A durable multi-parameter collaborative testing system for diaphragm valves

Abstract

This invention provides a multi-parameter collaborative testing system for the durability of diaphragm valves, relating to the field of valve durability testing technology. It includes: a drive positioning unit for adjusting the physical gap between a high-speed solenoid valve and the drive end of the diaphragm valve under test according to preset assembly parameters, obtaining a positioning reference signal containing real-time drive distance and alignment status; a cyclic excitation unit for generating periodic control commands based on the positioning reference signal, obtaining an alternating opening and closing action sequence acting on the diaphragm valve under test; and a multi-parameter acquisition unit for receiving the alternating opening and closing action sequence, synchronously monitoring the valve response time, pipeline pressure fluctuation, helium leakage rate, and particulate matter concentration triggered by this sequence, forming a raw test dataset containing a multi-parameter time series. This invention effectively solves the defect in existing diaphragm valve durability testing that requires interruption of the cycle and disassembly of the valve for performance testing, further improving the authenticity and repeatability of the test results.

Description

Technical Field

[0001] This invention relates to the field of valve durability testing technology, and in particular to a multi-parameter collaborative testing system for the durability of diaphragm valves. Background Technology

[0002] Atomic layer deposition (ALD) technology, with its atomic-level thickness control and excellent film uniformity, has become a key thin film fabrication process in advanced semiconductor manufacturing. As a core actuator in this process, the ALD diaphragm valve needs to achieve microsecond-level response and ultra-high sealing performance in high-temperature, highly corrosive gas environments. Its durability and reliability directly determine process yield and equipment uptime.

[0003] Currently, durability verification of ALD diaphragm valves typically employs a discrete testing approach: the valve is installed on a dedicated test bench, run a certain number of cycles, then disassembled and transferred to different equipment such as a leak detection bench, particulate counter, and flow characteristic test bench for individual parameter measurements, before being reinstalled and the cycle continues. However, this testing mode generally suffers from the following technical drawbacks: the testing process must be interrupted and the valve disassembled, making it difficult to complete durability cycling and performance monitoring in the same continuous process. This drawback leads to a series of related problems: for example, assembly differences and environmental interference introduced during disassembly and reassembly result in a large amount of missing intermediate data between two measurements, making it impossible to accurately determine at which cycle number the valve begins to fail; furthermore, each disassembly requires re-calibrating the equipment and establishing test conditions, which is not only time-consuming and labor-intensive but also causes additional wear on the valve sealing interface due to repeated disassembly and reassembly, further interfering with the accurate determination of the cause of failure. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a multi-parameter collaborative testing system for the durability of diaphragm valves, which can solve the defect that the existing diaphragm valve durability test must interrupt the cycle and disassemble the valve to perform performance testing, and further increase the authenticity and repeatability of the test results.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a multi-parameter collaborative testing system for the durability of a diaphragm valve includes: The drive positioning unit is used to adjust the physical gap between the high-speed solenoid valve and the drive end of the diaphragm valve under test according to the preset assembly parameters, so as to obtain a positioning reference signal containing the real-time drive distance and the centering status. The cyclic excitation unit is used to generate periodic control commands based on the positioning reference signal to obtain an alternating opening and closing action sequence acting on the diaphragm valve under test. The multi-parameter acquisition unit is used to receive the alternating opening and closing action sequence, and simultaneously monitor the valve response time, pipeline pressure fluctuation, helium leakage rate and particulate matter concentration triggered by this sequence, forming a raw test dataset containing multi-parameter time series. The data fusion processing unit is used to analyze the temporal dimension and physical structure mapping dimension of the original test dataset, dynamically lock the sampling features of the multidimensional parameter evolution trajectory, and fit and construct a constrained ellipsoid space representing the multidimensional performance degradation state of the valve based on the sampling features of the multidimensional parameter evolution trajectory. Principal component tensor projection dimensionality reduction is performed on the constrained ellipsoid space to extract the gradient drift and dissipation volume expansion rate of the tensor trace of the degradation sensitive direction. The gradient drift and dissipation volume expansion rate are input into a preset multidimensional decay correlation model to obtain the feature parameter matrix representing the performance degradation trajectory of the diaphragm valve. The durability determination unit is used to match and analyze the feature parameter matrix with a preset durability failure threshold to obtain a durability evaluation result that includes life prediction results, abnormal alarm instructions and comprehensive test reports.

[0006] In a second aspect, a computing device includes: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to execute the system.

[0007] Thirdly, a computer-readable storage medium storing a program that, when executed by a processor, performs the system.

