Fault diagnosis method, device and equipment and readable storage medium

Abstract

The invention discloses a fault diagnosis method, device and equipment and a readable storage medium, and relates to the technical field of machining, and the method comprises the steps: collecting multi-dimensional operation data in the operation process of a suction nozzle attachment type pneumatic module in real time; performing classification processing on the multi-dimensional operation data according to a suction nozzle working period to obtain target operation data; and inputting the target operation data into a diagnosis and prediction model, identifying a pneumatic module fault type and a suction nozzle consumable service life state through the diagnosis and prediction model, and outputting a fault diagnosis result and consumable service life prediction information. The method aims to solve the problems that in the prior art, the running state of a pneumatic module and the service life of suction nozzle consumables are lack of systematic pre-judgment in advance, a single parameter or artificial experience is relied on, so that faults are passively checked, and consumable replacement is unreasonable.

Description

Technical Field

[0001] This invention relates to the field of machining technology, and in particular to a fault diagnosis method, apparatus, equipment, and readable storage medium. Background Technology

[0002] Pneumatic modules are core actuators in automated production lines of intelligent manufacturing, widely used in the precise handling and attachment of various materials. Their operational reliability directly determines the continuous operating efficiency of the production line, product attachment yield, and overall production cost. Current technologies lack a systematic ability to predict the operating status of pneumatic modules and the lifespan of nozzle consumables, relying heavily on single-parameter monitoring or manual experience assessment. This makes it difficult to comprehensively capture changes in equipment operating status under complex conditions and accurately identify the gradual wear and tear of core consumables. Consequently, equipment failures can only be addressed reactively after shutdown, leading to either premature replacement of core consumables, wasting resources, or delayed replacement, causing downtime losses and severely impacting the stability and economy of the production line. Summary of the Invention

[0003] The main objective of this invention is to propose a fault diagnosis method, apparatus, device, and readable storage medium, aiming to solve the problems in the prior art of lacking systematic advance prediction of the operating status of pneumatic modules and the lifespan of nozzle consumables, relying on single parameters or manual experience to passively troubleshoot faults, and unreasonable replacement of consumables.

[0004] To achieve the above objectives, the fault diagnosis method proposed in this invention includes: Real-time acquisition of multi-dimensional operational data during the operation of nozzle-attach pneumatic modules; The multi-dimensional operational data is categorized and processed according to the working period of the suction nozzle to obtain the target operational data; The target operating data is input into the diagnostic prediction model, which identifies the pneumatic module fault type and nozzle consumable life status, and outputs fault diagnosis results and consumable life prediction information.

[0005] The pneumatic module fault diagnosis method provided by this invention solves the problems in existing technologies, such as the lack of systematic advance prediction of pneumatic module operating status and nozzle consumable lifespan, reliance on single parameters or manual experience leading to passive fault diagnosis, and unreasonable consumable replacement, by combining real-time acquisition of multi-dimensional operating data, classification processing according to nozzle working time, and analysis by a diagnostic prediction model. Specifically, it first collects multi-dimensional operating data in real time during the operation of nozzle-attaching pneumatic modules, comprehensively covering various status parameters of the equipment under complex working conditions, breaking the limitations of single-parameter monitoring; then, it classifies the multi-dimensional operating data according to nozzle working time, filtering out target operating data directly related to the core nozzle attachment process, improving the data's relevance and effectiveness; finally, it inputs the target operating data into the diagnostic prediction model, and through the model's intelligent analysis capabilities, accurately identifies the fault type of the pneumatic module and the gradual wear process of the nozzle consumables, ultimately outputting clear fault diagnosis results and consumable lifespan prediction information. In this way, multi-dimensional data acquisition solves the problem that single-parameter monitoring is insufficient to fully capture changes in equipment status under complex working conditions, ensuring that no status information is missed. Categorizing and processing data according to the working period of the nozzle allows the data to focus on core processes, providing accurate input for model analysis and improving the accuracy of diagnosis and prediction. The diagnostic and prediction model realizes the transformation from "passive investigation" to "proactive prediction," which can not only identify fault symptoms in advance to avoid sudden downtime, but also predict the life of nozzle consumables, making the timing of consumable replacement more reasonable. This reduces the waste of resources caused by premature replacement and avoids downtime losses caused by delayed replacement, ultimately ensuring the continuous operation efficiency of the production line, improving product bonding yield, and reducing overall production costs. Attached Figure Description

[0006] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0007] Figure 1 The system schematic diagram provided by this invention; Figure 2 This is a flowchart illustrating a first embodiment of a fault diagnosis method according to the present invention; Figure 3 This is a flowchart illustrating a second embodiment of a fault diagnosis method according to the present invention; Figure 4 This is a flowchart illustrating a third embodiment of a fault diagnosis method according to the present invention; Figure 5 This is a flowchart illustrating the fourth embodiment of a fault diagnosis method according to the present invention; Figure 6 This is a structural block diagram of the first embodiment of the fault diagnosis device of the present invention; Figure 7 This is a schematic diagram of the fault diagnosis device structure of the hardware operating environment involved in the embodiment of the present invention.

[0008] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0009] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0010] Reference Figure 7 , Figure 7 This is a schematic diagram of the fault diagnosis device structure of the hardware operating environment involved in the embodiment of the present invention.

[0011] like Figure 7 As shown, the fault diagnosis device may include: a processor 1001 (i.e., a central controller PLC), a communication bus 1002, a user interface 1003, a network interface 1004, a memory 1005, and an analog module (not shown in the figure). The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen (i.e., a touch screen) and input units such as touch buttons. Optionally, the user interface 1003 may also include standard wired interfaces and wireless interfaces (such as RS485 interfaces). The network interface 1004 may optionally include standard wired interfaces and wireless interfaces (such as Ethernet interfaces and Wi-Fi interfaces) for data interaction with peripherals such as vision modules and various sensors. The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as an industrial-grade SD card or a solid-state drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001. The analog module is connected to the processor 1001 and is used to receive signals from analog output devices such as negative pressure detection sensors and air pressure detection sensors, and convert them into digital signals that can be recognized by the processor 1001.

[0012] Those skilled in the art will understand that Figure 7The structure shown does not constitute a limitation on the fault diagnosis equipment and may include more or fewer components than shown, or combine certain components, or have different component arrangements. For example, it may also include solenoid valve drive circuits, alarm mechanism interfaces, cylinder control interfaces, loading and unloading mechanism drive modules, vacuum generator control interfaces, etc., to adapt to the complete control requirements of nozzle attachment pneumatic modules, corresponding one-to-one with the system components shown in Figure 1 (fault and life prediction system schematic diagram).

[0013] like Figure 7 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a network communication module, a user interface module, and a fault diagnosis program (i.e., a fault prediction program for the nozzle attachment pneumatic module and the lifespan prediction program for nozzle consumables).

[0014] exist Figure 7 In the fault diagnosis device shown, the network interface 1004 is mainly used for data communication with peripherals such as the vision module, air pressure detection sensor, and gas flow sensor; the analog signal module is mainly used to receive the analog signal output by the negative pressure detection sensor, convert it, and transmit it to the processor 1001; the user interface 1003 is mainly used for data interaction with the user, such as displaying device status, alarm information, sensor values, etc. through the touch screen, and receiving parameter thresholds and operation commands set by the user; the processor 1001, as the core control unit, is also used to control the movement of the loading and unloading conveying mechanism, the position trajectory of the motion axis, the movement of the nozzle and cylinder, and to collect and process the values ​​of various sensors; the processor 1001 and the memory 1005 in the fault diagnosis device of the present invention can be set in the fault diagnosis device, and the fault diagnosis device calls the fault diagnosis program stored in the memory 1005 through the processor 1001 and executes the fault diagnosis method provided in the embodiment of the present invention.

[0015] Please refer to Figure 1, which is a schematic diagram of the system of this invention (i.e., the core architecture diagram of the fault diagnosis system). This system achieves the entire process of data acquisition, control execution, fault diagnosis, and lifespan prediction through the collaborative work of its components. The specific connection relationships and functional implementations of each component are as follows: Central controller (PLC, corresponding to) Figure 7The processor 1001, as the core control unit of the system, establishes control and data transmission links with components such as analog modules, touch screens, solenoid valves, vacuum generators, and motor drivers via a communication bus. Its core functions include: receiving data collected by various sensors (such as negative pressure values, air pressure values, and gas flow rates), processing it, and performing fault diagnosis and lifespan prediction calculations; outputting control commands to control the loading and unloading actions of the material handling mechanism, controlling the movement of the motion axis along a preset trajectory, and controlling the extension and retraction of the cylinder and the suction and release actions of the nozzle; and receiving parameter setting commands from the touch screen and feeding back system operating parameters, equipment status, and alarm information to the touch screen display.

