Agricultural machinery operation fault early warning method and system

Abstract

The invention discloses an agricultural machinery operation fault early warning method and system, and relates to the field of agriculture. The problems of low prediction precision and low response efficiency caused by low data quality due to lack of comprehensive data fusion of multiple subsystems of agricultural machinery in an existing agricultural machinery operation fault early warning system and method are solved. According to the invention, multi-dimensional data from a plurality of subsystems such as a power system, a transmission and walking system, a hydraulic system, a working part, an electrical and control system and environment and working condition sensing can be integrated, and the data quality is improved; the high-precision fault probability of the equipment is calculated in real time through mathematical models such as data normalization, dynamic weight adjustment, fault propagation correction and nonlinear mapping, repair reporting is automatically triggered when the fault probability reaches advanced early warning, and the response speed is high, so that continuous monitoring of the health state of the equipment and early warning of fault risks are achieved; and the initiative and the accuracy of equipment maintenance are greatly improved. The method is mainly applied to agricultural machinery operation fault early warning.

Description

Technical Field

[0001] The present invention relates to the field of agriculture. Background Art

[0002] With the continuous improvement of the requirements for production efficiency and precision management in modern agriculture, agricultural machinery and equipment such as tractors, combine harvesters, and seeders operate for a long time under high load and multi-condition conditions. The risk of failures such as engine overheating, hydraulic failure, and transmission overload increases significantly, seriously affecting the operation continuity and economic benefits.

[0003] Traditional fault warning methods mainly rely on manual inspections or regular maintenance, making it difficult to achieve real-time and comprehensive monitoring of the equipment operation status; existing systems mostly based on single sensors or empirical rules lack the comprehensive integration and intelligent analysis of data from multiple subsystems such as power, transmission, hydraulics, working components, electricity, and environment; insufficient signal preprocessing and denoising means result in low-quality model inputs and low prediction accuracy; fixed thresholds and no adaptive optimization mechanism make it difficult to adapt to equipment aging and condition changes; prediction models are mostly shallow statistical or rule-based judgments, without using deep time series learning, and the prediction accuracy and lead time are limited; moreover, the warning levels are not detailed enough and lack linkage with remote operation and maintenance or automatic control, resulting in low response efficiency. These deficiencies make it difficult for existing fault warnings to meet the intelligent maintenance needs of large-scale and multi-equipment.

[0004] Therefore, there is an urgent need for an efficient agricultural machinery operation fault warning method and system that integrates multi-modal data fusion, adaptive threshold adjustment, and deep learning prediction. Summary of the Invention

[0005] The purpose of the present invention is to solve the problems existing in the existing agricultural machinery operation fault warning system and method, such as the lack of comprehensive data fusion of multiple subsystems of agricultural machinery, resulting in low data quality, low prediction accuracy, and low response efficiency. The present invention provides an agricultural machinery operation fault warning method and system.

[0006] An agricultural machinery operation fault warning method, the warning method includes:

[0007] S1. Initialization stage: Initialize the normal value N i of each sensor i of all subsystems of the agricultural machinery, the weight w i , and the lower limit L i and upper limit U i of the range, where i is the sensor serial number index;

[0008] S2. Data acquisition: Collect the original measurement value x i,j (t) of each sensor i of all subsystems of the agricultural machinery at each moment t during the jth sampling period; the initial value of j is 1; t = 1, 2... T, where T is the total number of sampling moments within the sampling period;

[0009] S3. Data preprocessing:

[0010] Based on L i and U i , obtain H i = 0.5(U i - L i ), where H i is the half range width of sensor i;

[0011] Based on N i and H i , calculate the normalized value x i,j ′(t) of the original measurement value of each sensor i in the j-th sampling period. After that, perform filter compensation on each normalized value x i,j ′(t) to obtain the filter compensation value of sensor i in the j-th sampling period