[0008] The above-described solution of the present invention has at least the following beneficial effects: Because the driving positioning unit precisely adjusts the physical gap between the high-speed solenoid valve and the driven end of the diaphragm valve under test through closed-loop control, and outputs a positioning reference signal to the cyclic excitation unit, the cyclic excitation unit generates an alternating opening and closing action sequence based on this signal and transmits it to the multi-parameter acquisition unit. The multi-parameter acquisition unit simultaneously acquires multi-parameter time series such as response time, pressure fluctuation, leakage rate, and particulate matter concentration, and then hands them over to the data fusion processing unit. The data fusion processing unit locates the performance degradation characteristic moments using the second derivative, constructs a constrained ellipsoidal space characterizing the multi-dimensional performance degradation state of the valve, and performs principal component tensor projection dimensionality reduction to extract the gradient drift and dissipative volume of the degradation-sensitive direction tensor trace. The product expansion rate is input into a preset multidimensional attenuation correlation model to obtain a feature parameter matrix. Finally, the durability judgment unit matches and analyzes the feature parameter matrix with a preset failure threshold and outputs the life prediction result and alarm command. Therefore, it realizes the seamless connection of data flow between units and multi-parameter collaborative processing. It overcomes the technical defects of the existing technology, which requires interruption of testing and repeated disassembly and reassembly of valves to perform performance testing, resulting in discontinuous monitoring and difficulty in accurately locating failure nodes. Thus, it achieves the technical effect of completing full-cycle multi-parameter synchronous monitoring, accurately locating the failure start time and acceleration inflection point, and significantly improving the authenticity and repeatability of the test without disassembling the valve. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of a multi-parameter collaborative testing system for the durability of a diaphragm valve provided in an embodiment of the present invention; Figure 2 This is a flowchart illustrating a multi-parameter collaborative testing system for the durability of a diaphragm valve, provided by an embodiment of the present invention. The system generates periodic control commands based on the positioning reference signal to obtain an alternating opening and closing sequence acting on the diaphragm valve under test. Figure 3 This is a three-dimensional structural simulation diagram of the constrained ellipsoidal space, which characterizes the multidimensional performance degradation state of the valve, provided by the experimental example of the present invention. Figure 4 This is a trajectory diagram of the shrinkage evolution of the ellipsoid volume with the number of cycles provided in the experimental example of the present invention; Figure 5 This is a simulation diagram of the decay trajectory in a three-dimensional feature space of the multidimensional decay correlation model provided in the experimental example of the present invention. Detailed Implementation

[0010] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0011] like Figure 1 As shown, an embodiment of the present invention proposes a multi-parameter collaborative testing system for the durability of a diaphragm valve, comprising: The drive positioning unit is used to adjust the physical gap between the high-speed solenoid valve and the drive end of the diaphragm valve under test according to the preset assembly parameters, so as to obtain a positioning reference signal containing the real-time drive distance and the centering status. The cyclic excitation unit is used to generate periodic control commands based on the positioning reference signal to obtain an alternating opening and closing action sequence acting on the diaphragm valve under test. The multi-parameter acquisition unit is used to receive the alternating opening and closing action sequence, and simultaneously monitor the valve response time, pipeline pressure fluctuation, helium leakage rate and particulate matter concentration triggered by this sequence, forming a raw test dataset containing multi-parameter time series. The data fusion processing unit is used to analyze the temporal dimension and physical structure mapping dimension of the original test dataset, dynamically lock the sampling features of the multidimensional parameter evolution trajectory, and fit and construct a constrained ellipsoid space representing the multidimensional performance degradation state of the valve based on the sampling features of the multidimensional parameter evolution trajectory. Principal component tensor projection dimensionality reduction is performed on the constrained ellipsoid space to extract the gradient drift and dissipation volume expansion rate of the tensor trace of the degradation sensitive direction. The gradient drift and dissipation volume expansion rate are input into a preset multidimensional decay correlation model to obtain the feature parameter matrix representing the performance degradation trajectory of the diaphragm valve. The durability determination unit is used to match and analyze the feature parameter matrix with a preset durability failure threshold to obtain a durability evaluation result that includes life prediction results, abnormal alarm instructions and comprehensive test reports.

[0012] In this embodiment of the invention, the precise closed-loop adjustment and stable periodic excitation of the driving distance between the high-speed solenoid valve and the tested diaphragm valve are achieved through the cooperation of the driving positioning unit and the cyclic excitation unit. The multi-parameter acquisition unit simultaneously acquires time series of multiple parameters such as response time, pressure fluctuation, leakage rate, and particulate matter concentration during the valve durability cycle without interrupting the test or disassembling the valve. The data fusion processing unit uses the second derivative to locate the performance degradation characteristic moment, constructs a constrained ellipsoidal space characterizing the multidimensional performance degradation state of the valve, and extracts gradient drift and dissipation volume expansion rate through principal component tensor projection dimensionality reduction, thereby accurately depicting the valve performance degradation trajectory. The durability judgment unit matches and analyzes the feature parameter matrix with the failure threshold, and automatically outputs life prediction, graded alarms, and comprehensive reports. The entire collaborative testing system realizes continuous monitoring of the entire cycle, multiple parameters, and without valve disassembly, effectively improving the integrity of test data, the accuracy of failure judgment, and the repeatability of evaluation results.