[0016] Analog module (corresponding) Figure 7 Analog Module: One end connects to the negative pressure detection device, air pressure detection sensor, and gas flow sensor via a signal line to receive analog output signals; the other end connects to the central controller via a communication bus to convert the analog signals into digital signals and transmit them to the central controller, ensuring accurate acquisition and transmission of sensor data.

[0017] Negative pressure detection device (negative pressure detection sensor): Installed on the gas pipeline between the nozzle and the vacuum generator, it is connected to the analog module through a signal line to collect the negative pressure value at different time periods during the operation of the nozzle (adsorption start, adsorption in place, transmission and attachment period) in real time, and transmit the signal to the analog module to provide data support for adsorption status judgment and nozzle life assessment.

[0018] Air pressure detection sensor: Installed on the main air supply pipeline of the pneumatic module, it is connected to the digital interface or analog module of the central controller through a signal line. It collects real-time air pressure data throughout the entire operation of the equipment, classifies it according to the working status of the nozzle, and transmits it to the central controller. The central controller generates corresponding air pressure curves to reflect the stability of the air supply system.

[0019] Gas flow sensor: Installed on the air inlet pipe of the vacuum generator, it is connected to the high-speed counter interface or analog module of the central controller via a signal line. It collects real-time gas flow data throughout the entire operation of the equipment, classifies it according to the working status of the nozzle, and transmits it to the central controller. The central controller generates corresponding gas flow curves to help determine whether there are abnormalities such as leaks or blockages in the gas path.

[0020] Touchscreen (corresponding) Figure 7The display screen of the user interface 1003 is connected to the central controller via a communication bus. As a human-machine interface, its functions include: receiving user operation commands, setting parameters of various transmission mechanisms, motion axes, and pneumatic modules (such as motor speed, action delay, vacuum parameter threshold, etc.); displaying various parameters of the current system, including motor position, values ​​of various sensors, equipment operating status, alarm status, etc., so that users can intuitively grasp the system operation status.

[0021] Suction nozzle: As the core component for performing adsorption and adhesion actions, it is connected to the vacuum generator through an air pipeline and linked to the cylinder through a mechanical structure; it receives adsorption / release commands from the central controller, generates negative pressure to adsorb materials under the action of the vacuum generator, or releases negative pressure to release materials; its head condition (wear, deformation, aging, damage) is collected and analyzed by a vision module to provide a basis for life prediction.

[0022] Cylinder: It is connected to the solenoid valve and vacuum generator through the air pipeline, and to the solenoid valve drive circuit of the central controller through the signal line; it receives the extension and retraction command of the central controller and drives the suction nozzle to realize the lifting, translating and other actions to ensure that the suction nozzle accurately reaches the adsorption position and the attachment position; its extension and retraction status is collected by the cylinder position sensor and fed back to the central controller to determine whether the action is in place.

[0023] The loading and unloading conveying mechanism includes components such as a motor, a motor driver, and a conveying guide rail. The motor driver is connected to the central controller via a signal line, receives transmission commands from the central controller, drives the motor to run, and drives the conveying guide rail to realize the loading and unloading conveying action, transporting the tooled products to the designated adsorption position, or transporting the finished products to the unloading area.

[0024] Material loading and unloading mechanism: The mechanical arm is linked with the cylinder and the suction nozzle, and its movement trajectory is controlled by the central controller. It assists the suction nozzle in picking up, placing and transferring materials. Its positioning accuracy data is collected by the position sensor and transmitted to the central controller to determine whether there is a positioning deviation in the mechanism.

[0025] Vision Module: This includes an upper vision module and a lower vision module. The upper vision module is mounted on the upper and lower transmission modules, while the lower vision module is fixed to the machine base along the upper and lower transmission path. Both are connected to the central controller via Ethernet. The function of the upper vision module is to photograph the position of the tooling product after it has been transported to its designated position, generate position coordinate data, and transmit it to the central controller to guide the nozzle alignment and attachment. The function of the lower vision module is to photograph the nozzle head and transmit it to the central controller to assess the nozzle's wear level and any abnormal conditions.

[0026] Alarm mechanism: It is connected to the alarm drive circuit of the central controller via a signal line, receives alarm commands from the central controller, and issues an audible and visual alarm signal when a serious fault is detected (such as nozzle damage or sudden drop in supply pressure) or when consumables need to be replaced immediately, so as to remind on-site operators to deal with the problem in a timely manner.

[0027] Solenoid valve and vacuum generator: The solenoid valve is connected to the central controller via a signal line and receives control commands from the central controller to control the opening and closing of the air circuit; the vacuum generator is connected to the solenoid valve and the suction nozzle via an air circuit pipeline, and generates negative pressure under the control of the solenoid valve to provide power for the suction nozzle to adsorb materials. Its working status is monitored in real time by the central controller.

[0028] Please refer to Figure 2, which is a flowchart illustrating a first embodiment of a fault diagnosis method according to the present invention. In one embodiment, the fault diagnosis method proposed by the present invention includes the following steps: S100: Real-time acquisition of multi-dimensional operational data during the operation of nozzle-attach pneumatic modules; It should be noted that the execution entity in this embodiment is the processor 1001 (i.e., the central controller PLC) of the fault diagnosis device. The fault diagnosis device works collaboratively with the system components shown in Figure 1 to jointly complete data acquisition, processing, fault diagnosis and life prediction.

[0029] Multi-dimensional operational data is a core dataset reflecting the working status of the pneumatic module and the performance of the nozzle consumables. Its core function is to provide comprehensive and accurate raw data support for fault diagnosis and lifespan prediction, avoiding diagnostic biases caused by missing data. This dataset specifically includes air pressure parameters, gas flow parameters, vacuum negative pressure values, the number of nozzle suction failures, nozzle arrival time, and nozzle head image data. It covers pneumatic system operating parameters, nozzle working status parameters, and visual inspection data, comprehensively covering key dimensions affecting equipment failure and consumable lifespan. Among these, air pressure parameters also need to be correlated with the installation status of the air supply device and pipelines, and mechanical wear, indirectly reflecting air pressure instability caused by improper installation or pipeline aging.

[0030] In this step, the processor 1001 collects various types of data through the system connection link shown in Figure 1 in the following ways: It receives analog signals from the negative pressure detection sensor via the analog module, converts them into digital signals to obtain vacuum negative pressure parameters; it receives signals from the air pressure detection sensor via the digital interface or analog module to obtain air pressure parameters; it receives pulse signals from the gas flow sensor via the high-speed counter interface, converts them into flow values ​​to obtain gas flow parameters; it receives tooling product position data acquired by the upper vision module and nozzle head image data acquired by the lower vision module via the Ethernet interface of the vision module; it collects cylinder extension / retraction status data via the cylinder position sensor and mechanism motion accuracy data via the loading / unloading mechanism status sensor; it counts the number of suction failures and nozzle arrival time within a single working cycle; and it receives vacuum parameter threshold settings transmitted from the touchscreen. The acquisition frequency is set to 50Hz to ensure real-time and continuous data acquisition. Each acquired data is accompanied by a millisecond-level timestamp, which is associated with the nozzle's action commands (adsorption start command, adsorption stop command), cylinder extension / retraction commands, and loading / unloading mechanism action commands, providing a precise time reference for subsequent time-based classification processing. During the data acquisition process, various sensors work synchronously, and the processor 1001 stores the data in real time to the data buffer of the memory 1005 to ensure that no data is lost.

[0031] S200. Classify the multi-dimensional operation data according to the working period of the suction nozzle to obtain the target operation data; It should be noted that the nozzle's working time is divided into core action periods and non-core action periods based on its working logic. The core action periods include the suction start period, the suction completion period, and the suction transfer and attachment period. The non-core action periods are the nozzle closing periods (i.e., the intervals during which no suction action is performed). The core purpose of this period division is to categorize the chaotic raw data according to the working scenario, strengthen the correlation between the data and the module's working status, and provide structured data for subsequent model calculations.

[0032] The adsorption start period refers to the time from when the processor 1001 issues the adsorption command to the nozzle until the negative pressure reaches the preset adsorption threshold (e.g., -65kPa). This period reflects the initial adsorption capacity of the nozzle. The adsorption completion period refers to the time from when the negative pressure reaches the adsorption threshold until the material carried by the nozzle stabilizes and comes to rest. This period reflects the adsorption stability. The adsorption, transport, and attachment period refers to the time from when the material carried by the nozzle begins to move until the material completes attachment and the negative pressure is released. This period reflects the continuity of the adsorption state during transport. The nozzle closing period refers to the time from when the processor 1001 issues the adsorption closing command until the next adsorption opening command is issued. This period reflects the stability of the pneumatic parameters in the system's standby state.

