzing the hardware appearance damage aggregation characteristic value of the set water quality monitoring station.

[0012] Further, the specific steps of analyzing the appearance damage feature set of the set water quality monitoring station are as follows: in the input layer of the appearance damage identification model, receiving the hardware appearance image data of the set water quality monitoring station and performing preprocessing; in the damage extraction layer of the appearance damage identification model, based on the preprocessed hardware appearance image data of the set water quality monitoring station, extracting the damage feature vector of the set water quality monitoring station; in the damage output layer of the appearance damage identification model, based on the damage feature vector of the set water quality monitoring station, outputting the appearance damage feature set of the set water quality monitoring station.

[0013] Further, the specific formula for calculating the hardware inspection fault hidden danger characteristic value of the set water quality monitoring station is as follows: ; wherein, , , The hardware inspection fault hidden danger characteristic value, the hardware appearance damage aggregation characteristic value and the hardware operation robust evolution characteristic value of the set water quality monitoring station are sequentially, , , The database.

[0014] Further, the specific steps of maintaining the hardware of the set water quality monitoring station based on the hardware inspection fault hidden danger characteristic value are as follows: comparing and analyzing the hardware inspection fault hidden danger characteristic value of the set water quality monitoring station with the preset hardware inspection fault hidden danger characteristic threshold interval; and taking the preset maintenance measure on the hardware of the set water quality monitoring station based on the comparison and analysis result.

[0015] The present application has the following advantages:

[0016] (1) The water quality monitoring station hardware inspection and monitoring method combined with the Internet of Things realizes unified evaluation of the running state and the appearance damage by acquiring the hardware operation data collection and the appearance image and fusing to generate the fault hidden danger characteristic value representing the overall state, and thus performs all-round dynamic evaluation on the water quality monitoring station hardware, so that the maintenance personnel can timely identify potential hidden dangers and accordingly take more accurate preventive or repair maintenance measures to ensure that the hardware health state is always within a stable and controllable range, thereby improving the accuracy of the inspection result and effectively ensuring the continuous and stable operation of the monitoring station under long-period operation conditions.

[0017] (2) The water quality monitoring station hardware inspection and monitoring method combined with the Internet of Things realizes automatic identification of the subtle damage area in the image by introducing a pre-trained appearance damage identification model and inputting the collected hardware appearance image data into the model for processing, and can uniformly represent different categories of damage to form a holistic feature, and the model performs pre-processing before feature extraction, which can effectively reduce the interference caused by external factors, making the identification result more stable and objective, and thus generating a hardware appearance damage aggregation characteristic value to realize efficient appearance damage identification and improve the inspection efficiency.

[0018] (3) The water quality monitoring station hardware inspection and monitoring method combined with the Internet of Things can continuously track and dynamically evaluate the running state of the water quality monitoring station hardware by introducing an analysis method of the hardware running robust evolution characteristic value, and the difference and change trend between adjacent time points depict the evolution trajectory of the hardware state, so that the inspection result has longitudinal time correlation, which can effectively identify subtle fluctuations and potential signs of deterioration in the running state, avoiding the masking of long-term risks due to the stability of short-time data, and the hardware running robust evolution characteristic value as a continuity index forms a comprehensive monitoring of the hardware health condition.

[0019] Of course, implementing any product of the present application does not necessarily need to achieve all the advantages described above at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1The application provides a hardware inspection and monitoring method for a water quality monitoring site combined with the Internet of Things.

[0021] Figure 2 The application provides a hardware inspection and monitoring method for a water quality monitoring site combined with the Internet of Things.

[0022] Figure 3 The application provides a hardware inspection and monitoring method for a water quality monitoring site combined with the Internet of Things.

[0023] Figure 4 The application provides a hardware inspection and monitoring method for a water quality monitoring site combined with the Internet of Things. DETAILED DESCRIPTION

[0024] Please refer to Figure 1 The application provides a hardware inspection and monitoring method for a water quality monitoring site combined with the Internet of Things, which comprises the following steps: continuously acquiring (i.e., acquiring from the first time point of the set period) hardware operation data of a plurality of time points of a set water quality monitoring site (in the embodiment, the hardware can be a sampling pump) in a set period (such as one day), and uploading based on the Internet of Things, which specifically comprises: at each time point, using a sensor arranged on the sampling pump to collect corresponding hardware operation data in real time, and transmitting the collected instantaneous operation data to an Internet of Things access gateway through wireless communication of the water quality monitoring site, and then uploading the data to a remote server through a cellular network, NB-IoT or LoRa Internet of Things communication link for storage and analysis.

[0025] Based on the hardware operation data of each time point of the set water quality monitoring site after uploading, the hardware operation robustness characteristic value of the corresponding time point is analyzed, and the corresponding time point of the set water quality monitoring site is monitored and warned, which specifically comprises: judging whether the hardware operation robustness characteristic value of each time point of the set water quality monitoring site is lower than a preset hardware operation robustness characteristic threshold value, if lower than the preset hardware operation robustness characteristic threshold value, issuing a warning to the corresponding personnel at the time point, otherwise, continuing to monitor.

[0026] The inspection processing is performed at the last time point of the set period, that is, based on the hardware operation robustness characteristic value of each time point of the set water quality monitoring site, the hardware operation robustness evolution characteristic value of the set water quality monitoring site is analyzed; the hardware appearance image data of the set water quality monitoring site is obtained (uploaded to the remote server based on the Internet of Things), the pre-trained appearance damage identification model is combined, the hardware appearance damage aggregation characteristic value of the set water quality monitoring site is analyzed, and the hardware operation robustness evolution characteristic value is combined, the hardware inspection fault hidden danger characteristic value of the set water quality monitoring site is analyzed; based on the hardware inspection fault hidden danger characteristic value, the hardware of the set water quality monitoring site is maintained.