[0013] In a preferred embodiment of the present invention, the physical gap between the high-speed solenoid valve and the drive end of the diaphragm valve under test is adjusted according to preset assembly parameters to obtain a positioning reference signal containing real-time drive distance and alignment status, including: The process involves reading preset assembly parameters, parsing the target gap value and alignment deviation tolerance from them to obtain the gap adjustment target. Specifically, this includes: first, reading a pre-configured assembly parameter file in a structured format, such as XML or JSON, which contains fields such as the target gap value, allowable error range, and alignment deviation tolerance; then, further parsing the numerical fields in this file to extract the target gap value. (e.g., 0.05 mm), radial offset tolerance (e.g., 0.01 mm) and angular deviation tolerance (e.g., 0.5 degrees); The three extracted parameters are combined into a structured gap adjustment target object according to the internal data protocol: This gap adjustment target object contains parameter type identifier, value and unit, and is passed to the subsequent closed-loop control module as the basis for execution.

[0014] Based on the gap adjustment target, a precision linear sliding stage is driven to move a high-speed solenoid valve. Simultaneously, the current physical gap value is acquired in real time, and closed-loop control is implemented until the current gap value falls within the error range of the target gap value. Then, the sliding stage position is locked, and a preliminary positioning completion signal is obtained. Specifically, this includes: based on the target gap value in the gap adjustment target... and error range The system sends a command to the servo driver of the precision linear slide stage, specifying the direction of movement (positive for moving closer to the drive end, negative for moving away) and the step size (e.g., 0.001 mm). The servo driver then drives the slide stage to move the high-speed solenoid valve axially. Simultaneously, a displacement sensor mounted on the side of the slide stage, such as a linear encoder or laser displacement sensor, continuously acquires the current physical gap value with a sampling period of 1 ms. The deviation is fed back to the closed-loop controller in real time; the closed-loop controller internally uses a proportional-integral-derivative algorithm to first calculate the deviation. Based on this, the moving speed and direction of the sliding table are dynamically adjusted according to the magnitude and rate of change of the deviation: when the deviation is large, it moves quickly; when the deviation is close to zero, it decelerates to fine adjustment. This closed-loop adjustment is repeated until the current gap value is reached. gap value with target absolute deviation Less than the preset error range ,For example Millimeters; at this time, the controller sends a locking signal to the electromagnetic brake of the sliding table, the brake moves to fix the position of the sliding table, and generates a preliminary positioning completion signal. This preliminary positioning completion signal includes the final stable physical gap value and the locking status flag, which is used to indicate that the physical gap has been precisely adjusted to the position.

[0015] After receiving the preliminary positioning completion signal, the alignment detection module is activated to measure the radial offset and angular deviation between the centerline of the high-speed solenoid valve and the centerline of the drive end of the diaphragm valve under test, obtaining alignment status data. Specifically, this includes: automatically activating the alignment detection module after receiving the preliminary positioning completion signal; the alignment detection module first controls two laser displacement sensors to measure two different cross-sections on the centerline of the high-speed solenoid valve, which are then recorded as cross-sections. and cross section Coordinate measurements are performed, and simultaneously, two cross-sections corresponding to the axis of the driven end of the diaphragm valve under test are measured. and cross section Perform coordinate measurements; set the cross section. The coordinates of the solenoid valve axis are The axis coordinates of the diaphragm valve drive end are ;section The coordinates of the locations are respectively and All coordinate values ​​are in millimeters and based on the same global coordinate system; based on this, the radial offset between the two axis lines is calculated. and angular deviation The formula is as follows:

[0016] in, The direction vector of the solenoid valve's axis of rotation. The direction vector of the axis of the diaphragm valve drive end. , Represents the magnitude of the vector; the calculated radial offset. (Unit: millimeters) and angular deviation The data (in degrees) is encapsulated as alignment status data, along with the current section's identifier and timestamp, for subsequent verification of alignment requirements.

[0017] The alignment status data is compared with the alignment deviation tolerance. If the deviation exceeds the tolerance, a fine-tuning actuator is triggered to compensate until all deviation indicators meet the tolerance requirements. An alignment confirmation signal is then output. This process specifically includes: adjusting the radial offset in the alignment status data... and angular deviation Compare with the radial offset tolerance obtained in step 1.1 respectively and angular skew tolerance Compare; if or This triggers a fine-tuning actuator, such as a piezoelectric ceramic micro-stage or a compensation mechanism driven by a differential head, to perform a compensation operation: when the radial offset exceeds the limit, the actuator moves slightly along the X and Y axes in the horizontal plane; when the angular deviation exceeds the limit, the actuator rotates slightly around the Z axis; after each compensation, the above measurement and calculation actions are automatically repeated to obtain a new result. and This is repeated, comparing the result with the tolerance again, until all deviation indicators are met. and Once all indicators are qualified, an alignment confirmation signal is output. The signal includes the final offset and skew values ​​that meet the requirements, as well as a confirmation timestamp, indicating that the high-speed solenoid valve and the drive end of the diaphragm valve under test have reached the optimal alignment state.