[0033] In this step, processor 1001 first adds corresponding time period tags to each piece of raw data based on the timestamps accompanying the data and the records of nozzle action commands, cylinder action commands, and loading / unloading mechanism action commands. Then, it integrates similar data within the same time period to form a time period-data correspondence; for example, all vacuum negative pressure value data during the adsorption start period are categorized into an adsorption start negative pressure dataset. Next, through data validity verification, invalid and redundant data that exceeds reasonable ranges, is duplicated, or lacks key information is removed to avoid interfering with subsequent model calculations. Finally, the categorized data are associated according to the work cycle to form structured target operating data, ensuring data integrity and logic. Simultaneously, a separate threshold configuration dataset is established for vacuum parameter threshold setting data and associated with the corresponding work cycle's operating data to analyze whether the threshold settings are too high or too low.

[0034] S300. Input the target operating data into the diagnostic prediction model, identify the pneumatic module fault type and nozzle consumable life status through the diagnostic prediction model, and output the fault diagnosis result and consumable life prediction information.

[0035] It should be noted that the diagnostic prediction model is a multi-dimensional ensemble algorithm model pre-trained based on a large amount of experimental data and historical operating data, stored in memory 1005, and has the dual functions of fault type identification and consumable life status assessment. The core of this model is to achieve accurate determination of equipment status through multi-dimensional feature vector matching and threshold comparison, replacing the traditional judgment method that relies on human experience.

[0036] Fault types include pneumatic module pipeline leaks, air circuit blockages, insufficient pressure supply, excessive pressure supply, abnormal vacuum parameter threshold settings (too high or too low), uneven contact surfaces between the nozzle and auxiliary materials, abnormal cylinder operation, and positioning deviations in the loading and unloading mechanism. Among these, pipeline leaks and insufficient pressure supply may be caused by improper installation of the air supply device or mechanical wear of the pipeline. The life status of the nozzle consumables includes normal status, life warning status, and immediate replacement status. Different statuses correspond to different parameter threshold ranges and characteristic manifestations (such as nozzle head wear, deformation, aging, and damage). Fault diagnosis results include whether a fault exists, the specific type of fault, the time of fault occurrence, fault cause analysis, and recommended handling measures. Consumable life prediction information includes life status, remaining life percentage, and recommended replacement time.

[0037] In this step, after the processor 1001 inputs the structured target operating data into the diagnostic prediction model, the model first extracts key features from the data (such as negative pressure fluctuation amplitude, air pressure stability, abnormal flow rate, nozzle wear area, the matching degree between vacuum parameter thresholds and actual needs, cylinder action response speed, and positioning accuracy of loading and unloading mechanisms, etc.), and compares them with the preset fault feature library and life assessment standards. Then, it identifies whether the pneumatic module has a fault and the specific type of fault, while simultaneously assessing the remaining life of the nozzle consumables and determining their corresponding status. Finally, it outputs the fault diagnosis results and consumable life prediction information, which are pushed in two ways: first, to the touchscreen (user interface 1003) for intuitive visualization; second, to the system backend to form a log record, facilitating management personnel to coordinate maintenance plans. If a serious fault is detected (such as a sudden drop in supply pressure, nozzle damage, or cylinder jamming), the processor 1001 simultaneously triggers the alarm mechanism, issuing an audible and visual alarm signal to remind on-site operators to handle the situation promptly.

[0038] The pneumatic module fault diagnosis method provided by this invention solves the problems in existing technologies, such as the lack of systematic advance prediction of pneumatic module operating status and nozzle consumable lifespan, reliance on single parameters or manual experience leading to passive fault diagnosis, and unreasonable consumable replacement, by combining real-time acquisition of multi-dimensional operating data, classification processing according to nozzle working time, and analysis by a diagnostic prediction model. Specifically, it first collects multi-dimensional operating data in real time during the operation of nozzle-attaching pneumatic modules, comprehensively covering various status parameters of the equipment under complex working conditions, breaking the limitations of single-parameter monitoring; then, it classifies the multi-dimensional operating data according to nozzle working time, filtering out target operating data directly related to the core nozzle attachment process, improving the data's relevance and effectiveness; finally, it inputs the target operating data into the diagnostic prediction model, and through the model's intelligent analysis capabilities, accurately identifies the fault type of the pneumatic module and the gradual wear process of the nozzle consumables, ultimately outputting clear fault diagnosis results and consumable lifespan prediction information. In this way, multi-dimensional data acquisition solves the problem that single-parameter monitoring is insufficient to fully capture changes in equipment status under complex working conditions, ensuring that no status information is missed. Categorizing and processing data according to the working period of the nozzle allows the data to focus on core processes, providing accurate input for model analysis and improving the accuracy of diagnosis and prediction. The diagnostic and prediction model realizes the transformation from "passive investigation" to "proactive prediction," which can not only identify fault symptoms in advance to avoid sudden downtime, but also predict the life of nozzle consumables, making the timing of consumable replacement more reasonable. This reduces the waste of resources caused by premature replacement and avoids downtime losses caused by delayed replacement, ultimately ensuring the continuous operation efficiency of the production line, improving product bonding yield, and reducing overall production costs.

[0039] Thus, the fault diagnosis method provided by this invention, through a dynamic process of "real-time data acquisition - time period classification processing - precise model judgment," can solve the problems of traditional nozzle-attach pneumatic modules that rely on manual experience to judge faults, cannot predict consumable lifespan, rely on experience to set vacuum parameter thresholds, and suffer from consumable waste, high equipment downtime rate, and low production yield due to the lack of real-time monitoring of cylinder and pick-and-place mechanism status. Specifically, the collaborative acquisition of system components shown in Figure 1 ensures the comprehensiveness and structure of data; the diagnostic prediction model realizes automatic judgment of faults and lifespan, replacing manual experience judgment and reducing the professional requirements for maintenance personnel; by providing early warning of faults and consumable lifespan status, it is convenient to formulate planned maintenance and replacement plans, reduce equipment downtime and consumable waste, and improve equipment operation stability and production yield; and by analyzing the rationality of vacuum parameter threshold settings, it avoids adsorption abnormalities caused by thresholds that are too high or too low.

[0040] Please refer to Figure 3 , Figure 3 This is a flowchart illustrating a second embodiment of a fault diagnosis method according to the present invention. In this embodiment, the multi-dimensional operational data includes gas pressure parameters, gas flow rate parameters, vacuum negative pressure value parameters, number of suction failures by the suction nozzle, and suction nozzle positioning time; step S100 includes: S110. Collect the vacuum negative pressure value parameters during the adsorption process of the nozzle through the negative pressure detection sensor, and divide the data into adsorption start time, adsorption completion time and adsorption transfer and adhesion time according to the adsorption stage. It should be noted that the negative pressure detection sensor is a high-precision pressure sensor (such as the SMC ZSE30A series), installed on the gas path between the nozzle and the vacuum generator. Its measurement range is -100kPa to 0kPa, with a measurement accuracy of ±0.1kPa, enabling it to accurately capture the dynamic changes in negative pressure during nozzle adsorption. The core function of this sensor is to convert the physical quantity of vacuum negative pressure into a transmittable electrical signal, providing quantitative data for judging the nozzle adsorption status and consumable lifespan.

[0041] The division of the adsorption start period, adsorption completion period, and adsorption transfer and attachment period is consistent with the first embodiment. The vacuum negative pressure value parameters of different periods reflect different nozzle states: the negative pressure value of the adsorption start period reflects the initial adsorption capacity. If the negative pressure value rises slowly, it may indicate nozzle wear or slight gas leakage. The negative pressure value of the adsorption completion period should remain stable. If the fluctuation range exceeds ±2kPa, there may be an adsorption instability problem. The negative pressure value of the adsorption transfer and attachment period reflects the continuity of adsorption during the transfer process. If the negative pressure value drops sharply, it may indicate that the material has fallen or the nozzle seal has failed.

[0042] In this step, the negative pressure detection sensor collects the negative pressure value at different times during the adsorption process of the nozzle in real time. The value is transmitted to the analog module of the processor 1001 via an analog signal (4-20mA). The analog module converts the value into a digital signal, which is then stored in the memory 1005 by the processor 1001 according to the corresponding time period label. This forms the adsorption start negative pressure dataset, the adsorption completion negative pressure dataset, and the transmission and attachment negative pressure dataset. Each dataset is arranged in the order of timestamps to ensure the time sequence of the data.