[0027] The specific formula for calculating the hardware inspection fault hidden danger characteristic value of the set water quality monitoring site is as follows: ; wherein, is the hardware inspection fault hidden danger characteristic value of the set water quality monitoring site, is the hardware appearance damage aggregation characteristic value of the set water quality monitoring site, is the damage aggregation adjustment coefficient stored in the database, is the hardware operation robustness evolution characteristic value of the set water quality monitoring site, is the robust evolution adjustment coefficient stored in the database, is the cooperative adjustment coefficient stored in the database, , and in the present embodiment example, the damage aggregation adjustment coefficient , the robust evolution adjustment coefficient , and the cooperative adjustment coefficient stored in the database are 0.468, 0.532, and 1.246, respectively.

[0028] The specific steps of maintaining the hardware of the set water quality monitoring site based on the hardware inspection fault hidden danger characteristic value are as follows: comparing and analyzing the hardware inspection fault hidden danger characteristic value of the set water quality monitoring site with the preset hardware inspection fault hidden danger characteristic threshold interval; based on the comparison and analysis result, the set water quality monitoring site hardware takes the preset maintenance measures, which are specifically:

[0029] If the hardware inspection fault hidden feature value of the set water quality monitoring station is lower than the lower limit of the preset hardware inspection fault hidden feature threshold interval, it is determined that the hardware is normal, only the normal recording operation is performed, and the monitoring is continued, and no additional maintenance is required; if the hardware inspection fault hidden feature value of the set water quality monitoring station is within the preset hardware inspection fault hidden feature threshold interval, it is determined that the hardware has potential hidden dangers, preventive maintenance measures are performed, such as reducing the sampling pump operating frequency to reduce mechanical load and electro-hydraulic impact, and sending a detection notification to the maintenance personnel to prompt them to check the sampling pump during routine maintenance; if the hardware inspection fault hidden feature value of the set water quality monitoring station is higher than the upper limit of the preset hardware inspection fault hidden feature threshold interval, it is determined that the hardware has serious hidden dangers, and emergency maintenance measures are performed, including immediate shutdown, switching to a backup pump, and sending a high-priority alarm notification to the maintenance personnel.

[0030] Specifically, as shown in Figure 2 The hardware operation data includes an electro-hydraulic imbalance index, a magnetic flux deviation rate value, a vibration difference factor, a winding temperature difference factor, a water power cavitation risk value, and a pump cavity hydraulic load value. The specific steps of analyzing the hardware operation robustness feature value of each time point of the set water quality monitoring station are as follows:

[0031] Based on the hardware operation data of each time point of the set water quality monitoring station, the hardware operation abnormal feature set of the corresponding time point is analyzed, including the mechanical and electrical load response feature value and the pump cavity hydraulic bearing feature value.

[0032] The communication transmission stability feature value of each time point of the set water quality monitoring station is obtained, and the hardware operation robustness feature value of the corresponding time point is analyzed in combination with the hardware operation abnormal feature set.

[0033] The electro-hydraulic imbalance index is the matching relationship between the sampling pump electric input power and the hydraulic output power. It can be obtained by setting flow and pressure sensors in the sampling pump pipeline and arranging voltage and current sensors at the motor input end to obtain instantaneous flow value, pump cavity outlet pressure value, pump motor voltage value, and pump motor current value, and performing ratio processing, i.e. (instantaneous flow value x pump cavity outlet pressure value) / (pump motor voltage value x pump motor current value), to obtain the electro-hydraulic ratio, and performing ratio processing with the electro-hydraulic ratio reference value (which can be obtained by obtaining the historical electro-hydraulic ratio of several historical time points and taking the average value), and taking the result as the electro-hydraulic imbalance index. The lower the electro-hydraulic imbalance index, the more likely the pump is to be idle or clogged.

[0034] The magnetic flux deviation rate value is the deviation degree between the sampling pump motor winding magnetic flux and the reference magnetic flux (which can be obtained by obtaining the historical sampling pump motor winding magnetic flux at several historical time points and taking the average value), which can be obtained by arranging Hall sensors or magnetic field sensors near the motor winding to obtain the real-time magnetic flux value of the motor winding, and performing deviation processing with the reference magnetic flux, that is, |motor winding real-time magnetic flux value-reference magnetic flux| / reference magnetic flux, and taking the result as the magnetic flux deviation rate value. If the magnetic flux deviation rate value is large, it indicates that the motor may have rotor eccentricity, blockage or winding short circuit risk.

[0035] The vibration difference factor is the instantaneous vibration imbalance characteristic of the sampling pump during operation, which can be obtained by arranging three-axis acceleration sensors on the sampling pump body to obtain real-time instantaneous acceleration components in three directions, and calculating the instantaneous acceleration component difference (taking the absolute value) in any two directions, and extracting the maximum instantaneous acceleration component difference and the minimum instantaneous acceleration component difference, and performing ratio processing, that is, the maximum instantaneous acceleration component difference / the minimum instantaneous acceleration component difference, and taking the result as the vibration difference factor. If the vibration difference factor is large, it indicates that the instantaneous vibration of the pump body in a certain direction is much larger than that in other directions, indicating that the sampling pump may have abnormal working conditions such as rotor eccentricity, bearing wear or impeller jamming.

[0036] The winding temperature difference factor is the temperature difference between different parts of the sampling pump motor winding, which can be obtained by arranging temperature sensors at different positions of the sampling pump motor winding to obtain the temperature value at each position, and extracting the maximum temperature, the minimum temperature and the average temperature, and performing ratio processing, that is, (temperature maximum-temperature minimum) / temperature average, and taking the result as the winding temperature difference factor. If the winding temperature difference factor is large, it indicates that the heat dissipation of the motor winding is uneven, indicating that the sampling pump may have risks such as overload operation, local winding overheating or poor heat dissipation.