[0012] S4. According to the filter compensation values of each sensor at each moment in the j-th sampling period and the corresponding weight w i of the sensor, calculate the overall fault score S j (t) at each moment in the j-th sampling period; According to the filter compensation values of all sensors at each moment in the j-th sampling period, calculate the fault propagation correction term ΔS j (t) at each moment in the j-th sampling period; Use ΔS j (t) to correct S j (t) to obtain the corrected overall fault score S aj (t) at each moment in the j-th sampling period;

[0013] S5. Segmentally map S aj (t) to obtain the fault probability P j (t) at time t in the j-th sampling period;

[0014] S6. According to the value of P j (t), determine the warning level L j (t) at time t in the j-th sampling period.

[0015] If the warning level L j (t) is determined to be a high-level warning, then report for repair and end the warning; otherwise, execute step S7;

[0016] S7. Judge whether j is greater than 1. If the result is no, let j = j + 1 and execute step S2; if the result is yes, let j = j + 1 and calculate μ i,j and σ i,j based on the original measurement values collected in the (j - 1)-th sampling period, where μi,j is the mean value of the sampling values of sensor i in the j-th sampling period, and σ i,j is the standard deviation of the sampling values of sensor i in the j-th sampling period;

[0017] According to μ i,j and σ i,j the normal value N of each sensor i i , the weight w i , and the lower limit L of the range i and the upper limit U i are updated, and then step S2 is executed.

[0018] Preferably,

[0019]

[0020] Preferably, in step S3, the implementation method for calculating the normalized value x i,j ′(t) of the original measurement value of each sensor i is:

[0021]

[0022] Preferably, in step S3, the implementation method for obtaining the filtering compensation value of sensor i is:

[0023]

[0024] where θ is the abnormal threshold and c is the compensation factor.

[0025] Preferably, in step S4, n is the total number of sensors.

[0026] Preferably, in step S4, m i is the preset fault propagation coefficient of sensor i, and n is the total number of sensors.

[0027] Preferably, in step S4, S aj (t) = S j (t) + ΔS j (t).

[0028] Preferably, in step S7, the implementation method for calculating μ i,j and σ i,j includes:

[0029]

[0030] Preferably,

[0031] the implementation method for updating the normal value N of sensor i i is: Ni = μ i,j ;

[0032] The implementation method for updating the lower limit L of the measurement range of sensor i is: L i = μ i - 2σ i,j ; i,j ;

[0033] The implementation method for updating the upper limit U of the measurement range of sensor i is: U i = μ i + 2σ i,j ; i,j ;

[0034] The implementation method for updating the weight w of sensor i is: i S41. Define the indication function of sensor i at time t within the j-th sampling period

[0035] S42. Calculate the abnormal ratio of sensor i within the j-th sampling period according to I

[0036] S43. Adjust w j according to Δ

[0037] to obtain the adjusted weight w i,j ' of sensor i = w i (1 + 0.1Δ i ) i ; i,j ;

[0038] S44. Update w i according to the adjusted weights of all sensors to obtain the updated weight of sensor i Complete the weight update of sensor i, where n is the total number of sensors.

[0039] The agricultural machinery operation fault warning system includes a storage device, a processor, and a computer program stored in the storage device and executable on the processor. The processor executes the computer program to implement the agricultural machinery operation fault warning method as described.

[0040] Advantages of the present invention:

[0041] The present invention provides an agricultural machinery operation fault warning method and system for continuous real-time fault monitoring based on multi-sensor data normalization, weight adaptive adjustment, and intelligent warning of fault propagation, which can accurately evaluate the equipment status and predict the fault risk.

[0042] Explore intelligent diagnostic technologies based on multi-sensor data fusion, weighted fault modeling, and adaptive early warning decision-making. The fault early warning method and system of the present invention can integrate multi-dimensional data from multiple subsystems such as the power system, transmission and walking system, hydraulic system, working components, electrical and control system, and environmental and working condition perception, improve data quality, and through mathematical models such as data normalization, dynamic weight adjustment, fault propagation correction, and non-linear mapping, calculate the high-precision fault probability of the equipment in real time, and automatically trigger a repair request when the fault probability reaches the advanced warning level, thereby realizing continuous monitoring of the equipment health status and early warning of fault risks, greatly improving the initiative and accuracy of equipment maintenance.