[0043] S120. The air pressure parameters and gas flow parameters of the pneumatic module during the entire operation process are collected by the air pressure detection sensor and the gas flow sensor, respectively, and the data are divided into the suction nozzle opening period and the suction nozzle closing period according to the suction nozzle status. It should be noted that the air pressure detection sensor (such as Festo SDE5-D10-O-Q6E-P-M8) is installed on the main air supply pipeline of the pneumatic module. The measurement range is 0~1MPa and the measurement accuracy is ±0.005MPa. It is used to monitor the air supply pressure status of the pneumatic system. Its core function is to capture the dynamic and static changes in the air supply pressure and provide quantitative basis for judging whether there is an abnormality in the air supply system.

[0044] The nozzle-opening period is the period during which the nozzle performs adsorption-related actions (including three sub-periods: adsorption start, adsorption in place, and adsorption transfer and attachment). The air pressure parameters during this period reflect the stability of the air supply when the adsorption action is performed. The normal operating pressure range is preset to 0.5~0.6MPa. The nozzle-closed period is the system standby period. The air pressure parameters during this period reflect the ability to maintain the air supply pressure. The normal pressure fluctuation should not exceed ±0.01MPa.

[0045] In this step, the air pressure sensor transmits the collected air pressure data to the processor 1001 in real time through the digital signal interface (PNP output). The processor 1001 records the data based on the nozzle action command and classifies and stores the data according to the nozzle opening period and the nozzle closing period, forming air pressure datasets for the nozzle opening period and air pressure datasets for the nozzle closing period. Each dataset contains the air pressure value, timestamp and fluctuation range information for the corresponding period, providing a basis for subsequent data processing.

[0046] S130. The pneumatic module control system counts the number of suction failures and the time it takes for the suction nozzle to complete its action within a single working cycle. It should be noted that the processor 1001 of the pneumatic module control system, which is also the fault diagnosis equipment, uses a dual verification method to count the number of suction failures: first, the negative pressure detection sensor fails to detect a negative pressure value reaching the adsorption threshold within a preset time (e.g., 800ms); second, the vision module fails to detect successful material adsorption (i.e., no material outline is recognized in the image). Meeting either condition is considered a suction failure. The core function of this statistical data is to reflect the adsorption reliability of the suction nozzle. Too many failures usually indicate nozzle wear, aging, or abnormal airflow.

[0047] The nozzle positioning time refers to the total time from when the processor 1001 issues the adsorption command until the nozzle moves to the preset adsorption position and completes adsorption (the negative pressure reaches the adsorption threshold). It is counted by the timer of the processor 1001 with a timing accuracy of 1ms. The core function of this time parameter is to indirectly reflect the mechanical state and adsorption efficiency of the nozzle. An excessively long positioning time may indicate nozzle jamming, wear, or insufficient air pressure.

[0048] In this step, after a single working cycle of the suction nozzle ends, the processor 1001 automatically accumulates the number of suction failures within that cycle and records the time when the suction nozzle reaches its position. It then associates and stores these two data points with the corresponding working cycle number to ensure that a complete data link is formed with other multi-dimensional data within that cycle.

[0049] S140. Image data of the nozzle head is acquired through the vision module to analyze the wear level and abnormal condition of the nozzle.

[0050] It should be noted that the vision module includes an upper vision and a lower vision (such as the Keyence IV2 series vision sensors). The upper vision is mounted on the upper and lower transmission module, while the lower vision is fixed on the machine tool along the upper and lower transmission path. The two work together to achieve omnidirectional imaging of the nozzle head. The core function of this module is to intuitively capture the physical state of the nozzle head through image data, providing direct evidence for analyzing anomalies such as wear, deformation, and damage, thus overcoming the limitations of single-parameter detection.

[0051] The data acquisition timing is after every 15 adsorption-attachment cycles. The processor 1001 controls the nozzle to move to the preset image position (avoiding the interference area of ​​the workstation). The lower vision takes a bottom image of the nozzle head from below, and the upper vision takes a side image of the nozzle head from the side. The image resolution is 1280×960 pixels, and the image format is BMP, which is convenient for subsequent image processing and feature extraction.

[0052] In this step, after the vision module completes the image capture, the image data is transmitted to the processor 1001 via Ethernet. The processor 1001 associates and stores the image data with the corresponding work cycle number, providing raw image support for subsequent extraction of feature parameters such as nozzle wear area and deformation.

[0053] This embodiment ensures the comprehensiveness, accuracy, and relevance of data by clearly defining the data collection subjects, collection periods, collection accuracy, and transmission methods for various multi-dimensional operational data. Different types of data reflect the equipment status and consumable performance from different dimensions, complementing and verifying each other. This provides high-quality raw data support for subsequent data processing and model calculations, avoiding diagnostic biases caused by missing or inaccurate data, and laying a reliable foundation for fault diagnosis and lifespan prediction.

[0054] Please refer to Figure 4 , Figure 4 This is a schematic flowchart of a third embodiment of a fault diagnosis method according to the present invention. In this embodiment, step S200 includes: S210. Remove invalid and redundant data from the data during the nozzle opening period and the data during the nozzle closing period.

[0055] It should be noted that invalid and redundant data refer to data that cannot reflect the true state of the device, contains errors, or is duplicated. The core purpose of removing this type of data is to improve data quality and avoid interfering with subsequent model calculations. Invalid data includes outliers that are outside the normal range (such as extreme values ​​caused by sensor malfunctions), and redundant data includes duplicate records and invalid data generated by data acquisition interruptions.

[0056] In this step, the processor 1001 first records the suction nozzle action instructions, dividing the pressure parameters into a pressure dataset for the suction nozzle opening period (labeled "Q1-XXX") and a pressure dataset for the suction nozzle closing period (labeled "Q2-XXX"). Similarly, the gas flow parameters are divided into a flow dataset for the suction nozzle opening period (labeled "L1-XXX") and a flow dataset for the suction nozzle closing period (labeled "L2-XXX"). Each dataset is arranged in ascending order of timestamps. Then, data validity thresholds are set (such as pressure parameter fluctuation thresholds, flow parameter stability thresholds, and data acquisition time interval thresholds). Based on these thresholds, the datasets are filtered to remove invalid and redundant data. Finally, the filtered datasets are completed. If a single data point is missing (timestamp interval within 30ms), the missing data is supplemented using linear interpolation to ensure the continuity and integrity of the dataset.

[0057] S220. Based on a single work cycle, associate the number of suction failures of the nozzle, the time of the nozzle action, the negative pressure data at the start of suction, the negative pressure data at the end of suction, the negative pressure data of suction transmission and attachment, the dataset of the nozzle opening period after removing the invalid and redundant data, and the dataset of the nozzle closing period to form a target operation data set indexed by the work period.

[0058] It should be noted that the core purpose of data association is to establish a complete data link based on a single working cycle of the suction nozzle, integrating parameter data and operational status data of different types and time periods to form a structured target operational data set, ensuring the correlation and integrity of the data, and providing comprehensive and coherent input data for model calculation.

[0059] The working period index uses the start timestamp of a single working cycle of the nozzle as a unique identifier to generate a working cycle number (such as "Cycle-XXX"). Each working cycle number corresponds to a complete adsorption-attachment cycle and associates all data within that cycle to ensure that the data corresponds one-to-one with the working cycle.

[0060] In this step, the processor 1001 first uses the work cycle number as an index to bind the adsorption start negative pressure data (N1-XXX), adsorption completion negative pressure data (N2-XXX), and transfer attachment negative pressure data (N3-XXX) within the same work cycle with the suction nozzle opening period air pressure dataset (Q1-XXX) and suction nozzle opening period flow dataset (L1-XXX) to form a core action data group. Then, it associates the suction nozzle closing period air pressure dataset (Q2-XXX) and suction nozzle closing period flow dataset (L2-XXX) within the same work cycle with the core action data group to supplement the pneumatic system parameters during non-working periods. Finally, it adds the number of suction nozzle failures and the suction nozzle action completion time within the work cycle to the data group to form a complete target operation data entry containing pneumatic parameters and operating status parameters. Multiple target operation data entries are arranged in ascending order according to the work cycle number to form a target operation data set, which is stored in the extended storage area of ​​the memory 1005 in a structured data format (such as JSON format) to facilitate model calling and data traceability.