[0037] The hydraulic cavitation risk value is the risk of cavitation occurring in the sampling pump during operation, which can be obtained by arranging a micro-dissolved gas sensor in the sampling pump cavity to detect the dissolved gas concentration (such as dissolved oxygen) in the water sample in real time. At the current time point, the deviation processing is performed with the dissolved gas concentration at the last time point, that is, the absolute value of the difference between the dissolved gas concentration at the current time point and the dissolved gas concentration at the last time point / the interval time length value between the current time point and the last time point. The result is taken as the hydraulic cavitation risk value. If the hydraulic cavitation risk value increases, it indicates that the dissolved gas in the water sample is released in a large amount in a short time, and it is determined that cavitation may occur in the sampling pump cavity, thereby causing the sampling pump to have risks such as flow reduction, efficiency reduction or impeller erosion, and the hydraulic cavitation risk value at the first time point is 0.

[0038] The pump cavity hydraulic load value is the hydraulic bearing characteristic of the pump cavity of the sampling pump during operation, which is used to characterize the actual hydraulic load level of the sampling pump during water sampling. It can be obtained by arranging pressure sensors at the inlet and outlet of the sampling pump to obtain the inlet pressure value and the outlet pressure value, and performing ratio processing, i.e. inlet pressure value / outlet pressure value, and taking the result as the pump cavity hydraulic load value. When the value is low, it indicates that the pump inlet pressure is insufficient, and the sampling pump may be at risk of abnormal suction or idling.

[0039] The specific formula for calculating the hardware operation robustness characteristic value of the set water quality monitoring station at a certain time point is as follows: ; wherein, is the hardware operation robustness characteristic value of the set water quality monitoring station at a certain time point, is the electromechanical load response characteristic value of the set water quality monitoring station at a certain time point, is the load response adjustment coefficient stored in the database, is the pump cavity hydraulic bearing characteristic value of the set water quality monitoring station at a certain time point, is the hydraulic bearing adjustment coefficient stored in the database, is the communication transmission stability characteristic value of the set water quality monitoring station at a certain time point, is the operation adjustment coefficient stored in the database, is the smoothing adjustment coefficient stored in the database, and in the present embodiment, the load response adjustment coefficient , the hydraulic bearing adjustment coefficient , the operation adjustment coefficient , and the smoothing adjustment coefficient stored in the database are respectively 0.579, 0.421, 0.846, and 0.001.

[0040] The specific implementation example of calculating the hardware operation robustness characteristic value of the set water quality monitoring station at a certain time point is as follows. The existing data includes the electromechanical load response characteristic value, the pump cavity hydraulic bearing characteristic value, and the communication transmission stability characteristic value of the set water quality monitoring station at 5 time points (randomly selected), as shown in Table 1 and Figure 3

[0041] Table 1 Timing data example of set water quality monitoring station

[0042] Electromechanical load response characteristic value Pump chamber hydraulic bearing characteristic value Communication transmission stability characteristic value Time point 1 0.243 0.318 0.673 Time point 2 0.518 0.694 0.752 Time point 3 0.467 0.235 0.694 Time point 4 0.329 0.528 0.768 Time point 5 0.364 0.442 0.816

[0043] The load response adjustment coefficient stored in the database is 0.579.

[0044] The hydraulic bearing adjustment coefficient stored in the database is 0.421. ​​​

[0045] the running adjustment coefficient stored in the database is: 0.846;

[0046] the smoothing adjustment coefficient stored in the database is: 0.001;

[0047] Substituting the data in Table 1 and the above coefficients into the specific formula for setting the hardware running robust eigenvalue of a certain time point of the water quality monitoring station, we obtain:

[0048] the hardware running robust eigenvalue of the first time point of the set water quality monitoring station = 0.846 x ((0.579 / (1+0.243)) + (0.421 / (1+0.318)) ) x V0.673 = 0.545;

[0049] the hardware running robust eigenvalue of the second time point of the set water quality monitoring station = 0.846 x ((0.579 / (1+0.518)) + (0.421 / (1+0.694)) ) x V0.752 = 0.462;

[0050] the hardware running robust eigenvalue of the third time point of the set water quality monitoring station = 0.846 x ((0.579 / (1+0.467)) + (0.421 / (1+0.235)) ) x V0.694 = 0.518;

[0051] the hardware running robust eigenvalue of the fourth time point of the set water quality monitoring station = 0.846 x ((0.579 / (1+0.329)) + (0.421 / (1+0.528)) ) x V0.768 = 0.527;

[0052] the hardware running robust eigenvalue of the fifth time point of the set water quality monitoring station = 0.846 x ((0.579 / (1+0.364)) + (0.421 / (1+0.442)) ) x V0.816 = 0.547.

[0053] The specific steps of analyzing the hardware operation abnormal feature set of each time point of the set water quality monitoring station are as follows: based on the magnetic flux deviation rate value, the vibration difference factor and the winding temperature difference factor of each time point of the set water quality monitoring station, the corresponding time point of the electromechanical load response characteristic value is analyzed, which is specifically: the magnetic flux deviation rate value, the vibration difference factor and the winding temperature difference factor of each time point of the set water quality monitoring station are standardized, the weighted processing result is obtained based on the standardized processing result, the weighted processing result is processed by Sigmoid function, and the result is mapped between 0 and 1, so as to obtain the corresponding time point of the electromechanical load response characteristic value (used for representing the abnormal degree of the sampling pump motor and mechanical structure in the running process); based on the electro-hydraulic imbalance index, the water power cavitation risk value and the pump cavity hydraulic load value of each time point of the set water quality monitoring station, the pump cavity hydraulic bearing characteristic value of the corresponding time point is analyzed, which is specifically: the electro-hydraulic imbalance index, the water power cavitation risk value and the pump cavity hydraulic load value of each time point of the set water quality monitoring station are standardized, the weighted processing result is obtained based on the standardized processing result, the weighted processing result is processed by Sigmoid function, and the result is mapped between 0 and 1, so as to obtain the corresponding time point of the electromechanical load response characteristic value (used for representing the abnormal degree of the sampling pump in the running process), and in the weighting process, the electro-hydraulic imbalance index and the pump cavity hydraulic load value after standardization are taken as the opposite, such as 1 / (1+electro-hydraulic imbalance index after standardization).