[0018] Based on the locked sliding stage position, the current physical gap value, and the alignment confirmation signal, a positioning reference signal containing the real-time drive distance and alignment status is formed. Simultaneously, the positioning reference signal is transmitted to the cyclic excitation unit, specifically including: setting the final physical gap value corresponding to the locked sliding stage position... The centering confirmation signal output and the recorded and locked final drive distance parameters are integrated. Following a predefined data frame format, these data are sequentially encapsulated as follows: frame header (fixed flag), drive distance value (floating-point, mm), centering status confirmation flag (Boolean, 1 indicates confirmation), final physical gap value (floating-point), timestamp (64-bit integer), and check bit. After encapsulation, a complete positioning reference signal is formed. This positioning reference signal not only includes the real-time drive distance value but also a flag indicating that the centering status has been confirmed and precise timing information. Finally, this positioning reference signal is transmitted in real-time to the cyclic excitation unit as the precise initial condition for generating subsequent periodic control commands.

[0019] like Figure 2 As shown, in another preferred embodiment of the present invention, periodic control commands are generated based on the positioning reference signal to obtain an alternating opening and closing sequence acting on the diaphragm valve under test, including: The locked drive distance parameter and alignment status confirmation flag are parsed from the positioning reference signal to obtain the excitation parameter initialization conditions. Specifically, this includes: receiving the positioning reference signal transmitted from the drive positioning unit; parsing this signal according to a predefined data frame format; and extracting the drive distance value field, alignment status confirmation flag field, and timestamp field. Specifically, after confirming that the alignment status confirmation flag is true, the drive distance value is used as the locked drive distance parameter and temporarily stored together with the alignment status confirmation flag in the excitation parameter register. This register is a dedicated area in a dual-port random access memory (DPRAM), with a size of 32 bytes, divided into four... The system has three channels: Channel 0 stores the driving distance parameter (floating-point), Channel 1 stores the acknowledgment flag (Boolean), Channel 2 stores the timestamp, and Channel 3 stores the checksum, used for quick reading of subsequent excitation parameter initialization conditions. Based on this, it automatically verifies whether the driving distance is within a preset safe range (e.g., 0 to 0.2 mm), and compares the current timestamp with the last received timestamp to check its monotonically increasing nature to prevent data loss or out-of-order delivery. After the verification is passed, the excitation parameter initialization conditions are generated. These conditions include an enable signal (active high) that allows the start of the cyclic test and an initial driving distance reference value, and the conditions are latched, awaiting the triggering of subsequent steps.

[0020] Based on the initialization conditions of the excitation parameters, a preset cyclic test recipe is loaded. This preset cyclic test recipe includes the target total number of cycles, operating frequency, duty cycle, and test pressure value, forming a waveform parameter set. Specifically, this includes: loading a pre-configured cyclic test recipe from a local non-volatile recipe library based on the enable signal in the initialization conditions of the excitation parameters. Specifically, this local non-volatile recipe library consists of an EEPROM or Flash memory, internally divided into several recipe storage areas, each corresponding to a test scheme, indexed by recipe number. The pre-configured cyclic test recipe is stored in one of the storage areas in the form of a structured table. Each row of the table records a parameter item, including: parameter name (e.g., target total number of cycles), value, unit, and upper and lower limits. The recipe specifically includes the target total number of cycles, operating frequency, duty cycle, and test pressure value, for example, 0.6 MPa.

[0021] Simultaneously, the values ​​in the formula are checked for consistency with the drive distance parameter in the excitation parameter initialization conditions to ensure that the operating frequency matches the solenoid valve response capability corresponding to the drive distance (i.e., the frequency does not exceed the maximum operating frequency of the solenoid valve) and the test pressure does not exceed the rated pressure of the pipeline, such as 0.8 MPa. After the check is passed, these parameters are combined into a waveform parameter set according to the internal data protocol. The internal data protocol adopts a compact binary format, and the protocol frame consists of the following in sequence: protocol identifier, data length, parameter type field, parameter field, and end flag. The waveform parameter set contains all the time-domain parameters (including period, pulse width, and number of pulses) and amplitude parameters required to generate the control waveform, i.e., the drive voltage level.

[0022] The waveform parameter set is input into the programmable logic controller (PLC), which generates a corresponding high-speed solenoid valve drive pulse sequence. Each pulse in the drive pulse sequence corresponds to an opening or closing command of the diaphragm valve under test, resulting in the original pulse control sequence. Specifically, the generated waveform parameter set is transmitted to the waveform generation module inside the PLC via a real-time industrial Ethernet (such as EtherCAT) or a high-speed serial communication interface. Specifically, the PLC is an embedded computer system designed for industrial real-time control. Its core components include: a central processing unit (CPU) responsible for executing user-written control programs; a hardware high-speed counter module for accurately counting the number of pulses; a multi-channel timer module for generating microsecond-level precision timer interrupts; and a digital output module for outputting high and low level signals. After receiving the waveform parameter set, the PLC first loads the target total number of cycles into the preset value register of the high-speed counter, converts the operating frequency into the interrupt period of the timer, where the interrupt period = 1 / operating frequency, and converts the duty cycle into the high-level duration, where the high-level duration = interrupt period × duty cycle.

