Intelligent diagnosis and predictive maintenance method for heavy-load stereoscopic warehouse equipment
By using multi-source sensor data fusion and Bayesian theorem to identify faults in heavy-duty automated warehouse equipment, the problem of incomplete equipment status reflection in traditional methods is solved. This enables accurate fault prediction and maintenance optimization, reduces unplanned downtime and maintenance costs, and extends equipment lifespan.
Patent Information
- Application Number
- CN202610022949.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-08
- Publication Date
- 2026-05-12
AI Technical Summary
Traditional heavy-duty automated warehouse equipment health monitoring relies on a single sensor, which cannot comprehensively and accurately reflect the equipment status. Early fault signals are easily masked, leading to delayed warnings or false alarms. It is difficult to locate the root cause of the fault and cannot achieve predictive maintenance, resulting in frequent unplanned downtime and high maintenance costs.
The system uses multi-source sensors (vibration, stress, temperature, and acoustics) to collect data in real time, performs multi-dimensional feature extraction and adaptive data fusion, and combines Bayes' theorem to identify fault modes and assess equipment health, predict remaining service life, and generate maintenance priority decisions.
It improves detection accuracy and reliability, significantly enhances fault early warning capabilities, reduces maintenance costs, and extends equipment life.
Smart Images
Figure CN122020365A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of industrial Internet of Things and intelligent diagnostics, and in particular to a method for intelligent diagnostics and predictive maintenance of heavy-duty automated warehouse equipment. Background Technology
[0002] Traditional heavy-duty automated warehouses (such as heavy steel pipe warehouses) primarily rely on single-type sensors (such as vibration sensors) or simple threshold alarm mechanisms for equipment health monitoring. Under harsh conditions such as heavy loads, strong impacts, and high dust levels (where a single steel pipe can weigh several tons), traditional methods face significant bottlenecks: single-sensor signals cannot comprehensively and accurately reflect the true state of the equipment under combined loads; early fault signals are weak and easily masked by environmental noise and operating condition fluctuations, leading to delayed warnings or false alarms; simultaneously, due to the lack of fusion analysis of multi-source information and in-depth diagnostics based on equipment health status, existing systems struggle to pinpoint the root cause of faults and cannot achieve predictive maintenance, resulting in frequent unplanned downtime, high maintenance costs, and significant safety hazards. Summary of the Invention
[0003] The purpose of this invention is to overcome the shortcomings of the existing technology. To achieve the above objective, an intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment is adopted to solve the problems mentioned in the background technology.
[0004] A method for intelligent diagnosis and predictive maintenance of heavy-duty automated storage and retrieval systems includes the following steps: S1. Real-time synchronous acquisition of multi-source sensor data of key equipment in the automated warehouse, including ASRS stacker cranes and rack structures, and the multi-source sensor data includes at least vibration data, stress data, temperature data, and acoustic data. S2. Perform multi-dimensional feature extraction on the multi-source sensor data to construct a multi-dimensional feature vector containing vibration feature vector, stress feature vector, and temperature feature vector; S3. Based on the reliability coefficients and real-time signal quality scores of each sensor, adaptive data fusion is performed on the multidimensional feature vector using a weighting formula. S4. Based on the fused feature data, calculate the anomaly score using the anomaly detection scoring function and compare it with the threshold to determine the device status. S5. When the abnormal score exceeds the threshold, the intelligent diagnosis process is initiated, which includes: calculating the failure mode probability based on Bayes' theorem to identify the failure type, assessing the equipment health score by combining the failure probability and severity, predicting the remaining service life of the equipment based on the health degradation, and finally generating decision recommendations that include maintenance priorities.
[0005] As a further aspect of the present invention: in step S1, the vibration data is collected by a vibration sensor deployed in the stacker crane drive system, the stress data is collected by a stress sensor deployed in the key load-bearing components of the rack, the temperature data is collected by a temperature sensor deployed in the motor and bearing, and the acoustic data is collected by an acoustic sensor deployed near the equipment.