[0054] The specific steps of obtaining the communication transmission stability characteristic value of each time point of the set water quality monitoring station are as follows: obtaining the hardware transmission network state data (including signal strength value, interference noise index, energy efficiency ratio value, carrier offset factor) of each time point of the set water quality monitoring station, and performing standardization processing (i.e. standardizing the signal strength value, the interference noise index, the energy efficiency ratio value and the carrier offset factor of each time point of the set water quality monitoring station); based on the hardware transmission network state data of each time point of the set water quality monitoring station after standardization, the communication transmission stability characteristic value of the corresponding time point is analyzed, which is specifically: the signal strength value, the interference noise index, the energy efficiency ratio value and the carrier offset factor of each time point of the set water quality monitoring station after standardization are weighted, the weighted processing result is processed by Sigmoid function, and the result is mapped between 0 and 1, so as to obtain the communication transmission stability characteristic value of the corresponding time point, and in the weighting process, the interference noise index and the carrier offset factor after standardization are taken as the opposite, such as 1 / (1+interference noise index).

[0055] The signal strength value is the instantaneous signal receiving strength of the sampling pump during data uploading in the water quality monitoring station. The signal strength value can be obtained by setting a radio frequency power detection sensor at the antenna interface of the sampling pump to obtain the data transmission signal power between the sampling pump and the upper computer in real time, and taking the data transmission signal power as the signal strength value. If the signal strength value is low, it indicates that the wireless link between the sampling pump and the upper computer is in a weak signal state, and there is a risk of transmission interruption.

[0056] The interference noise index is the degree of interference of the communication link when the sampling pump uploads data in the water quality monitoring station. The interference noise index can be obtained by arranging a radio frequency power sensor at the receiving end of the sampling pump antenna to obtain the power value of the received signal in real time and taking the power value as the received signal power. A noise power sensor is arranged at the noise branch of the radio frequency receiving circuit of the sampling pump to obtain the background noise power in real time. The signal-to-noise ratio is obtained by ratio processing. A bit error detection sensor is arranged at the demodulation circuit of the sampling pump to obtain the bit error rate at the current time in real time. The bit error rate / signal-to-noise ratio is obtained by ratio processing, and the result is taken as the interference noise index. If the interference noise index is large, it indicates that the sampling pump is affected by electromagnetic interference or background noise in the current running environment, resulting in unstable data uploading link.

[0057] The energy efficiency ratio is the instantaneous energy efficiency characteristic of the sampling pump when uploading data in the water quality monitoring station. The energy efficiency ratio can be obtained by arranging voltage and current sensors in the communication power supply circuit of the sampling pump to obtain the supply voltage value and supply current value in real time. A data rate detection sensor is arranged at the communication data interface of the sampling pump to measure the current transmission rate, i.e. bit rate, in real time. Then, the bit rate / (supply voltage value x supply current value) is obtained by ratio processing, and the result is taken as the energy efficiency ratio. If the energy efficiency ratio is low, it indicates that the data transmission rate of the sampling pump per unit energy consumption is insufficient, indicating that the communication link of the sampling pump is in an unstable state in the current running state, and there is a risk of link abnormality.

[0058] The carrier offset factor is the instantaneous frequency offset characteristic between the transmitting and receiving carriers of the sampling pump when uploading data in the water quality monitoring station. The carrier offset factor can be obtained by arranging a frequency detection sensor in the radio frequency receiving path of the sampling pump to obtain the actual frequency value of the received signal carrier in real time, and comparing the actual frequency value with the reference local oscillator frequency value set in the sampling pump, i.e. |actual frequency value-reference local oscillator frequency value| / reference local oscillator frequency value. The result is taken as the carrier offset factor. If the carrier offset factor is too large, it indicates that there is a frequency synchronization error in the current running process of the sampling pump, which may cause the demodulation of data packets to fail or the communication link to be interrupted, thereby reducing the data uploading stability of the sampling pump.

[0059] In the embodiment, by comprehensively collecting and analyzing the hardware operation data of the sampling pump, and introducing standardization and weighting processing, unified quantitative evaluation of the hardware operation state can be realized. Specifically, the hardware operation data collectively constitutes an operation abnormality feature set, so that the operation state is no longer limited to isolated judgment of single-point data, but the overall performance characteristics are reflected through multi-index fusion. At the same time, the communication transmission stability characteristic value is introduced, which associates the reliability of the data transmission link with the hardware operation state, so that the abnormality detection is more reliable. Secondly, through standardization and negation processing, the influence between different dimensions and scales can be weakened, ensuring that each index plays its due role under a unified scale. Finally, the obtained hardware operation robustness characteristic value not only reflects the current state in real time, but also provides a solid foundation for subsequent analysis of robust evolution trend, thereby effectively improving the accuracy of inspection and early warning, and further ensuring the continuous and stable operation of the monitoring site hardware.