[0023] Furthermore, the timer module is triggered cyclically according to the set interrupt cycle. Each time it is triggered, the CPU unit executes the interrupt service routine: first, the digital output channel is set (outputting a high level), then a delay timer is started, with the delay time being the duration of the high level. After the delay ends, the digital output channel is reset (outputting a low level). This process is repeated until the number of output pulses recorded by the high-speed counter reaches the target total number of cycles. During this process, each pulse's rising and falling edges are accompanied by a time stamp accurate to the microsecond level, generated by the timer's hardware capture function, thereby obtaining the original pulse control sequence. The sequence is output in parallel from the digital output module in digital form, with each channel corresponding to a drive signal for a valve under test.

[0024] The original pulse control sequence is sent to the power drive module of the high-speed solenoid valve, where it is amplified into a solenoid valve excitation signal with sufficient drive current, resulting in an amplified excitation pulse sequence. Specifically, the generated original pulse control sequence is sent to the power drive module of the high-speed solenoid valve via an opto-isolated parallel data line. The power drive module internally includes MOSFET or IGBT power switches, corresponding gate drive circuits, and freewheeling diodes. When each pulse in the original pulse control sequence is high, the power drive module converts the low-voltage control signal, for example, 5 volts, into a higher voltage and current sufficient to drive the solenoid valve coil, such as 24 volts and 2 amps, forming the amplified excitation pulse sequence. When the pulse is low, the output is 0. Simultaneously, the power drive module also has built-in overcurrent and short-circuit protection functions to ensure automatic shutdown in case of solenoid valve coil abnormalities. The amplified excitation pulse sequence is completely consistent with the original pulse control sequence in waveform timing, but with significantly enhanced driving capability.

[0025] Based on the amplified excitation pulse sequence, the high-speed solenoid valve is controlled to alternately turn on and off according to a set frequency and duty cycle. This drives the diaphragm valve under test to perform corresponding opening and closing actions through the pilot air path, forming an alternating opening and closing action sequence acting on the diaphragm valve under test. Specifically, the amplified excitation pulse sequence is directly applied to both ends of the solenoid coil of the high-speed solenoid valve. Specifically, the high level of each excitation pulse causes the armature inside the solenoid valve to overcome the spring force and close, thereby opening the air inlet of the pilot air path and closing the exhaust port; when the low level is low, the armature is reset under the action of the spring force, closing the air inlet and opening the exhaust port; the switching action of the pilot air path further drives the diaphragm of the diaphragm valve under test to move, so that the diaphragm valve performs opening and closing actions according to the same frequency and duty cycle. On this basis, the actual action state of the valve is fed back in real time by a position sensor or pressure sensor integrated on the valve body, and a closed-loop comparison is performed with the drive pulse to verify the consistency of the action. This forms a continuous, stable, and precisely controllable alternating opening and closing action sequence, which directly acts on the diaphragm valve under test, driving it to complete the entire durability cycle test.

[0026] In a preferred embodiment of the present invention, the alternating opening and closing sequence is received, and the valve response time, pipeline pressure fluctuation, helium leakage rate, and particulate matter concentration triggered by this sequence are monitored simultaneously to form a raw test dataset containing multi-parameter time series, including: Using the alternating on / off action sequence as a trigger reference, the monitoring channels of the multi-parameter acquisition are activated to obtain the acquisition synchronization trigger signal. Specifically, this includes: receiving the alternating on / off action sequence transmitted from the cyclic excitation unit, and using the rising edge of each pulse of the sequence as a trigger reference; specifically, the rising edge signal of the alternating on / off action sequence is captured by a high-speed digital input channel using hardware edge detection; the high-speed digital input channel is composed of a high-speed opto-isolator, a Schmitt trigger shaping circuit, and edge detection logic in a programmable logic device (CPLD) cascaded together: the high-speed opto-isolator electrically isolates the external signal from the internal logic circuit, and converts the signal level to the internal standard level; the Schmitt trigger shaping circuit shapes the waveform of the isolated signal, eliminates jitter at the leading and trailing edges, and outputs a steep square wave; the edge detection logic in the CPLD captures the rising edge of the square wave in real time, and once a low-to-high level transition is detected, the state is immediately latched and subsequent actions are triggered.

[0027] The maximum sampling frequency of the channel is no less than 10 MHz, capable of recognizing pulse signals with a pulse width of no less than 100 nanoseconds, ensuring microsecond-level time detection accuracy. When the first rising edge is detected, the hardware triggering logic inside the CPLD immediately generates a data acquisition synchronization trigger signal. This trigger signal is simultaneously broadcast to each monitoring channel of the multi-parameter acquisition in a multi-channel parallel interrupt manner, including the response time capture channel, pressure acquisition channel, leakage rate reading channel, and particulate matter counting channel. These channels function in concert: the response time capture channel uses the trigger signal as a starting point to accurately measure the time delay of valve action; the pressure acquisition channel begins continuous sampling under the synchronization of the trigger signal, providing real-time changes in the gas path status; the leakage rate reading channel uses the trigger signal as a reference to periodically acquire sealing performance data; and the particulate matter counting channel begins accumulating or reading the number of particles after synchronization triggering. Each channel has an independent hardware trigger, which simultaneously starts its own data acquisition process after receiving the signal, thereby achieving microsecond-level synchronization of the entire channel.