[0006] As a further aspect of the present invention, step S2 specifically includes the following steps: The multidimensional feature vector is: ; The feature vectors for each dimension are as follows: The vibration characteristic vector This includes the root mean square value, peak value, kurtosis, peak factor, and dominant frequency of the vibration signal;
[0007] The stress feature vector This includes maximum stress, average stress, stress variance, and stress variation trend;
[0008] The temperature feature vector This includes the current temperature, temperature gradient, temperature change trend, and maximum temperature difference; .
[0009] As a further aspect of the present invention, step S3 specifically includes the following steps: The weight allocation formula is as follows:
[0010] in, For the first The fusion weights of the individual sensors; The sensor reliability coefficient; Assess the quality of sensor signals; The reliability coefficient of the sensor The sensor is pre-calibrated based on historical stability data under heavy load, dust, and temperature change conditions; the sensor signal quality score is... Dynamic calculations are performed based on the signal-to-noise ratio, completeness, and validity of the real-time collected data.
[0011] As a further aspect of the present invention, step S4 specifically includes the following steps: The anomaly detection scoring function is:
[0012] in, For the first The detection sensitivity of each sensor, For the first 3D eigenvalues and These are the mean and standard deviation of the feature under normal conditions, respectively.
[0013] As a further aspect of the present invention: in step S5, the calculation of the failure mode probability based on Bayes' theorem specifically involves: Calculation in obtaining multidimensional feature vectors Under the given conditions, the posterior probability of equipment failure j is: .
[0014] As a further aspect of the present invention, step S5 specifically includes the following steps: The device health score Hscore is calculated using the following formula:
[0015] in, For pre-set, fault-reflecting The severity coefficient of the impact.
[0016] As a further aspect of the present invention, step S5 specifically includes the following steps: The remaining useful life (RUL) is predicted using an index model based on baseline lifespan and the degree of health degradation. The calculation formula is as follows:
[0017] in, This is the baseline lifespan of the equipment under normal operating conditions. The attenuation coefficient is related to the equipment type. Historical degradation of equipment health score; The maintenance priority Based on comprehensive equipment health score Remaining service life And the estimated impact of downtime costs due to failures. The formula for multi-factor weighted calculation is as follows:
[0018] in, Impact of downtime costs due to malfunctions; , , These are the weighting coefficients for each factor.
[0019] As a further aspect of the present invention, it also includes specific steps: For the rated heavy-load operating conditions of the key equipment in the automated warehouse, multi-source sensor data under normal operating conditions are collected to statistically obtain the mean values of each feature dimension required in the anomaly detection scoring function under normal conditions. and standard deviation .
[0020] As a further aspect of the present invention, it also includes generating fault diagnosis results, health assessments, and remaining service life. It provides visual output of maintenance decision suggestions and automatically triggers maintenance work orders when the maintenance priority exceeds a preset threshold.
[0021] Compared with the prior art, the present invention has the following technical advantages: The above technical solution utilizes multiple sensors, including vibration, stress, temperature, and acoustic sensors, to simultaneously collect operational data from key equipment in heavy-duty automated warehouses and extract multi-dimensional features. An adaptive weight allocation algorithm based on reliability and signal quality is employed to fuse these multi-source features. The fused features and anomaly scoring function are used for status determination. When an anomaly is detected, Bayes' theorem is used for fault mode identification, comprehensively assessing equipment health and predicting remaining lifespan, ultimately generating quantified maintenance priority decisions. Key features include: adaptability to heavy-duty conditions (specifically designed for the heavy-duty characteristics of steel pipe warehouses); multi-sensor fusion (improving detection accuracy and reliability); early fault identification (significantly enhancing fault warning capabilities); and intelligent maintenance decision-making (reducing maintenance costs and extending equipment lifespan). Attached Figure Description
[0022] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings: Figure 1 This is a schematic diagram illustrating the steps of a maintenance method according to an embodiment of this application. Detailed Implementation
[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0024] Please refer to Figure 1 In this embodiment of the invention, a method for intelligent diagnosis and predictive maintenance of heavy-duty automated warehouse equipment includes the following steps: S1. Real-time synchronous acquisition of multi-source sensor data of key equipment in the automated warehouse, including ASRS stacker cranes and rack structures, and the multi-source sensor data includes at least vibration data, stress data, temperature data, and acoustic data. In this embodiment, the vibration data is collected by vibration sensors deployed in the stacker crane drive system, the stress data is collected by stress sensors deployed in key load-bearing components of the rack, the temperature data is collected by temperature sensors deployed in the motor and bearings, and the acoustic data is collected by acoustic sensors deployed near the equipment.