[0060] Specifically, the specific steps of analyzing the hardware operation robust evolution characteristic value of the set water quality monitoring site are as follows: based on the hardware operation robust characteristic value of each time point of the set water quality monitoring site, the operation robustness feature difference value of several groups of adjacent time points of the set water quality monitoring site is analyzed, which is specifically: the hardware operation robustness characteristic values of any two adjacent time points of the set water quality monitoring site are subjected to difference processing to obtain the operation robustness feature difference value of each group of adjacent time points; based on the operation robustness feature difference value of several groups of adjacent time points of the set water quality monitoring site, the hardware operation robust evolution characteristic value of the set water quality monitoring site is analyzed, which is specifically: the operation robustness feature difference value of each group of adjacent time points is subjected to judgment processing with 0, if higher than 0, the operation robustness feature difference value of the group of adjacent time points is marked as a positive operation robustness feature difference value, otherwise, it is marked as a negative operation robustness feature difference value, and the sum of the positive operation robustness feature difference value and the sum of the negative operation robustness feature difference value are calculated, and the ratio is processed, i.e. the sum of the positive operation robustness feature difference value / (the sum of the positive operation robustness feature difference value+the sum of the negative operation robustness feature difference value), to obtain the hardware operation robust evolution characteristic value of the set water quality monitoring site.

[0061] In the embodiment, by difference processing of the adjacent time point operation robustness characteristic value, quantitative expression of the change trend of the hardware operation state with time is realized, which is no longer limited to static judgment of single-point time. Secondly, by distinguishing the difference results into positive and negative two categories and accumulating them respectively, the dual information of gradual improvement and gradual deterioration of the operation state can be captured at the same time, so as to avoid misjudgment caused by instantaneous fluctuation or accidental abnormality. Finally, the evolution characteristic value can clearly reflect the overall evolution direction and change speed of the hardware operation robustness, so that the inspection and monitoring have the ability of trend early warning and long-term evolution evaluation, thereby ensuring that the water quality monitoring site can stably play its role.

[0062] Specifically, as shown in Figure 4 The hardware appearance image data specifically refers to pixel values and two-dimensional coordinates of each pixel point in the hardware appearance image, and the appearance damage identification model includes an input layer, a damage extraction layer, and a damage output layer.

[0063] The specific steps of analyzing the hardware appearance damage aggregate characteristic value of the set water quality monitoring station are as follows: inputting the hardware appearance image data of the set water quality monitoring station into the pre-trained appearance damage identification model, analyzing the appearance damage feature set of the set water quality monitoring station, including cavitation micro-pit texture characteristic value, sealing liquid seepage erosion characteristic value, and scale deposition structure characteristic value; based on the appearance damage feature set of the set water quality monitoring station, analyzing the hardware appearance damage aggregate characteristic value of the set water quality monitoring station.

[0064] The specific formula for calculating the hardware appearance damage aggregate characteristic value of the set water quality monitoring station is as follows: ; wherein, is the hardware appearance damage aggregate characteristic value of the set water quality monitoring station, is the cavitation micro-pit texture characteristic value of the set water quality monitoring station, is the cavitation micro-pit adjustment coefficient stored in the database, is the sealing liquid seepage erosion characteristic value of the set water quality monitoring station, is the sealing liquid seepage adjustment coefficient stored in the database, is the scale deposition structure characteristic value of the set water quality monitoring station, is the scale deposition adjustment coefficient stored in the database, is the interaction adjustment coefficient stored in the database, and in the present embodiment example, the cavitation micro-pit adjustment coefficient stored in the database, the sealing liquid seepage adjustment coefficient , the scale deposition adjustment coefficient , and the interaction adjustment coefficient are 0.537, 0.421, 0.667, and 0.333, respectively.

[0065] The specific steps of analyzing the appearance damage feature set of the set water quality monitoring station are as follows: in the input layer of the appearance damage identification model, the hardware appearance image data of the set water quality monitoring station is received and preprocessed, which is specifically: the hardware appearance image is subjected to light balance processing, that is, the pixel gray value of the input image is subjected to histogram equalization processing to stretch the contrast of the image, thereby weakening the brightness unevenness caused by the difference in external light intensity, so that the fine damage area on the surface of the pump body is more prominent in vision; the image subjected to light balance is subjected to noise filtering processing, that is, the image pixel points are subjected to smoothing operation by using a median filter or a Gaussian filter to suppress random noise points introduced by the acquisition device or environmental interference, thereby avoiding being misidentified as effective damage texture in the subsequent damage feature extraction process; the filtered image is subjected to edge enhancement processing, that is, the image is subjected to convolution operation by using a Laplacian operator or a Sobel operator to highlight the gradient change of the surface structure edge and damage boundary of the pump body, so that the micro-cracks, pitting or fouling profile are more clearly visible; the image subjected to edge enhancement is subjected to background removal processing, which is specifically: based on color threshold segmentation or region growing method, the pump body region and the background region are separated, only the pixel points of the pump body region are reserved as the foreground region, and the pixel points of the background region are set to zero, thereby eliminating the interference of background stray information on damage identification.

[0066] In the damage extraction layer of the appearance damage identification model, the damage feature vector of the set water quality monitoring station is extracted based on the preprocessed hardware appearance image data of the set water quality monitoring station; in the damage output layer of the appearance damage identification model, the appearance damage feature set of the set water quality monitoring station is output based on the damage feature vector of the set water quality monitoring station, which is specifically: the cavitation pit texture feature, the sealing liquid infiltration erosion feature and the fouling deposition structure feature in the damage feature vector are subjected to Sigmoid function processing, and the results are mapped between 0 and 1 to obtain the cavitation pit texture feature value, the sealing liquid infiltration erosion feature value and the fouling deposition structure feature value.