[0043] The present invention has the ability of adaptive optimization. By analyzing historical data, it continuously adjusts the fault propagation coefficient and sensor weights to improve the sensitivity and prediction accuracy of anomaly detection. Through the present invention, real-time monitoring of the equipment operation status can be realized, unplanned downtime can be reduced, maintenance strategies can be optimized, and different equipment and operating environments can be adapted, improving the overall operation efficiency, and having broad application value. Brief Description of the Drawings

[0044] Figure 1 is a flowchart of a fault early warning method for agricultural machinery operations according to the present invention;

[0045] Figure 2 is a schematic diagram of the principle of the agricultural machinery subsystem. Detailed Embodiments

[0046] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0047] It should be noted that, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0048] Detailed Embodiment 1. In combination with Figure 1 and Figure 2 describe this embodiment. An agricultural machinery operation fault early warning method described in this embodiment, the early warning method includes:

[0049] S1. Initialization stage: Initialize the normal value N i of each sensor i of all subsystems of the agricultural machinery, i weight w i , and the lower limit L i and upper limit U, where i is the index of the sensor serial number; in specific applications, all subsystems of the agricultural machinery include the power system, the transmission and walking system, the hydraulic system, the working system, the electrical and control system, and the environmental monitoring system;

[0050] S2. Data acquisition: In the j-th sampling period, collect the original measurement values x i,j (t) of each sensor i of all subsystems of the agricultural machinery at each moment t; the initial value of j is 1; t = 1, 2... T, where T is the total number of sampling moments in the sampling period;

[0051] S3. Data preprocessing:

[0052] According to L i and U i , obtain H i = 0.5(U i - L i ), and H i is the half-range width of sensor i;

[0053] According to N i and H i , calculate the normalized value x i,j ′(t) of the original measurement value of each sensor i in the j-th sampling period, and then perform filter compensation on each normalized value x i,j ′(t) to obtain the filter compensation value

[0054] S4. According to the filter compensation values of each sensor at each moment in the j-th sampling period and the corresponding weight w i , calculate the overall fault score S j (t) at each moment in the j-th sampling period; according to the filter compensation values of all sensors at each moment in the j-th sampling period, calculate the fault propagation correction term ΔS j (t) at each moment in the j-th sampling period; use ΔS j (t) to correct S j (t) to obtain the corrected overall fault score S aj (t);

[0055] S5. Segmentally map S aj (t) to obtain the fault probability P j (t) at moment t in the j-th sampling period; specifically,

[0056]

[0057] S6. According to P j(t) value, determine the warning level L at time t in the jth sampling period j (t),

[0058]

[0059] If the warning level is L j (t) If it is determined to be a high-level warning, a repair is reported and the warning ends; otherwise, step S7 is executed;

[0060] S7. Determine whether j is greater than 1. If not, set j = j + 1 and execute step S2. If yes, set j = j + 1 and execute step S8.

[0061] S8. Calculate μ based on the original measurement value collected in the j-1th sampling period i,j and σ i,j , where μ i,j is the mean value of the sampling value of sensor i in the jth sampling period, σ i,j is the standard deviation of the sampling value of sensor i in the jth sampling period;

[0062] According to μ i,j and σ i,j The normal value N of each sensor i i , weight w i , and the lower limit of the range L i With upper limit U i After the update is completed, step S2 is executed.

[0063] This implementation provides a continuous real-time fault monitoring based on multi-sensor data normalization, weight adaptive adjustment, and fault propagation intelligent early warning, which can accurately assess equipment status and predict fault risks. Figure 2 As shown in Table 1, 31 types of key sensors are used to collect power, transmission, hydraulic, working parts, electrical and environmental monitoring data, and each data is linearly normalized to eliminate the influence of different measurement units on the calculation. In addition, data fusion is combined to improve data quality, thereby improving prediction accuracy.