[0061] Please refer to Figure 5 , Figure 5 This is a flowchart illustrating a fourth embodiment of a fault diagnosis method according to the present invention. In this embodiment, a data validity determination threshold is set, which includes a pressure parameter fluctuation threshold, a gas flow parameter stability threshold, and a data acquisition time interval threshold; step S220 includes: S221. If the fluctuation range of the air pressure parameter or the gas flow rate parameter exceeds the fluctuation threshold of the air pressure parameter or the stability threshold of the gas flow rate parameter within the same time period, it is determined to be invalid data and discarded. It should be noted that the data validity threshold is the core basis for distinguishing between valid data and invalid / redundant data. Its setting is based on the design parameters of the nozzle-attached pneumatic module, industry standards, and a large amount of historical operating data to ensure the rationality and relevance of the threshold. At the same time, it supports users to dynamically adjust it according to actual working conditions via the touch screen.

[0062] Pressure fluctuation thresholds: For the nozzle-open period, the threshold is set to ±0.03 MPa, as the supply pressure experiences normal dynamic fluctuations during suction, resulting in a relatively wide threshold range. For the nozzle-closed period, the threshold is set to ±0.01 MPa, as the supply pressure should remain stable in standby mode, requiring a strict threshold range. The core function of this threshold is to determine whether the pressure data fluctuations are within the normal range; values ​​exceeding this range may indicate invalid data.

[0063] Gas flow rate stability threshold: ±1.5 L / min during nozzle-on period, allowing for normal fluctuations in gas flow during adsorption; ±0.3 L / min during nozzle-off period, ensuring no significant flow rate change in standby mode. This threshold is crucial for determining the stability of flow data; exceeding it may indicate gas path abnormalities or invalid data. Data acquisition interval threshold: 30 ms, set in conjunction with the sensor's acquisition frequency (50 Hz) to ensure continuous data acquisition and prevent data loss or redundancy due to acquisition interruptions. This threshold is also crucial for determining the continuity of data acquisition; exceeding the interval may indicate redundant data resulting from acquisition interruptions.

[0064] In this step, the processor 1001 stores the aforementioned thresholds in a parameter register, which the user can modify via the touchscreen. After modification, the data acquisition process must be restarted for the changes to take effect, ensuring the flexibility and standardization of threshold adjustments. For example, in the air pressure data sequence during the nozzle-on period, if the difference between a data point and the previous data point is 0.05 MPa, exceeding the fluctuation threshold of ±0.03 MPa, and the deviation of this data point from the average value of the dataset exceeds twice the standard deviation, it is considered invalid data. Similarly, in the flow rate data during the nozzle-off period, if a data point has a value of 1.2 L / min, exceeding the fluctuation threshold of ±0.3 L / min, and there is no reasonable explanation for the operating conditions, it is also considered invalid data.

[0065] In this step, the processor 1001 calculates the fluctuation amplitude of the air pressure dataset and flow dataset for each data point in the same time period, compares it with the corresponding fluctuation threshold, deletes invalid data that exceeds the threshold from the dataset, and records the deletion log (including data point value, timestamp, and reason for deletion) for easy traceability later.

[0066] S222. If the repetition rate of the gas pressure parameter or the gas flow parameter collected in the same time period exceeds a preset ratio, or the collection time interval exceeds the data collection time interval threshold, it is determined to be redundant data and is removed.

[0067] It should be noted that the repetition rate refers to the proportion of consecutively repeated parameter data points within the same time period to the total number of data points. For example, if the preset proportion is set to 30%, then data points with a consecutive repetition rate exceeding 30% are considered redundant data. The acquisition time interval refers to the difference in timestamps between two adjacent data points; if it exceeds 30ms, it is considered redundant data generated by an acquisition interruption. The core of this judgment logic is to eliminate meaningless, repetitive, or discontinuous data to avoid consuming storage resources and interfering with model calculations.

[0068] For example, if four or more consecutive air pressure data points within the same time period have completely identical values ​​(error less than ±0.001 MPa), and the duplication rate exceeds 30%, then the first data point is retained, and subsequent duplicate data are deleted. Similarly, if the timestamp interval between two adjacent flow rate data points is 50 ms, and the data exceeds the 30 ms acquisition time interval threshold, it is considered redundant and discarded. In this step, the processor 1001 first performs a duplicate detection on the datasets from the same time period, calculates the duplicate rate and compares it with a preset ratio to remove duplicate and redundant data; then it checks the timestamp interval of the data points and removes redundant data that exceeds the threshold of the collection time interval; finally, it reorders the timestamps of the datasets after deleting redundant data to ensure the temporal continuity of the data sequence.

[0069] In one embodiment, the associated steps in step S230 include: S231. Using the single working cycle of the suction nozzle as the time reference, establish a unique time period index identifier. The time period index identifier corresponds one-to-one with the suction start time period, the suction arrival time period, the transmission and attachment time period, the suction nozzle opening time period, and the suction nozzle closing time period. It should be noted that a single working cycle of the suction nozzle refers to the complete cycle from the issuance of the suction command to the completion of attachment, negative pressure release, and entry into standby mode. The duration is usually a fixed value (e.g., 5 seconds) and can be adjusted according to actual working conditions. The core purpose of establishing a unique time period index is to assign a unique identifier to each working cycle and each time period within it, achieving precise binding between data and time periods, and providing a unified index basis for subsequent data association.

[0070] The time period index identifier consists of "work cycle number - time period type code". The work cycle number is the cumulative number of work cycles since the processor 1001 started the device (e.g., the 200th work cycle number is "200"). The time period type code is used to distinguish different nozzle working time periods. The specific coding rules are as follows: adsorption start time period (01), adsorption in place time period (02), adsorption transfer and attachment time period (03), nozzle open time period (04), and nozzle close time period (05). Each time period index identifier uniquely corresponds to a specific time period within a work cycle. For example, "200-01" corresponds to the adsorption start time period of the 200th work cycle.

[0071] In this step, the processor 1001 automatically generates a work cycle number at the beginning of each work cycle, combines it with the time period type code to form a time period index identifier, stores it in the index register, and associates it with the start timestamp and end timestamp of each time period to form an index-time period mapping table, which facilitates fast lookup when data is associated.

[0072] S232. Bind the adsorption start negative pressure data, adsorption completion negative pressure data or transmission attachment negative pressure data under the same time period index to the corresponding gas pressure parameter dataset and gas flow parameter dataset for the nozzle opening period or the nozzle closing period to form a parameter data group for a single time period. It should be noted that the core purpose of the binding operation is to integrate parameter data of different types within the same time period into a complete parameter data group for a single time period, ensuring data consistency across time periods and avoiding data confusion between different time periods. The binding logic corresponding to the index identifiers of different time periods is different, and precise matching based on the time period type code is required.

[0073] For example, the time period index identifier "N-01" (the adsorption start time of the Nth working cycle) corresponds to the adsorption start negative pressure data (N1-XXX), and the adsorption start time belongs to the nozzle opening time. Therefore, this adsorption start negative pressure data is bound to the corresponding sub-datasets in the nozzle opening time pressure dataset (Q1-XXX) and the nozzle opening time flow dataset (L1-XXX). The time period index identifier "N-05" (the nozzle closing time of the Nth working cycle) has no negative pressure data, so only the nozzle closing time pressure dataset (Q2-XXX) and the nozzle closing time flow dataset (L2-XXX) are bound.

[0074] In this step, the processor 1001 queries various parameter data under the same time period index identifier based on the index-time period mapping table, performs precise matching according to the time period range, eliminates data with non-overlapping time ranges, integrates the remaining data into parameter data groups for a single time period, and marks the corresponding time period index identifier to ensure the time period uniqueness of the data group.

[0075] S233. Associate the parameter data group with the number of suction failures and the suction action arrival time of the nozzle statistically recorded in the same working cycle to form a target operation data entry containing complete time period parameters and operation status data. It should be noted that the number of suction failures and the time it takes for the suction nozzle to complete its action are the core data reflecting the operating status of the suction nozzle. The core purpose of associating them with the parameter data group is to integrate the time period parameter data with the operating status data to form a complete target operating data entry, providing comprehensive data support for model calculation that includes "parameter-status".

[0076] The number of suction failures and the time it takes for the nozzle to complete its action within the same work cycle are associated with all parameter data sets within that cycle. In other words, each work cycle corresponds to a complete target operation data entry, which includes parameter data sets for all time periods within that cycle and the corresponding operation status data.

[0077] In this step, the processor 1001 uses the work cycle number as the association basis to integrate all single-time period parameter data groups within the same work cycle with the number of suction failures and suction action arrival time of that cycle, forming a structured target operation data entry. The entry format is: {Work cycle number: N, Time period index identifier set: {N-01 to N-05}, Parameter data group set: {Data-N-01 to Data-N-05}, Number of suction failures: FailCount-N, Action arrival time: ArriveTime-N}, ensuring the integrity and standardization of the data entry.