[0067] The specific steps of extracting the damage feature vector of the set water quality monitoring station are as follows:

[0068] The pre-processed hardware appearance image is subjected to large-scale filtering processing, such as smoothing operation of the image by using a large-size Gaussian filter to obtain a large-scale filtered image, and the pre-processed hardware appearance image and the large-scale filtered image are subjected to pixel-by-pixel difference operation to obtain a difference enhancement graph (if the pixel point is in a flat area, the pixel value difference of the two images is small, and the difference value is close to zero; if there is a local pitting pit at the pixel point, the gray scale of the original image changes obviously at this position, while the large-scale filtered image is smoothed at this position, resulting in a significant difference between the two, thereby forming a high gray response in the difference enhancement graph, and thus the difference enhancement graph can highlight the local subtle concave area of the pump body surface), and the difference enhancement graph includes the difference pixel values of a plurality of pixel points and the corresponding two-dimensional coordinates.

[0069] The difference pixel value of each pixel point is subjected to gray scale processing to obtain the difference pixel gray scale value of the corresponding pixel point, and a local window is formed by selecting the neighborhood (such as a 3x3 neighborhood pixel window centered on the pixel point) around each pixel point as the center, and the LBP value of each pixel point is analyzed based on the local binary method, and the average value is taken to obtain the comprehensive LBP value, and the difference pixel gray scale value of each pixel point in each local window is subjected to variance processing to obtain the difference pixel gray scale variance value of each local window, and the average value is taken to obtain the comprehensive difference pixel gray scale variance value, and the comprehensive LBP value is subjected to weighted processing to extract the cavitation pit texture feature (used to represent the local pits and pitting morphology caused by the liquid cavitation effect of the appearance surface of the sampling pump during operation, reflecting the cavitation damage degree of the pump body surface).

[0070] The pixel value of each pixel point in the pre-processed hardware appearance image is subjected to HSV color space conversion to extract the corresponding saturation component and brightness component of the corresponding pixel point, and a local perception domain is formed by selecting the neighborhood (such as a 5x5 neighborhood pixel window centered on the pixel point) around each pixel point as the center, and the global saturation component mean value and the global brightness component mean value of the hardware appearance image and the saturation component mean value and the brightness component mean value of each local perception domain are counted, and difference value processing (taking absolute value) is performed respectively, and weighted fusion is performed based on the difference value processing result to obtain the local color offset value of each local perception domain, and the local color offset mean value and the local color offset variance value are counted respectively.

[0071] The pre-processed hardware appearance image is subjected to edge detection processing, such as calculating the gradient amplitude and non-maximum suppression of the image by using a Canny operator to extract a set of significant edge pixel points on the surface of the pump body, including a plurality of edge pixel points, and clustering according to connectivity (judging the connectivity of each edge pixel point and its adjacent 8 pixel points in the up, down, left, right and diagonal directions, if the Euclidean distance between the adjacent pixel points is less than or equal to 1, it is determined that they belong to the same connected domain, and by traversing all edge pixel points, the pixel points belonging to the same connected domain are classified into a group), thereby forming a plurality of continuous edge curves, and respectively performing morphological processing, i.e. performing an erosion operation to delete isolated or composed of a small number of pixel noise edge fragments by scanning the edge curve with a pre-set structure element (such as a 3x3 square convolution kernel); then performing a dilation operation to expand and fill the edge curve with the same structure element, connecting and completing the edge segments at the broken parts, thereby obtaining a plurality of closed or approximately closed edge contours, each edge contour (including a plurality of edge pixel points) corresponding to a local edge structure on the surface of the pump body;

[0072] Based on the two-dimensional coordinates of all edge pixel points in the edge contour, a least squares method is used to fit the edge point set as a circle; wherein, when fitting as a circle, the center coordinates and radius of the fitted circle are obtained, for any edge pixel point in the edge contour, the actual distance from the center of the circle is calculated (based on the Euclidean distance formula), and the difference value is processed with the radius of the fitted circle, based on the difference value processing result, the root mean square is processed, the residual mean square is obtained, and the ratio processing is performed with the radius of the fitted circle, i.e. 1-(residual mean square / radius of fitted circle), to obtain the roundness of each edge contour, and the roundness threshold is judged and processed, the edge contour higher than or equal to the preset roundness threshold is retained, and is marked as a liquid permeation edge contour;

[0073] For each liquid permeation edge contour, all edge pixel points are sorted in the order of connectivity, i.e. the two-dimensional coordinates of adjacent edge pixel points are recorded in sequence, the first pixel point after sorting is defined as the starting point of the contour, the last pixel point after sorting is defined as the ending point of the contour, and the Euclidean distance (i.e. gap distance) between the two is calculated, the perimeter of the edge contour is calculated, which can be obtained by point-by-point accumulation of the Euclidean distance between adjacent pixel points and ratio processing, 1-(gap distance / length), to obtain the closure of each liquid permeation edge contour, and the roundness of each liquid permeation edge contour is read for weighted average processing to obtain the liquid permeation complete feature, and the local color shift mean and local color shift variance are combined for weighted fusion processing to extract the sealing liquid permeation erosion feature (used to represent the ring-shaped imprint and color shift formed at the sealing part due to liquid permeation caused by sealing failure of the sampling pump during operation, reflecting the leakage erosion degree of the sealing part of the pump body);

[0074] The pixel value of each pixel point in the preprocessed hardware appearance image is subjected to gray scale processing to obtain the pixel gray value of the corresponding pixel point, and the average gray value is counted. The pixel gray value of each pixel point is subjected to difference processing (taking absolute value) with the average gray value, and the result is judged with the preset gray difference threshold. The pixel points higher than the preset gray difference threshold are marked as abnormal brightness pixel points, and then clustering is performed according to connectivity to form several candidate deposition regions. In each candidate deposition region, a 3*3 local neighborhood pixel set is selected with each abnormal brightness pixel point as the center, the variance of the pixel gray values of all abnormal brightness pixel points in the neighborhood is calculated, and the average value is taken to obtain the local texture roughness of the candidate deposition region. The local texture roughness is compared with the preset reference texture roughness (the average value of the gray variances calculated by the same method in the region without deposition on the pump body surface), and when the comparison result is higher than the preset proportion (such as 1.5), the candidate deposition region is marked as a deposition region,