[0064] Secondly, according to the real-time calibration mechanism, the normal value N of the sensor is dynamically adjusted i , weight w i , and the lower limit of the range L i With upper limit U i , and calculates a comprehensive failure score. Using piecewise linear functions and Sigmoid nonlinear mapping, the failure score is converted into a specific failure probability and compared with a set threshold. If the failure probability exceeds the safety threshold (i.e., advanced warning), a repair report is triggered, notifying operations and maintenance personnel to take timely maintenance measures.

[0065] The method of the present invention has a complete structure, determined parameters, continuous process and rapid response, significantly improves the accuracy of fault prediction and the timeliness of early warning, effectively reduces the risk of unexpected equipment shutdown and maintenance costs, and at the same time improves the operation stability and production efficiency of the equipment, and has strong adaptability and broad market application prospects.

[0066] Table 1 Sensor parameter table of agricultural machinery subsystem

[0067]

[0068]

[0069]

[0070] Specifically, the original data collected by each sensor i is normalized according to the preset normal value (denoted as N i ) and the lower limit L of the range i and the upper limit U i (the values can be as shown in Table 1). The present invention adopts linear normalization to map the deviation degree to the interval [0,1] to reflect the deviation degree; therefore, in step S3, the implementation method of calculating the normalized value x i,j ′(t) of the original measurement value of each sensor i is as follows:

[0071]

[0072] In this preferred embodiment, the normal value N i and the half-range width H of the sensor i i are used for linear normalization to map the deviation degree to the interval [0,1] to eliminate the influence of dimension, facilitate comprehensive analysis, intuitively reflect the degree of abnormality, enhance the anti-noise ability, facilitate model training and optimization, flexibly adapt to multi-condition changes, and when applied, when x i,j (t) = N i , a i (t) = 0 (normal state). When x i,j (t) deviates from N i by reaching or exceeding the half-range width H i = 0.5[U i -L i , a i (t) = 1 (full score for abnormality). a i (t) represents the normalized abnormality degree of the sensor i at a certain moment.

[0073] Specifically, in step S3, the implementation method of obtaining the filtering compensation value of the sensor i is as follows:

[0074]

[0075] Among them, θ is the anomaly threshold, and c is the compensation factor. During application, generally, the anomaly threshold θ = 0.9 and the compensation factor c = 0.1 are taken to perform filtering compensation on the normalized value.

[0076] In this preferred embodiment, for each normalized value x i,j ′(t), filtering compensation is performed to obtain the filtering compensation value of sensor i The purpose is to further suppress the residual high-frequency noise and short-time pulse interference after preprocessing such as linear normalization, so as to obtain higher robustness and stability in subsequent performance, significantly reduce the false alarm and miss rate, improve the warning sensitivity and lead time, and achieve real-time, accurate warning and dynamic adaptation of agricultural machinery equipment failures.

[0077] During specific application, in step S4, n is the total number of sensors. The overall fault score at each moment is calculated through the weights of all sensors. This calculation method organically integrates multi-dimensional sensor data of multiple subsystems such as power, transmission, hydraulic pressure, working components, electricity, and environment into a single fault score; with the reasonable allocation of the weights of each sensor, the fault signals of key components are highlighted, while general noise and short-time interference are weakened; the weighted accumulation operation is simple and efficient, suitable for real-time execution on edge devices; combined with the smoothed anomaly degree after filtering compensation, the robustness and stability of the fault score can be significantly improved, reducing false alarms and missed reports; at the same time, this method supports flexible addition and subtraction of sensors and dynamic adjustment of weights, and has good scalability and scene adaptability. During specific application, in step S4, m i is the preset fault propagation coefficient of sensor i, and n is the total number of sensors. The fault propagation correction term ΔS is calculated through the filtering compensation value j (t) and the fault propagation coefficient of each sensor i. This calculation method combines propagation correction and weighted summation, which not only enhances the sensitivity to key fault signals but also avoids over-response caused by local noise or occasional anomalies, further improving the accuracy and stability of early warning.