[0025] Specifically, to achieve comprehensive status awareness of key equipment in the heavy-duty steel pipe automated warehouse, this embodiment deploys the following multi-sensor system, the specific configuration and functions of which are as follows: 1. Vibration Sensors: High-frequency shock resistant industrial accelerometers are employed and deployed in key transmission components of the ASRS stacker crane, including the output shaft bearing housing of the lifting motor, the bearing housing of the traveling wheels, and the gearbox housing. They primarily monitor vibration acceleration signals to extract features such as root mean square (RMS), peak value, kurtosis, peak factor, and dominant frequency, in order to identify mechanical faults such as bearing wear, broken gear teeth, imbalance, and misalignment.
[0026] 2. Stress Sensors: High-precision resistance strain gauges are attached to key load-bearing components of the automated warehouse racking system, particularly stress concentration points on uprights and beams in high-rise and heavy-load areas. They continuously monitor the dynamic stress and strain of the racking system during steel pipe storage and retrieval, obtaining maximum stress, average stress, stress variance, and trends to assess the structural integrity and overload risk of the racking system.
[0027] 3. Temperature Sensors: A combination of platinum resistance thermometers and infrared temperature sensors is used. The platinum resistance thermometer is embedded in the windings of the stacker crane's lifting motor and traveling motor, as well as the drive-end bearings, for precise contact temperature measurement. The infrared temperature sensor, on the other hand, monitors the temperature of the high-speed rotating bearing's outer surface non-contactly. Monitored parameters include real-time temperature, temperature rise gradient, trend, and temperature difference between different measuring points, used to provide early warning of overheating, poor lubrication, and other faults.
[0028] 4. Acoustic Sensors: Deploy broadband industrial microphones in the stacker crane's operating aisles and racking areas. They collect broadband sound signals during operation, and are particularly useful for capturing sudden abnormal noises (such as metal impacts and sharp friction sounds) and continuous abnormal noises, serving as an effective supplement to vibration monitoring, and are especially sensitive to faults such as loosening and scratching.
[0029] 5. Displacement Sensors: Laser rangefinders or high-precision encoders are used. Laser sensors are installed at both ends of the rack aisle to detect the stacker crane's positioning accuracy in the horizontal and lifting directions; encoders are integrated into the stacker crane's traveling and lifting motors for real-time feedback of position and speed. The data is used to monitor positioning deviations, operational vibrations, and other performance degradation.
[0030] 6. Current Sensor: A Hall effect current sensor is used, clamped onto the main power supply circuit of each drive motor of the stacker crane. By monitoring the amplitude, waveform, and harmonic components of the motor's operating current, and analyzing its load changes, it can be used to diagnose mechanical jamming, voltage fluctuations, or electrical faults in the motor itself.
[0031] Data Acquisition and Synchronization: All sensors are connected to the central processing unit via industrial fieldbus or IoT gateway. Synchronous acquisition is triggered by a unified timing controller to ensure time alignment of multi-source data, laying the foundation for subsequent fusion analysis. The acquisition frequency is set according to the fault characteristic frequency band; vibration and acoustic data are acquired at kHz level at high speed, while stress, temperature, displacement, and current data are acquired at Hz level.