[0075] Several deposition edge pixel points (which can be extracted by Canny edge detection) of each deposition region are extracted, the perimeter of the deposition region is calculated by point-by-point accumulation, and the number of pixel points of the deposition region is counted as its corresponding area, and the ratio is processed, that is, the perimeter 2 / (4*Pi*area), to obtain the geometric protrusion degree of each deposition region. The total number of pixel points in the preprocessed hardware appearance image is counted as the visible area of the pump body and the deposition area (i.e. the sum of the areas of the deposition regions), and the comparison is processed, that is, the deposition area / pump body visible area, to obtain the deposition coverage rate. The local texture roughness and the geometric protrusion degree of each deposition region are subjected to mean value processing, and the local texture roughness mean value and the geometric protrusion degree mean value are extracted, and then weighted with the deposition coverage rate to extract the fouling deposition structure feature (used to represent the irregular deposition region formed by the deposition of mineral matter or impurities on the appearance surface of the sampling pump during long-term operation, reflecting the fouling degradation degree of the pump body surface). The cavitation micro-pit texture feature, the sealing liquid infiltration erosion feature, and the fouling deposition structure feature are spliced into a damage feature vector.

[0076] The pre-training step of the appearance damage recognition model is as follows:

[0077] An annotated data set is obtained, which is composed of appearance image data of sampling pumps of a plurality of water quality monitoring stations. The appearance image data is collected by on-site manual inspection records and manually annotated by equipment maintenance experts. Each sample in the annotated data set includes a pump appearance image and a corresponding damage feature true value label. The true value label includes a cavitation pit texture feature label, a sealing liquid infiltration erosion feature label, and a scale deposition structure feature label, to ensure that the annotated data set has complete annotation information. The annotated data set is divided into a training set, a validation set, and a test set. For example, 80% of the data is used for training, 10% of the data is used for validation, and 10% of the data is used for testing. The proportion of samples of different damage types is balanced to improve the generalization ability of the model.

[0078] The appearance damage recognition model is trained. Specifically, in the training phase, each appearance image is input into the input layer of the model to extract the pixel points of the pump body region. In the damage extraction layer, the cavitation pit texture feature is extracted by using local binary pattern (LBP) analysis and gray variance statistics, the sealing liquid infiltration erosion feature is extracted by using color space offset and edge contour roundness and closeness fusion, and the scale deposition structure feature is extracted by using local roughness ratio, geometric protrusion degree, and coverage fusion. In the damage output layer, the Sigmoid activation function is used to map each feature value to the interval of 0-1 to output the corresponding damage probability distribution.

[0079] In the training process, the cross-entropy loss function is used as the optimization objective to minimize the difference between the predicted damage feature value and the manually annotated true value label. The Adam optimizer is used for iterative training to adjust the learning rate, batch size, and other hyperparameters to improve the training convergence speed and prediction accuracy. The performance of the model is evaluated by the validation set, and the network structure parameters (such as feature fusion layer weights and damage discrimination threshold) are dynamically adjusted to ensure that the model has good recognition ability under different damage types.

[0080] Finally, the test set is used to evaluate the trained appearance damage recognition model, and the precision (Precision), recall (Recall), and F1-score are calculated to verify the generalization ability and robustness of the model. When the test results meet the preset performance requirements, the trained model parameters are saved for subsequent calling in the actual monitoring process to realize online damage recognition and automatic warning of the appearance image of the sampling pump.

[0081] In the embodiment, by constructing and pre-training the appearance damage recognition model, the feature extraction of the hardware appearance image can be automatically completed, thereby improving the recognition accuracy. Through the introduction of the labeled data set and the hierarchical training mechanism, the model can learn multi-dimensional appearance feature expression, thereby maintaining high discrimination ability when facing different damage modes. In the damage extraction layer, the features related to the damage are extracted, thereby significantly improving the recognition accuracy and anti-interference ability. Through the loss function optimization and validation set tuning in the training stage, the model can continuously improve the sensitivity to the features. In the output layer, the Sigmoid function is used to normalize the results to the interval of 0-1, thereby ensuring the comparability and consistency of different features in quantization. Finally, the pre-trained model has strong stability and can quickly perform online recognition on the appearance image in the actual monitoring process, thereby realizing the early judgment of potential damage and improving the operation safety of the hardware of the water quality monitoring station.

[0082] Although preferred embodiments of the application have been described, those skilled in the art will be able to make additional changes and modifications without departing from the spirit and scope of the application. Therefore, the appended claims are intended to cover all such changes and modifications that fall within the scope of the application.

[0083] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A method for hardware inspection and monitoring of water quality monitoring station connected to the Internet of Things, characterized in that, The method comprises the following steps: obtaining hardware operation data of a set of water quality monitoring sites at a plurality of time points in succession and uploading the data based on the Internet of Things; based on the hardware operation data of the set of water quality monitoring sites at each time point after uploading, analyzing hardware operation robustness characteristic values of the set of water quality monitoring sites at the corresponding time points, and monitoring and warning the corresponding time points of the set of water quality monitoring sites; based on the hardware operation robustness characteristic values of the set of water quality monitoring sites at each time point, analyzing hardware operation robustness evolution characteristic values of the set of water quality monitoring sites; obtaining hardware appearance image data of the set of water quality monitoring sites, combining a pre-trained appearance damage recognition model, analyzing hardware appearance damage aggregation characteristic values of the set of water quality monitoring sites, and combining the hardware operation robustness evolution characteristic values, analyzing hardware inspection fault hidden danger characteristic values of the set of water quality monitoring sites; based on the hardware inspection fault hidden danger characteristic values, maintaining the hardware of the set of water quality monitoring sites.