[0028] Based on the acquisition synchronization trigger signal, the valve response time corresponding to each alternating opening and closing action is captured in real time to obtain response time sequence data containing timestamps. Specifically, the process includes: initiating the response time capture process based on the acquisition synchronization trigger signal; specifically, when the rising edge of each alternating opening and closing action corresponds to the opening command and the falling edge corresponds to the closing command, a start capture command is sent to a high-precision timer (resolution 0.1 microseconds); simultaneously, the actual completion time of the valve core is sensed in real time through a non-contact eddy current sensor installed at the valve core of the diaphragm valve under test or through a valve downstream pressure rising edge detection circuit; when the sensor output signal exceeds a set threshold, the current value of the timer is immediately latched to obtain the opening response time (from the rising edge of the command to the completion of valve core opening) and the closing response time (from the falling edge of the command to the completion of valve core closing); each response time value, along with its corresponding action sequence number, command type (including opening or closing), and the absolute timestamp of the trigger time (i.e., a 64-bit microsecond count since system startup), is encapsulated into a record and sequentially appended to the circular queue of the response time sequence data buffer.

[0029] While capturing the response time, the pipeline pressure value is continuously acquired, and the pressure waveform during each opening and closing action is recorded to obtain pressure fluctuation sequence data including timestamps. Specifically, after the acquisition synchronization trigger signal is activated, a multi-channel synchronous sampling analog-to-digital converter (ADC) is used to synchronously sample the signals of two piezoresistive pressure sensors installed upstream and downstream of the diaphragm valve under test at a sampling rate of not less than 2 kHz. The ADC is a successive approximation type or Δ-Σ type converter, with an integrated multi-channel sample-and-hold circuit, which can simultaneously freeze the analog input voltage of each channel to eliminate time deviation between channels and ensure... To ensure strict synchronization of upstream and downstream pressure data, an anti-aliasing filter and a programmable gain amplifier are connected in series at the front end of each sampling channel. The anti-aliasing filter is a second-order low-pass active filter with a cutoff frequency set to half the sampling rate, i.e., 1 kHz, used to filter out noise components higher than the Nyquist frequency and prevent high-frequency interference from folding back into the passband and causing measurement errors. The programmable gain amplifier selects the gain factor through digital instructions, such as 1, 2, 4, or 8 times, to amplify the sensor output signals of different ranges to the full-scale input range of the ADC, such as 0 to 5 volts, so as to obtain high resolution even with small pressure changes.

[0030] The two components are cascaded: the sensor signal is first filtered by an anti-aliasing filter to remove high-frequency noise, then fed into a programmable gain amplifier for amplitude conditioning, and finally output to the input of the ADC to complete analog-to-digital conversion; during each opening and closing operation, the ADC continuously records the pressure waveform data throughout the entire cycle from the start of opening to the end of closing, including the pressure drop edge caused by gas flow at the moment of opening, the pressure plateau value during steady-state flow, and the pressure rise edge at the moment of closing; each sampling point is supplemented with a relative time offset (in units of sampling period) after the rising edge trigger and a global absolute timestamp, forming a pressure fluctuation sequence data containing timestamps, which is temporarily stored in the dual-port RAM.

[0031] While acquiring pressure, the helium leakage rate is continuously read, and the leakage rate value is correlated with the current number of actions and the timestamp to obtain leakage rate sequence data including the timestamp. Specifically, while acquiring pressure, the helium leakage rate is continuously read. Specifically, the analog voltage output (0 to 10 volts, linearly corresponding to the leakage rate range) of the helium mass spectrometer leak detector is connected to an independent analog-to-digital converter channel for continuous acquisition at a configurable sampling rate, such as 20 Hz. After each sampling, the voltage value is immediately converted into a leakage rate value. At the same time, the cumulative number of current alternating on / off actions, the action counter value provided by the cyclic excitation unit through shared memory, and the current absolute timestamp are read. The leakage rate value, the number of actions, and the timestamp are correlated to form a triplet record. In order to reduce storage pressure while ensuring data integrity, two data streams are maintained in parallel: the original sampling sequence for transient analysis, and the statistical values ​​within each action cycle, including the average, maximum, and minimum values, ultimately generating leakage rate sequence data including the timestamp.