[0078] S234. Multiple target operation data entries are systematically summarized according to the time period index identifier to form the target operation data set, and each target operation data entry corresponds to a single working cycle of the nozzle.

[0079] It should be noted that the core purpose of ordered aggregation is to form a structured and traceable target operation data set, ensuring the logic and availability of the data set, making it easy for the model to call data in the order of the work cycle, and supporting queries by key fields such as work cycle number and time period index identifier.

[0080] The storage capacity limit of the target running data set is set by the user (default 10,000 entries). When the number of data entries reaches the limit, the processor 1001 automatically deletes the oldest 1,000 data entries (or executes according to the storage policy set by the user) to ensure the rational use of storage resources.

[0081] In this step, the processor 1001 arranges all valid target running data entries in ascending order according to the working cycle number, forming a target running data set, which is stored in the extended storage area of ​​the memory 1005. It supports fast querying by key fields such as working cycle number, time period index identifier, and number of failures, providing convenience for subsequent model calculations and data traceability.

[0082] In one embodiment, the diagnostic prediction model is constructed through the following steps: S310. Perform feature classification on the target operation data to determine numerical feature data and categorical feature data; the numerical feature data includes the vacuum negative pressure value parameter, the air pressure parameter, the gas flow rate parameter, and the nozzle action arrival time; the categorical feature data includes the nozzle damage indicator. It should be noted that the core purpose of feature classification is to improve data processing efficiency and model calculation accuracy by adopting differentiated preprocessing methods based on different data types. Numerical feature data refers to features that can be expressed through specific numerical values. Its core characteristics are continuity and computability, and it can be directly used for numerical calculations in the model. Categorical feature data refers to features that cannot be quantified numerically and can only be described by categories. Its core characteristic is discreteness, and it needs to be encoded and converted before it can be used for model calculations.

[0083] The characteristics of vacuum negative pressure parameters include the average negative pressure at the start of adsorption, the fluctuation range of negative pressure when adsorption is complete, and the minimum negative pressure for transmission and adhesion; the characteristics of air pressure parameters include the average air pressure and fluctuation range during the opening / closing of the nozzle; the characteristics of gas flow parameters include the average flow rate and fluctuation range during the opening / closing of the nozzle; the nozzle action completion time is a single numerical characteristic; the nozzle damage indicator is determined based on image data collected by the vision module, and is classified into three categories: "normal", "minor damage", and "severe damage", which directly reflects the category characteristics of the physical state of the nozzle.

[0084] In this step, the processor 1001 extracts features from each data entry in the target running dataset. According to the above definition, the extracted features are divided into numerical feature data and categorical feature data, and stored in different data buffers to prepare for subsequent preprocessing.

[0085] S320. Perform outlier processing and standardization on the numerical feature data to obtain the first feature data; It should be noted that the core purpose of outlier handling is to remove extreme outliers in numerical feature data to avoid them interfering with model training results; the core purpose of standardization is to unify the dimensions of different numerical feature data, eliminate model weight bias caused by differences in dimensions, and ensure that the impact of each feature on the model is fair and reasonable.

[0086] Outlier handling employs box plotting, whereby for each numerical feature, its quartiles (Q1, Q2, Q3) and interquartile range (IQR = Q3 - Q1) are calculated. Data values ​​less than Q1 - 1.5 × IQR or greater than Q3 + 1.5 × IQR are identified as outliers and processed using median replacement (i.e., the median of the feature data is used to replace the outliers). Standardization uses Z-score standardization, with the formula: $x' = \frac{x - \mu}{\sigma}$, where $x$ is the original data, $\mu$ is the mean of the feature data, and $\sigma$ is the standard deviation of the feature data. The standardized data has a mean of 0 and a standard deviation of 1.

[0087] In this step, the processor 1001 performs outlier processing and standardization on all numerical feature data in sequence according to the above method. The result after processing is the first feature data, which is stored in the model training data area.

[0088] S330. Extract features from the image data acquired by the vision module to obtain the wear area and deformation of the nozzle; encode and convert the categorical feature data to obtain the second feature data; It should be noted that the core purpose of visual image data feature extraction is to convert the visual information of the image into quantifiable numerical features, providing the model with a direct quantitative basis for the physical state of the nozzle; the core purpose of categorical feature data encoding conversion is to convert discrete category labels into numerical vectors that the model can recognize, ensuring that category features can participate in model calculations.

[0089] Image data feature extraction employs machine vision algorithms (such as edge detection, threshold segmentation, and morphological processing): First, the original image is processed by grayscale conversion, noise reduction (Gaussian filtering), and enhancement (histogram equalization) to improve image quality; then, the nozzle head contour is extracted using an edge detection algorithm (Canny algorithm), compared with a standard nozzle contour, and the area of ​​the difference region is calculated, which is the nozzle wear area; finally, a coordinate system is established with the center point of the standard contour as the reference, the coordinate deviation between the key feature points of the actual contour and the standard point is measured, and the maximum deviation value is taken as the nozzle deformation.

[0090] The categorical feature data encoding transformation adopts the one-hot encoding method. The encoding rules are: "normal" corresponds to vector [1,0,0], "slightly damaged" corresponds to vector [0,1,0], and "severely damaged" corresponds to vector [0,0,1]. The transformed numerical vector is the second feature data.

[0091] In this step, the processor 1001 calls the image processing algorithm to extract features from the image data acquired by the vision module to obtain the wear area and deformation of the nozzle; at the same time, it performs one-hot encoding conversion on the categorical feature data to obtain the second feature data, and stores the wear area and deformation of the nozzle together with the second feature data in the model training data area.

[0092] S340. Summarize the first feature data, the wear area of ​​the suction nozzle, the deformation amount, and the second feature data, and combine them with the statistical characteristics of the number of suction failures of the suction nozzle to construct a multi-dimensional feature vector; It should be noted that the statistical features include the cumulative number of nozzle failures, the number of consecutive failures, and the failure rate (the ratio of the number of failures to the total number of work cycles). Their core function is to reflect the historical trend of nozzle adsorption reliability and provide time-series feature support for the model.

[0093] The core function of multi-dimensional feature vectors is to integrate feature data from different sources and of different types into a unified input format, which facilitates feature matching and computation by the model and ensures that the model can fully utilize various types of data information for fault diagnosis and life prediction.

[0094] In this step, the processor 1001 first determines the dimensions and order of the feature vector. The feature vector contains 18 dimensions in the following order: average negative pressure at the start of adsorption, fluctuation range of negative pressure at the start of adsorption, average negative pressure at the end of adsorption, fluctuation range of negative pressure at the end of adsorption, average negative pressure during transmission and attachment, fluctuation range of negative pressure during transmission and attachment, average air pressure during the nozzle opening period, fluctuation range of air pressure during the nozzle opening period, average air pressure during the nozzle closing period, fluctuation range of air pressure during the nozzle closing period, average flow rate during the nozzle opening period, fluctuation range of flow rate during the nozzle opening period, average flow rate during the nozzle closing period, fluctuation range of flow rate during the nozzle closing period, nozzle action arrival time, nozzle wear area, nozzle deformation, nozzle damage identification encoding vector (3 dimensions), number of nozzle suction failures, and number of consecutive failures. Subsequently, for each target running data item, the corresponding data is extracted according to the above-mentioned dimension order to form an 18-dimensional feature vector, which is then stored in the model training dataset.

[0095] S350. Based on the multi-dimensional feature vector, a basic model is trained using a preset algorithm. A derivative model is generated by adjusting the negative pressure threshold, flow rate threshold, and air pressure threshold. The basic model and the derivative model are integrated to obtain the diagnostic prediction model.

[0096] It should be noted that the default algorithm is the random forest algorithm, which has the advantages of anti-overfitting, strong ability to handle high-dimensional data, and good robustness, making it suitable for modeling multi-dimensional feature data. The base model is the core model trained based on the original threshold parameters, and the derived model is the auxiliary model generated by adjusting the key threshold parameters. The core purpose of model integration is to combine the advantages of multiple models to improve the accuracy and generalization ability of diagnostic prediction.

[0097] Negative pressure threshold, flow rate threshold, and air pressure threshold are key parameters for the model to determine fault and lifespan status. The adjustment range is based on the normal state parameter range of historical data statistics, and is adjusted within ±10% and ±20% respectively, generating 5 different threshold combinations, corresponding to 5 derivative models.