[0078] During specific application, in step S4, S aj (t) = S j (t) + ΔS j (t).

[0079] And in step S7, the implementation methods for calculating μ i,j and σ i,j include:

[0080]

[0081] The implementation method for updating the normal value N of sensor i is: N i is updated as: N i = μ i,j ;

[0082] The implementation method for updating the lower limit L of the measurement range of sensor i i is updated as: L i = μ i,j - 2σ i,j ;

[0083] The implementation method for updating the upper limit U of the measurement range of sensor i i is updated as: U i = μ i,j + 2σ i,j ;

[0084] The implementation method for updating the weight w of sensor i i is updated as follows:

[0085] S41. Define the indication function of sensor i at time t within the jth sampling period

[0086] S42. Calculate the abnormal proportion of sensor i within the jth sampling period according to I j (t)

[0087] S43. Adjust w i,j according to Δ i to obtain the adjusted weight w i ' of sensor i = w i (1 + 0.1Δ i,j );

[0088] S44. Update w i according to the adjusted weights of all sensors to obtain the updated weight of sensor i, completing the weight update of sensor i. n is the total number of sensors

[0089] In this preferred embodiment, a real-time calibration mechanism is provided to update the normal value N i , weight w i , lower limit L i of the measurement range, and upper limit U i of each sensor i in each sampling period, effectively capturing subtle abnormalities of the device under high load and multiple working conditions; dynamically adjusting the weights of each sensor according to the abnormal proportion within the statistical time window, and calculating the overall fault score in combination with the weights of each sensor to improve the accuracy of the acquired data, thereby improving the early warning accuracy

[0090] Specific Embodiment 2. The agricultural machinery operation fault warning system described in this embodiment includes a storage device, a processor, and a computer program stored in the storage device and executable on the processor. The processor executes the computer program to implement the agricultural machinery operation fault warning method as described above.

[0091] Principle Analysis:

[0092] The system collects power, transmission, hydraulic, working component, electrical, and environmental monitoring data through 31 types of key sensors, and performs linear normalization processing on each data to eliminate the influence of different measurement units on the calculation.

[0093] Secondly, according to the sensor fault propagation coefficient and the dynamic anomaly detection mechanism, the weights of the sensors are adjusted, and the comprehensive fault score is calculated in combination with the preset fault propagation model.

[0094] The system uses a piecewise linear function and a Sigmoid non-linear mapping method to convert the fault score into a specific fault probability, and compares it with the set threshold. If the fault probability exceeds the safety threshold, the system triggers a real-time warning to notify the operation and maintenance personnel to take maintenance measures in time.

[0095] Although the present invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the present invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed, as long as they do not depart from the spirit and scope of the present invention as defined by the appended claims. It should be understood that the features described in different dependent claims and in this document can be combined in a manner different from that described in the original claims. It should also be understood that the features described in connection with a single embodiment can be used in other described embodiments.