[0032] While collecting the leakage rate, the particle count at the gas outlet is continuously read to obtain particulate matter concentration sequence data including timestamps. Specifically, this includes: simultaneously collecting the leakage rate and continuously reading the particle count at the gas outlet; specifically, the digital output interface of the particulate matter analyzer (such as RS-485 or Ethernet, using Modbus TCP protocol) is read in a periodic polling manner; the particulate matter analyzer defaults to reading at a first polling interval (such as 0.5 seconds); each reading operation acquires the cumulative particle count of multiple particle size channels (e.g., 0.1 μm, 0.2 μm, 0.5 μm, 1.0 μm); to overcome the collaborative analysis error that may be caused by the response lag of the particulate matter analyzer, this system adopts a collaborative correction mechanism, which includes the following: Specifically, intelligent polling acceleration begins first. The data fusion processing unit continuously monitors the changing trend of the leakage rate sequence. A sliding window (window width, for example, 10 sampling points) is used to calculate the first difference of the leakage rate sequence in real time. Then, the absolute value of the difference sequence is taken and the average value of the window is calculated as the instantaneous leakage rate change rate. When this change rate exceeds a preset threshold (this threshold can be retrieved from the test formula according to the valve model, and is usually set by default to 50% of the current average leakage rate or 3 times the standard deviation of the historical stable period change rate), it is determined that the valve has entered the performance accelerated degradation stage. At this time, the polling interval of the particulate matter tester is automatically shortened from the first polling interval to the second polling interval, for example, 0.1 seconds, through ModbusTCP commands, thereby effectively improving the acquisition density of particulate matter concentration data to capture transient particle generation events in the accelerated degradation process.

[0033] Furthermore, dynamic interpolation alignment is performed. When performing multi-dimensional alignment and merging, the zero-order preservation method is no longer used for particulate matter concentration sequences; instead, a dynamic interpolation method based on a physical degradation model is employed. The core of this dynamic interpolation method is to utilize the strong correlation between the leakage rate, which characterizes seal decay, and particulate matter concentration to construct a first-order dynamic model. This first-order dynamic model takes the form of estimating the current rate of change in particulate matter concentration as a linear function of the current leakage rate. ;in C Particulate matter concentration (unit: particles / ft) 3 ), Leakage rate time series (unit: Pa) m 3 / s), and These are the model coefficients; the training process of the model is as follows: In the early stage of testing (the valve degradation stabilization stage), the coefficients are fitted using the sampling values ​​of the particulate matter analyzer itself and its corresponding leakage rate data through the least squares method. and .

[0034] When interpolation is required between two samplings by the particulate matter analyzer, the most recent particulate matter concentration sampling value is used as the initial value. Numerical integration is performed using the first-order dynamic model and the continuously monitored leakage rate sequence, such as the Euler method or the trapezoidal method, to estimate the particulate matter concentration at any time point within the interval. This enables high-resolution reconstruction of particulate matter data on the time axis, aligning it precisely to the same pressure sampling time as other parameters, effectively improving the synchronization of multi-parameter collaborative analysis.

[0035] The response time series data, pressure fluctuation series data, leakage rate series data, and particulate matter concentration series data are aligned and merged in multiple dimensions according to the same timestamp to form an original test dataset containing multiple parameter time series. Specifically, this includes: aligning and merging the four generated series data in multiple dimensions according to the same absolute timestamp; using the timestamp of the pressure fluctuation series as a reference, using nearest neighbor matching for the response time series; mapping the leakage rate series to each pressure sampling time using a linear interpolation method; and applying the above dynamic interpolation alignment method to the particulate matter concentration series, using the leakage rate series as a driver, to reconstruct the particulate matter concentration to each pressure sampling time. Each merged super-record includes a global timestamp, number of actions, start response time, stop response time, upstream pressure, downstream pressure, instantaneous leakage rate, and particle count values ​​for each particle size.

[0036] All records are arranged in ascending chronological order and preprocessed by outlier removal (e.g., bad pixels exceeding the sensor's range) and missing value filling (e.g., linear interpolation), forming a complete raw test dataset containing multi-parameter time series. This dataset is written to a circular buffer in a local solid-state drive in columnar storage format, while a copy is maintained in a memory-mapped file for real-time access by the data fusion processing unit. Preferably, to adapt to applications in ultra-clean environments such as atomic layer deposition (ALD), this diaphragm valve durability multi-parameter collaborative testing system also includes a contamination source elimination mechanism. Specifically, before installing the diaphragm valve under test, the background particle concentration of the testing system itself is calibrated; that is, the system is fully run without the valve or with a known clean blind flange installed. The dynamic positioning unit and cyclic excitation unit ensure that all mechanical moving parts and optical sensors are in normal working condition. The background particle concentration baseline generated by the system's own operation is continuously collected and recorded by the particulate matter tester (e.g., collected for 30 minutes, and the average value and standard deviation are taken). In subsequent formal testing, the real-time collected raw particulate matter concentration data is automatically subtracted from the background baseline to obtain the corrected particulate matter concentration data. All subsequent data analysis, ellipsoid construction, and degradation judgment are based on this corrected data. If the corrected particulate matter concentration still shows an upward trend and exceeds the preset durability failure threshold, it can be confirmed that the particulate matter originates from the wear or sealing failure of the valve under test. This ensures the effectiveness and distinguishability of particulate matter concentration, a key monitoring indicator, in a clean environment.