[0098] In this step, processor 1001 first divides the multi-dimensional feature vectors into training and testing sets in a 7:3 ratio; then, it trains the base model using the random forest algorithm, setting parameters such as the number of decision trees to 100 and the maximum depth to 10, with fault type and consumable life status as target labels to minimize prediction error; next, it trains 5 derived models based on 5 sets of adjusted threshold combinations; finally, it integrates the base model and the 5 derived models using a weighted voting method, assigning weights to each model according to its prediction accuracy on the test set (the higher the accuracy, the greater the weight), and the integrated model is the diagnostic prediction model, which is stored in the model storage area of ​​memory 1005.

[0099] In one embodiment, the method further includes: S360. Based on the fault diagnosis results and consumable life prediction information output by the diagnostic prediction model, the actual operating results and consumable replacement records are compared and correlated to obtain the prediction error of the diagnostic prediction model. It should be noted that actual operating results refer to the actual equipment status information, such as fault types and fault handling records confirmed by manual inspection; consumable replacement records refer to information such as the actual replacement time of the nozzle, its wear condition at the time of replacement, and remaining life assessment. The core purpose of correlation comparison is to quantify the predictive accuracy of the diagnostic prediction model and provide a basis for model optimization and adjustment.

[0100] Prediction error is measured across multiple dimensions: Failure Type Prediction Error Rate (number of incorrectly predicted failure types / total number of failure cases × 100%), Mean Failure Period Prediction Deviation (average of the absolute values ​​of the prediction deviations for the failure periods across all failure cases), Lifespan Status Prediction Error Rate (number of incorrectly predicted lifespan statuses / total number of lifespan assessment cases × 100%), and Mean Remaining Life Percentage Deviation (average of the absolute values ​​of the deviations for the remaining life percentage across all lifespan assessment cases). These metrics reflect the model's predictive performance from different perspectives, ensuring the comprehensiveness of error assessment.

[0101] In this step, the processor 1001 associates the predicted results output by the model with the actual operating results and consumable replacement records one by one according to the working cycle number, establishes a comparison data table, calculates the prediction error based on the above indicator calculation formula, and stores it in the error analysis data area.

[0102] S370. Based on the prediction error, and combined with the feature weights of the vacuum negative pressure value parameter, the air pressure parameter, the gas flow rate parameter, the nozzle action arrival time, the nozzle wear area, the deformation amount, the nozzle damage mark, and the number of nozzle suction failures contained in the multi-dimensional feature vector, the feature weights of the diagnostic prediction model and the negative pressure threshold, the flow rate threshold, and the air pressure threshold are adjusted. It should be noted that feature weights reflect the contribution of each feature to the model's prediction results and are calculated using the feature importance evaluation method of the random forest algorithm. The core purpose of adjusting feature weights and various thresholds is to optimize model parameters and improve model prediction accuracy for dimensions with large prediction errors.

[0103] For example, if the fault type prediction error rate is too high, and analysis reveals that the feature importance of the nozzle wear area is lower than the preset value (e.g., 5%), leading to misjudgments of nozzle-related faults, then the weight of this feature should be appropriately increased. If the lifespan state prediction error rate is too high, and analysis reveals that the negative pressure threshold is set too strictly, leading to frequent false lifespan warnings, then the range of the negative pressure threshold should be appropriately widened. During parameter adjustment, only one parameter should be adjusted at a time, and the adjustment range should not exceed 20% of the original parameter value to avoid model performance degradation due to sudden parameter changes.

[0104] In this step, the processor 1001 first analyzes the sources of prediction error, identifies the dimensions with larger errors and their corresponding influencing features and threshold parameters; then, based on feature weight influence analysis, it adjusts the weights of the corresponding features and related thresholds; finally, it stores the adjusted parameters in the model parameter register to prepare for model retraining.

[0105] S380. Supplement the model training set with case data whose prediction error exceeds the preset range, and retrain and optimize the diagnostic prediction model based on the adjusted feature weights, negative pressure threshold, flow rate threshold, and air pressure threshold using the preset algorithm.

[0106] It should be noted that the preset error range is set based on the model design requirements. Specifically, the threshold for the error rate of fault type prediction is 5%, the threshold for the average deviation of fault time prediction is 100ms, the threshold for the error rate of lifetime state prediction is 8%, and the threshold for the average deviation of remaining lifetime percentage is 10%. Case data exceeding any of these error indicators are considered to have exceeded the error limit. The core purpose of supplementing the training set with this type of data is to enhance the model's adaptability to complex scenarios and abnormal operating conditions, and to improve the model's generalization ability.

[0107] The core purpose of retraining is to update the model's decision rules based on the optimized parameters and the supplemented training set, so that the model can correct previous prediction biases and continuously improve prediction accuracy.

[0108] In this step, processor 1001 first filters out case data with excessive errors and adds them to the model training set to ensure the diversity and representativeness of the training set cases. Then, based on the adjusted feature weights, negative pressure threshold, flow rate threshold, and air pressure threshold, the same random forest algorithm and parameters as the original model are used to retrain the model using the supplemented training set. During the training process, the prediction error index of the model on the test set is monitored in real time. When the error index drops below the preset threshold, training is stopped. Finally, the retrained model replaces the original diagnostic prediction model and is stored in the model storage area of ​​memory 1005, completing the model optimization and adjustment.

[0109] In addition, please refer to Figure 6 This invention also proposes a fault diagnosis device, comprising: a data acquisition module for real-time acquisition of multi-dimensional operational data during the operation of a nozzle-attach pneumatic module; a data processing module for classifying the acquired multi-dimensional operational data according to the nozzle's working period to obtain target operational data; a model module for storing a diagnostic prediction model, receiving the target operational data and performing calculations to identify the pneumatic module fault type and nozzle consumable life status; and an output module for outputting the fault diagnosis results and consumable life prediction information obtained by the model module.

[0110] It should be noted that the data acquisition module is the core unit of the device for acquiring external data, including sensor interface circuits, signal conversion circuits, and data buffer circuits. The sensor interface circuit provides various standard sensor interfaces (analog interface, digital interface, Ethernet interface, etc.), which are compatible with various types of sensors; the signal conversion circuit converts the sensor output signal into a standard digital signal that the device can recognize, with a conversion delay of less than 1ms; the data buffer circuit uses a high-speed RAM chip to temporarily store the acquired raw data and prevent data loss.

[0111] The data processing module is the core computing unit of the device, employing a 32-bit high-performance microprocessor (such as the STM32H7 series MCU) with a main frequency of up to 480MHz, possessing powerful data processing capabilities. Its main functions include data classification and processing, removal of invalid and redundant data, multi-dimensional data association, and data preprocessing, providing standardized input data for the model module.

[0112] The model module is the core unit for the device to realize fault diagnosis and life prediction, including a model storage circuit, a model calculation circuit, and a model optimization circuit. The model storage circuit uses non-volatile memory to store the diagnostic prediction model and related parameters; the model calculation circuit uses an FPGA chip, which has the advantage of strong parallel computing capabilities and a calculation latency of less than 50ms; the model optimization circuit works in conjunction with the data processing module and the output module to realize dynamic adjustment and retraining of model parameters.

[0113] The output module is the unit for interaction between the device and external systems, including a communication interface circuit, a display driver circuit, and an alarm driver circuit. The communication interface circuit provides multiple standard communication interfaces, supporting communication with external devices such as touch screens and system backends; the display driver circuit drives the touch screen to display diagnostic results and predictive information; and the alarm driver circuit triggers audible and visual alarms to alert users of serious abnormalities.

[0114] This fault diagnosis device, through its modular design, integrates data acquisition, processing, model calculation, and result output. It features a compact structure, strong compatibility, and can be adapted to different types of nozzle-attach pneumatic module equipment. It is characterized by accurate data acquisition, efficient processing, accurate prediction, and intuitive output.

[0115] In one embodiment, the fault diagnosis device further includes a quality feedback module, used to collect actual operating results and consumable replacement records, compare them with model output results, calculate prediction errors, and feed them back to the model module, providing data support for model optimization and adjustment. The quality feedback module is connected to an external system via a communication interface circuit, automatically collecting actual operating data to ensure the closed-loop nature of model optimization.

[0116] Since this device adopts all the technical solutions of all the above embodiments, it has at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be repeated here.

[0117] Furthermore, embodiments of the present invention also propose a fault diagnosis device, the device including a memory, a processor, and a fault diagnosis program stored in the memory and executable on the processor, the fault diagnosis program being configured to implement the steps of the fault diagnosis method as described in any of the preceding claims.