[0032] S2. Perform multi-dimensional feature extraction on the multi-source sensor data to construct a multi-dimensional feature vector containing vibration feature vector, stress feature vector, and temperature feature vector; In this embodiment, the specific steps include: The multidimensional feature vector is: ; The feature vectors for each dimension are as follows: The vibration characteristic vector This includes the root mean square value, peak value, kurtosis, peak factor, and dominant frequency of the vibration signal;
[0033] The stress feature vector This includes maximum stress, average stress, stress variance, and stress variation trend;
[0034] The temperature feature vector This includes the current temperature, temperature gradient, temperature change trend, and maximum temperature difference; .
[0035] S3. Based on the reliability coefficients and real-time signal quality scores of each sensor, adaptive data fusion is performed on the multi-dimensional feature vector using a weighting formula. The specific steps include: The weight allocation formula is as follows:
[0036] in, For the first The fusion weights of the individual sensors; The sensor reliability coefficient; Assess the quality of sensor signals; The reliability coefficient of the sensor The sensor is pre-calibrated based on historical stability data under heavy load, dust, and temperature change conditions; the sensor signal quality score is... Dynamic calculations are performed based on the signal-to-noise ratio, completeness, and validity of the real-time collected data. S4. Based on the fused feature data, an anomaly score is calculated using an anomaly detection scoring function and compared with a threshold to determine the device status. The specific steps include: The anomaly detection scoring function is:
[0037] in, For the first The detection sensitivity of each sensor, For the first 3D eigenvalues and These are the mean and standard deviation of the feature under normal conditions, respectively. S5. When the abnormal score exceeds the threshold, the intelligent diagnosis process is initiated, which includes: calculating the failure mode probability based on Bayes' theorem to identify the failure type, assessing the equipment health score by combining the failure probability and severity, predicting the remaining service life of the equipment based on the health degradation, and finally generating decision recommendations that include maintenance priorities.
[0038] In this embodiment, the failure mode probability calculation based on Bayes' theorem specifically involves: Calculation in obtaining multidimensional feature vectors Under the given conditions, the posterior probability of equipment failure j is: .
[0039] In this embodiment, step S5 specifically includes the following steps: The device health score Hscore is calculated using the following formula:
[0040] in, For pre-set, fault-reflecting The severity coefficient of the impact.
[0041] In this embodiment, step S5 specifically includes the following steps: The remaining useful life (RUL) is predicted using an index model based on baseline lifespan and the degree of health degradation. The calculation formula is as follows:
[0042] in, This is the baseline lifespan of the equipment under normal operating conditions. The attenuation coefficient is related to the equipment type. Historical degradation of equipment health score; The maintenance priority Based on comprehensive equipment health score Remaining service life And the estimated impact of downtime costs due to failures. The formula for multi-factor weighted calculation is as follows:
[0043] in, Impact of downtime costs due to malfunctions; , , These are the weighting coefficients for each factor.
[0044] This embodiment also includes specific steps: For the rated heavy-load operating conditions of the key equipment in the automated warehouse, multi-source sensor data under normal operating conditions are collected to statistically obtain the mean values of each feature dimension required in the anomaly detection scoring function under normal conditions. and standard deviation .
[0045] In this embodiment, the generated fault diagnosis results, health assessment, and remaining service life are also included. It provides visual output of maintenance decision suggestions and automatically triggers maintenance work orders when the maintenance priority exceeds a preset threshold.
[0046] The beneficial effects of this invention are: By implementing the technical solution of the present invention, the following technical effects are achieved: 1. Improved detection accuracy: Anomaly detection accuracy reaches over 95%, and false alarm rate is reduced by 60%; 2. Early detection capability: The average time for early fault detection is increased by 3-5 times; 3. Reduced maintenance costs: Predictive maintenance reduces unplanned downtime by 40%; 4. Extended equipment lifespan: Intelligent diagnostics can extend the equipment's lifespan by more than 20%.
[0047] Quantitative performance metrics: Anomaly detection response time: ; Fault location accuracy: ; Prediction accuracy: ; System availability: .
[0048] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention. The scope of the invention is defined by the appended claims and their equivalents, all of which should be included within the scope of protection of the invention.