Claims

1. A method for warning of agricultural machinery operation failures, characterized in that, The early warning method includes: S1. Initialization stage: Initialize the normal value N of each sensor i of all subsystems of the agricultural machinery i , weight w i , and the lower limit L of the measuring range i and the upper limit U i , where i is the sensor serial number index; S2. Data acquisition: Collect the original measurement values \(x_{i}(t)\) of all sensors \(i\) of all subsystems of the agricultural machinery at each moment \(t\) within the \(j\)-th sampling period. The initial value of \(j\) is 1; \(t = 1, 2,\cdots,T\), where \(T\) is the total number of sampling moments within the sampling period. i,j (t); The initial value of \(j\) is 1; \(t = 1, 2,\cdots,T\), where \(T\) is the total number of sampling moments within the sampling period. S3. Data preprocessing: According to L i and U i , obtain H i = 0.5(U i - L i ), where H i is the half range width of sensor i; According to N i and H i , after calculating the normalized value x i,j ′(t) of the original measurement value of each sensor i in the j-th sampling period, filter compensation is performed on each normalized value x i,j ′(t) to obtain the filter compensation value of sensor i in the j-th sampling period S4. According to the filtering compensation values of each sensor at each moment within the j-th sampling period and the weight w corresponding to this sensor i , calculate the overall fault score S j (t) at each moment within the j-th sampling period; according to the filtering compensation values of all sensors at each moment within the j-th sampling period, calculate the fault propagation correction term ΔS j (t) at each moment within the j-th sampling period; use ΔS j (t) to correct S j (t), and obtain the corrected overall fault score S aj (t) at each moment within the j-th sampling period; S5, S aj (t) Segment mapping, obtain the fault probability P at time t in the jth sampling period j (t); S6. Determine the warning level L j j (t) at time t within the j-th sampling period according to the value of P j j (t), If the warning level is L j (t) If it is determined to be a high-level warning, a repair is reported and the warning ends; otherwise, step S7 is executed; S7, determine whether j is greater than 1, if not, set j = j + 1, and execute step S2; if yes, set j = j + 1, and calculate μ based on the original measurement value collected in the j-1th sampling period i,j and σ i,j , where μ i,j is the mean value of the sampling value of sensor i in the jth sampling period, σ i,j is the standard deviation of the sampling value of sensor i in the jth sampling period; According to μ i,j and σ i,j After updating the normal value N i of each sensor i, the weight w i , and the lower limit L i of the range and the upper limit U i perform step S2.

2. A fault early warning method for agricultural machinery operation according to claim 1, characterized in that 3. The agricultural machinery operation fault warning method according to claim 1, characterized in that In step S3, the implementation method for calculating the normalized value x i,j ′(t) of the original measurement value of each sensor i is as follows:

4. A method for warning of agricultural machinery operation failures according to claim 1, characterized in that In step S3, the filtering compensation value of sensor i is obtained The implementation method is as follows: where θ is the abnormal threshold and c is the compensation factor.

5. A method for warning of agricultural machinery operation failures according to claim 1, characterized in that In step S4, n is the total number of sensors.

6. The agricultural machinery operation fault warning method according to claim 1, wherein, In step S4, m i is the fault propagation coefficient of the preset sensor i, and n is the total number of sensors.

7. A method for early warning of agricultural machinery operation failures according to claim 1, characterized in that, In step S4, S aj (t) = S j (t) + ΔS j (t).

8. A method for warning of agricultural machinery operation failures according to claim 1, characterized in that, In step S7, calculate μ i,j and σ i,j The implementation methods include:

9. A fault early warning method for agricultural machinery operation according to claim 8, characterized in that The normal value N of sensor i i is updated in the following way: N i = μ i,j ; Lower limit L of the measurement range of sensor i i The implementation method for updating is: L i = μ i,j - 2σ i,j ; The upper limit U of the range of sensor i i is updated as follows: U i = μ i,j + 2σ i,j ; The weight w of sensor i i is updated in the following way: S41. Define the indicator function of sensor i at time t in the jth sampling period S42. Calculate the abnormal ratio of sensor i within the j-th sampling period according to I j (t) S43. According to Δ i,j Adjust w i to obtain the adjusted weight w i ' = w i (1 + 0.1Δ i,j ); S44. Update \(w\) according to the adjusted weights of all sensors i to obtain the updated weight of sensor \(i\) Complete the weight update of sensor \(i\), where \(n\) is the total number of sensors.

10. An agricultural machinery operation fault warning system, comprising a storage device, a processor, and a computer program stored in the storage device and executable on the processor, characterized in that, The processor executes a computer program to implement a fault early warning method for agricultural machinery operation as described in claims 1 to 9.

Images