[0118] It should be noted that the hardware structure of this fault diagnosis equipment includes a memory, processor, communication interface, input devices, and output devices, all connected via a motherboard bus. The memory includes RAM and ROM; RAM stores temporary data, while ROM stores the operating system, fault diagnosis programs, etc. The processor uses an industrial-grade CPU (such as an Intel Core i5-1035G1) with high-efficiency computing power. The communication interface includes wired and wireless communication interfaces, supporting communication with various peripherals and external systems. Input devices include a touchscreen and physical buttons for receiving user operation commands. Output devices include a display screen, an audible and visual alarm, and a printer interface for outputting diagnostic results, triggering alarms, and printing reports.

[0119] The device's software operating environment is an industrial-grade operating system (such as Windows 10 IoT Enterprise). The fault diagnosis program is developed based on the C++ language, adopts a modular design, and relies on open-source libraries such as OpenCV and Scikit-learn to implement image processing and model calculation.

[0120] This fault diagnosis equipment, through the collaborative design of hardware and software, features comprehensive data acquisition, efficient processing, accurate prediction, convenient operation, and strong environmental adaptability. It can be widely used in nozzle attachment pneumatic module equipment in the intelligent manufacturing industry, providing comprehensive support for equipment maintenance and consumable management.

[0121] Furthermore, embodiments of the present invention also propose a storage medium storing a fault diagnosis program, wherein the fault diagnosis program, when executed by a processor, implements the steps of the fault diagnosis method as described in any of the preceding claims.

[0122] It should be noted that the storage medium can be a computer-readable storage medium such as a solid-state drive, SD card, USB flash drive, or hard disk drive, all of which are non-volatile to ensure secure data storage. The fault diagnosis program is stored in binary file format on the storage medium and contains code for functional modules such as data acquisition, processing, model calculation, result output, and model optimization.

[0123] Once the storage medium is connected to the fault diagnosis equipment, the equipment's processor reads the fault diagnosis program from the storage medium, loads it into the memory, and executes it to implement the various steps of the fault diagnosis method. The storage medium supports multiple reads and writes, has a long service life, and can meet the long-term use needs of industrial production sites.

[0124] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0125] In addition, for technical details not described in detail in this embodiment, please refer to the fault diagnosis method provided in any embodiment of the present invention, which will not be repeated here.

[0126] Furthermore, it should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0127] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0128] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0129] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A fault diagnosis method, characterized in that, The method includes: Real-time acquisition of multi-dimensional operational data during the operation of nozzle-attach pneumatic modules; The multi-dimensional operational data is categorized and processed according to the working period of the suction nozzle to obtain the target operational data; The target operating data is input into the diagnostic prediction model, which identifies the pneumatic module fault type and nozzle consumable life status, and outputs fault diagnosis results and consumable life prediction information.

2. The method as described in claim 1, characterized in that, The multi-dimensional operational data includes air pressure parameters, gas flow parameters, vacuum negative pressure value parameters, number of suction failures, and suction nozzle arrival time; the steps for real-time acquisition of multi-dimensional operational data during the operation of the suction nozzle attachment pneumatic module include: The vacuum negative pressure value parameters during the adsorption process of the suction nozzle are collected by a negative pressure detection sensor, and the data are divided into adsorption start period, adsorption completion period and adsorption transfer and adhesion period according to the adsorption stage. The air pressure and gas flow parameters of the pneumatic module are collected by the air pressure detection sensor and the gas flow sensor respectively throughout the operation process, and the data are divided into the suction nozzle opening period and the suction nozzle closing period according to the suction nozzle status. The pneumatic module control system counts the number of suction failures and the time it takes for the suction nozzle to complete its action within a single work cycle. Image data of the nozzle head is collected by a vision module to analyze the wear and abnormal condition of the nozzle.

3. The method as described in claim 2, characterized in that, The step of classifying the multi-dimensional operational data according to the working period of the suction nozzle to obtain the target operational data includes: Remove invalid and redundant data from the data during the nozzle opening period and the data during the nozzle closing period; Based on a single work cycle, the number of times the nozzle failed to pick up the suction, the time when the nozzle action was in place, the negative pressure data at the start of suction, the negative pressure data at the end of suction, the negative pressure data for suction transmission and attachment, the dataset of the nozzle opening period after removing invalid and redundant data, and the dataset of the nozzle closing period are associated to form a target operation data set indexed by the work period.

4. The method as described in claim 3, characterized in that, Set data validity judgment thresholds, which include pressure parameter fluctuation thresholds, gas flow parameter stability thresholds, and data acquisition time interval thresholds. The step of removing invalid and redundant data from the data during the nozzle-open period and the data during the nozzle-closed period includes: If the fluctuation range of the air pressure parameter or the gas flow rate parameter exceeds the fluctuation threshold of the air pressure parameter or the stability threshold of the gas flow rate parameter within the same time period, it is determined to be invalid data and is removed. If the repetition rate of the gas pressure parameter or the gas flow parameter collected within the same time period exceeds a preset ratio, or the collection time interval exceeds the data collection time interval threshold, it is determined to be redundant data and is removed.

5. The method as described in claim 3, characterized in that, The associated steps include: Using the single working cycle of the suction nozzle as the time reference, a unique time period index is established. The time period index corresponds one-to-one with the suction start time period, the suction arrival time period, the transfer and attachment time period, the suction nozzle opening time period, and the suction nozzle closing time period. The adsorption start negative pressure data, adsorption completion negative pressure data, or transmission attachment negative pressure data under the same time period index are bound to the corresponding gas pressure parameter dataset and gas flow parameter dataset for the corresponding nozzle opening period or nozzle closing period to form a parameter data group for a single time period. The parameter data set is associated with the number of suction failures and the suction action arrival time of the nozzle statistically recorded within the same work cycle to form a target operation data entry containing complete time period parameters and operation status data. Multiple target operation data entries are sequentially aggregated according to the time period index to form the target operation data set, and each target operation data entry corresponds to a single working cycle of the nozzle.

6. The method as described in claim 2, characterized in that, The diagnostic prediction model is constructed through the following steps: The target operating data is classified into numerical feature data and categorical feature data. The numerical feature data includes the vacuum negative pressure value parameter, the air pressure parameter, the gas flow rate parameter, and the nozzle action arrival time. The categorical feature data includes the nozzle damage indicator. The numerical feature data is subjected to outlier processing and standardization to obtain the first feature data; Feature extraction is performed on the image data acquired by the vision module to obtain the wear area and deformation of the nozzle. The categorical feature data is encoded and converted to obtain the second feature data; By summarizing the first feature data, the wear area of ​​the suction nozzle, the deformation amount, and the second feature data, and combining them with the statistical characteristics of the number of suction failures of the suction nozzle, a multi-dimensional feature vector is constructed. Based on the multi-dimensional feature vectors, a basic model is trained using a preset algorithm. A derivative model is generated by adjusting the negative pressure threshold, flow rate threshold, and air pressure threshold. The basic model and the derivative model are then integrated to obtain the diagnostic prediction model.

7. The method as described in claim 6, characterized in that, The method further includes: Based on the fault diagnosis results and consumable life prediction information output by the diagnostic prediction model, the actual operating results and consumable replacement records are compared and correlated to obtain the prediction error of the diagnostic prediction model. Based on the prediction error, and combined with the feature weights of the vacuum negative pressure value parameter, the air pressure parameter, the gas flow rate parameter, the nozzle action arrival time, the nozzle wear area, the deformation amount, the nozzle damage mark, and the number of nozzle suction failures contained in the multi-dimensional feature vector, the feature weights of the diagnostic prediction model and the negative pressure threshold, the flow rate threshold, and the air pressure threshold are adjusted. Case data whose prediction error exceeds the preset range are added to the model training set. Based on the adjusted feature weights, negative pressure threshold, flow rate threshold, and air pressure threshold, the diagnostic prediction model is retrained and optimized using the preset algorithm.

8. A fault diagnosis device, characterized in that, The device includes: The data acquisition module is used to collect multi-dimensional operational data in real time during the operation of the nozzle attachment pneumatic module; The data processing module is used to classify and process the collected multi-dimensional operational data according to the working period of the suction nozzle to obtain the target operational data; The model module is used to store diagnostic prediction models, receive the target operating data and perform calculations to identify the fault type of the pneumatic module and the life status of the nozzle consumables. The output module is used to output the fault diagnosis results and consumable life prediction information obtained by the model module.

9. A fault diagnosis device, characterized in that, The device includes: a memory, a processor, and a fault diagnosis program stored in the memory and executable on the processor, the fault diagnosis program being configured to perform the steps of fault diagnosis as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a fault diagnosis program, which, when executed by a processor, implements the fault diagnosis steps as described in any one of claims 1 to 7.

Images