Claims
1. A method for intelligent diagnosis and predictive maintenance of heavy-duty automated storage and retrieval systems, characterized in that, Includes the following steps: S1. Real-time synchronous acquisition of multi-source sensor data of key equipment in the automated warehouse, including ASRS stacker cranes and rack structures, and the multi-source sensor data includes at least vibration data, stress data, temperature data, and acoustic data. S2. Perform multi-dimensional feature extraction on the multi-source sensor data to construct a multi-dimensional feature vector containing vibration feature vector, stress feature vector, and temperature feature vector; S3. Based on the reliability coefficients and real-time signal quality scores of each sensor, adaptive data fusion is performed on the multidimensional feature vector using a weighting formula. S4. Based on the fused feature data, calculate the anomaly score using the anomaly detection scoring function and compare it with the threshold to determine the device status. S5. When the abnormal score exceeds the threshold, the intelligent diagnosis process is initiated, which includes: calculating the failure mode probability based on Bayes' theorem to identify the failure type, assessing the equipment health score by combining the failure probability and severity, predicting the remaining service life of the equipment based on the health degradation, and finally generating decision recommendations that include maintenance priorities.
2. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, In step S1, the vibration data is collected by vibration sensors deployed in the stacker crane drive system, the stress data is collected by stress sensors deployed in key load-bearing components of the rack, the temperature data is collected by temperature sensors deployed in the motor and bearings, and the acoustic data is collected by acoustic sensors deployed near the equipment.
3. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, The specific steps in step S2 include: The multidimensional feature vector is: ； The feature vectors for each dimension are as follows: The vibration characteristic vector This includes the root mean square value, peak value, kurtosis, peak factor, and dominant frequency of the vibration signal; The stress feature vector This includes maximum stress, average stress, stress variance, and stress variation trend; The temperature feature vector This includes the current temperature, temperature gradient, temperature change trend, and maximum temperature difference; 。 4. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, The specific steps in step S3 include: The weight allocation formula is as follows: in, For the first The fusion weights of the individual sensors; The sensor reliability coefficient; Assess the quality of sensor signals; The reliability coefficient of the sensor The sensor is pre-calibrated based on historical stability data under heavy load, dust, and temperature change conditions; the sensor signal quality score is... Dynamic calculations are performed based on the signal-to-noise ratio, completeness, and validity of the real-time collected data.
5. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, The specific steps in step S4 include: The anomaly detection scoring function is: in, For the first The detection sensitivity of each sensor, For the first 3D eigenvalues and These are the mean and standard deviation of the feature under normal conditions, respectively.
6. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, In step S5, the calculation of the failure mode probability based on Bayes' theorem specifically involves: Calculation in obtaining multidimensional feature vectors Under the given conditions, the posterior probability of equipment failure j is: 。 7. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 6, characterized in that, The specific steps in step S5 include: The device health score Hscore is calculated using the following formula: in, For pre-set, fault-reflecting The severity coefficient of the impact.
8. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, The specific steps in step S5 include: The remaining useful life (RUL) is predicted using an index model based on baseline lifespan and the degree of health degradation. The calculation formula is as follows: in, This is the baseline lifespan of the equipment under normal operating conditions. The attenuation coefficient is related to the equipment type. Historical degradation as a measure of equipment health score; The maintenance priority Based on comprehensive equipment health score Remaining service life And the estimated impact of downtime costs due to failures. The formula for multi-factor weighted calculation is as follows: in, Impact of downtime costs due to malfunctions; , , These are the weighting coefficients for each factor.
9. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, It also includes specific steps: For the rated heavy-load operating conditions of the key equipment in the automated warehouse, multi-source sensor data under normal operating conditions are collected to statistically obtain the mean values of each feature dimension required in the anomaly detection scoring function under normal conditions. and standard deviation .
10. The intelligent diagnosis and predictive maintenance method for heavy-duty automated warehouse equipment according to claim 1, characterized in that, This also includes the generated fault diagnosis results, health assessment, and remaining service life. It provides visual output of maintenance decision suggestions and automatically triggers maintenance work orders when the maintenance priority exceeds a preset threshold.