2. The method for hardware inspection and monitoring of water quality monitoring station connected to internet as claimed in claim 1 wherein, The hardware operation data includes electro-hydraulic imbalance index, magnetic flux deviation rate value, vibration difference factor, winding temperature difference factor, water power cavitation risk value, and pump cavity water power load value. The specific steps of analyzing the hardware operation robustness characteristic values of the set of water quality monitoring sites at each time point are as follows: based on the hardware operation data of the set of water quality monitoring sites at each time point, analyzing hardware operation abnormal characteristic sets of the set of water quality monitoring sites at the corresponding time points, including electromechanical load response characteristic values and pump cavity hydraulic bearing characteristic values; obtaining communication transmission stability characteristic values of the set of water quality monitoring sites at each time point, and combining the hardware operation abnormal characteristic sets, analyzing hardware operation robustness characteristic values of the set of water quality monitoring sites at the corresponding time points.

3. The method for hardware inspection and monitoring of water quality monitoring station connected to IoT as claimed in claim 2 wherein, The specific steps of analyzing the hardware operation abnormal characteristic sets of the set of water quality monitoring sites at each time point are as follows: based on the magnetic flux deviation rate value, vibration difference factor, and winding temperature difference factor of the set of water quality monitoring sites at each time point, analyzing electromechanical load response characteristic values of the set of water quality monitoring sites at the corresponding time points; based on the electro-hydraulic imbalance index, water power cavitation risk value, and pump cavity water power load value of the set of water quality monitoring sites at each time point, analyzing pump cavity hydraulic bearing characteristic values of the set of water quality monitoring sites at the corresponding time points.

4. The method for hardware inspection and monitoring of water quality monitoring station connected to IoT as claimed in claim 2 wherein, The specific steps of obtaining the communication transmission stability characteristic values of the set of water quality monitoring sites at each time point are as follows: obtaining hardware transmission network state data of the set of water quality monitoring sites at each time point, and performing standardization processing; based on the hardware transmission network state data of the set of water quality monitoring sites at each time point after standardization processing, analyzing communication transmission stability characteristic values of the set of water quality monitoring sites at the corresponding time points.

5. The method for hardware inspection and monitoring of water quality monitoring station connected to internet as claimed in claim 1 wherein, The specific steps of analyzing the hardware operation robustness evolution characteristic values of the set of water quality monitoring sites are as follows: based on the hardware operation robustness characteristic values of the set of water quality monitoring sites at each time point, analyzing operation robustness characteristic difference values of a plurality of groups of adjacent time points of the set of water quality monitoring sites; based on the operation robustness characteristic difference values of the plurality of groups of adjacent time points of the set of water quality monitoring sites, analyzing hardware operation robustness evolution characteristic values of the set of water quality monitoring sites.

6. The method for hardware inspection and monitoring of water quality monitoring station connected to internet as claimed in claim 1 wherein, The hardware appearance image data specifically refers to pixel values and two-dimensional coordinates of each pixel point in the hardware appearance image. The appearance damage recognition model includes an input layer, a damage extraction layer, and a damage output layer.

7. The method for hardware inspection and monitoring of water quality monitoring station connected to IoT as claimed in claim 6 wherein, The specific steps of analyzing the hardware appearance damage aggregation feature value of the set water quality monitoring site are as follows: The hardware appearance image data of the set water quality monitoring site is input into the pre-trained appearance damage identification model, and the appearance damage feature set of the set water quality monitoring site is analyzed, including cavitation pit texture feature value, sealing liquid infiltration erosion feature value and scaling deposition structure feature value. Based on the appearance damage feature set of the set water quality monitoring site, the hardware appearance damage aggregation feature value of the set water quality monitoring site is analyzed.

8. The method for hardware inspection and monitoring of water quality monitoring station connected to IoT as claimed in claim 7 wherein, The specific steps of analyzing the appearance damage feature set of the set water quality monitoring site are as follows: In the input layer of the appearance damage identification model, the hardware appearance image data of the set water quality monitoring site is received and preprocessed; In the damage extraction layer of the appearance damage identification model, the damage feature vector of the set water quality monitoring site is extracted based on the preprocessed hardware appearance image data of the set water quality monitoring site; In the damage output layer of the appearance damage identification model, the appearance damage feature set of the set water quality monitoring site is output based on the damage feature vector of the set water quality monitoring site.

9. The method for hardware inspection and monitoring of water quality monitoring station connected to IoT as claimed in claim 1 wherein, The specific formula for calculating the hardware inspection fault hidden danger feature value of the set water quality monitoring site is as follows: ； wherein, , , are respectively a hardware inspection fault hidden danger characteristic value of the water quality monitoring station, a hardware appearance damage aggregation characteristic value, and a hardware operation robust evolution characteristic value of setting the water quality monitoring station, , , are respectively a damage aggregation adjustment coefficient, a robust evolution adjustment coefficient, and a synergistic adjustment coefficient stored in the database.

10. The method for hardware inspection and monitoring of water quality monitoring station connected to IoT as claimed in claim 1 wherein, The specific steps of maintaining the set water quality monitoring site hardware based on the hardware inspection fault hidden danger feature value are as follows: The hardware inspection fault hidden danger feature value of the set water quality monitoring site is compared and analyzed with the pre-set hardware inspection fault hidden danger feature threshold interval; Based on the comparison and analysis result, the set water quality monitoring site hardware takes the pre-set maintenance measures.

Images