[0037] In a preferred embodiment of the present invention, a constrained ellipsoidal space characterizing the multidimensional performance degradation state of the valve is constructed by fitting the sampling features of the multidimensional parametric evolution trajectory, including: The response time series, pressure fluctuation series, leakage rate series, and co-corrected particulate matter concentration series are extracted from the original test dataset to obtain four independent parameter time series arrays. Specifically, the original test dataset is stored in columnar storage format, and each record contains a timestamp. Number of actions Enable response time Closing response time Upstream pressure Downstream pressure Instantaneous leakage rate and particle counts of various sizes Read in order according to field name: and Merged into response time series Each time point is taken As a comprehensive response time; and The difference As a pressure fluctuation sequence ;Will As a leakage rate sequence Sum of particle counts for each particle size channel. As particulate matter concentration sequence This results in four independent parameter timing arrays: 、 、 、 ,in For the first Each sampling time ( ), This represents the total number of sampling points.

[0038] For each parameter time series array, perform second derivative operations to locate the point with the maximum curvature. Use the time position corresponding to this point as the performance degradation feature moment. Extract parameter values ​​near each feature moment to construct a multi-dimensional parameter feature point cloud. Specifically, this includes: first, performing Gaussian smoothing preprocessing on the time series array to reduce noise; then calculating the first and second derivatives; to respond to the time series... For example, its first derivative Using the central difference formula:

[0039] in Sampling interval; second derivative for:

[0040] curvature The calculation formula is:

[0041] Find The time point when the maximum value is reached This is the point with the largest curvature in the response time series; similarly, for the pressure fluctuation series... Leakage rate sequence and particulate matter concentration sequence Perform the above operations separately to obtain the time position corresponding to the point of maximum curvature for each point. 、 and The time locations corresponding to these points with the largest curvature are taken as the characteristic moments of performance degradation, and denoted as the set. Based on this, a time window is extracted near each characteristic time point, for example, the window width is... Each sampling point is used to extract all parameter values ​​within a window, forming a multidimensional parameter feature point cloud. .

[0042] Each point in the multidimensional parameter feature point cloud is spatially mapped according to three physical structure mapping dimensions: the valve core diaphragm micro-variation region, the pilot air inlet cross-section, and the sealing lip contact zone. This yields the three-dimensional spatial coordinates of each feature point in the physical structure coordinate system, forming a structured feature point set. Specifically, this involves spatially mapping each point in the multidimensional parameter feature point cloud according to three physical structure mapping dimensions: the valve core diaphragm micro-variation region, the pilot air inlet cross-section, and the sealing lip contact zone. Specifically, the valve core diaphragm micro-variation region corresponds to the sensitive mapping of the response time series, the pilot air inlet cross-section corresponds to the sensitive mapping of the pressure fluctuation series, and the sealing lip contact zone corresponds to the joint sensitive mapping of the leakage rate series and the particulate matter concentration series. Therefore, a structured feature point set is defined from the original four-dimensional parameter... To three-dimensional physical structure coordinates Mapping function:

[0043] in, These are the minimum and maximum values ​​of the response time series, respectively. 、 、 Similar definition; Let be the weighting coefficient, satisfying , wherein the weighting coefficient The determination method can be based on principal component analysis of historical failure data. The specific steps are as follows: First, obtain the full life cycle data of the same model of diaphragm valve recorded in historical durability tests to form a sample matrix. Each column of this matrix corresponds to the response time. Pressure fluctuations P Leakage rate L and particulate matter concentration C Each row corresponds to a sampling time. Principal component analysis (PCA) is then performed on this sample matrix to obtain the loading matrix of each principal component. Based on the loading values ​​of the first, second, and third principal components on each original parameter, orthogonal rotation (such as the maximum variance method) is used to polarize the loading values ​​towards 0 or 1, thereby determining the correspondence between each physical structure mapping dimension and the most relevant parameter, and thus determining the initial values ​​of the weighting coefficients. For example, if after rotation the loading of the first principal component on the response time is close to 1, and the loading on other parameters is close to 0, then the valve core diaphragm micro-variation region (X-axis) is mainly related to the response time. A value of 0.85 to 0.95 is acceptable. (The remaining allocations related to pressure, leakage, and particles) are taken as 0.05 to 0.15.

[0044] Finally, the weighting coefficients are fine-tuned through calibration experiments with no fewer than three sets of known failure results: the full life data of each set of calibrated valves are input into the system, and the weighting coefficients are iteratively adjusted with the goal of ensuring that the volume shrinkage rate of the constrained ellipsoidal space is within a preset range for the predicted life (e.g., the error is less than ±10% of the current predicted life) to obtain the final set of weighting coefficients. As another implementation method, a sensitivity analysis-based method can also be used to determine the weighting coefficients. This method first sets initial values ​​for all weighting coefficients (e.g., evenly distributed according to experience), and then applies small perturbations (e.g., ±5%) to each weighting coefficient in turn, keeping other coefficients unchanged. The entire data processing and life prediction process is run, and the relative rate of change of the final predicted life caused by the perturbation is calculated, i.e., the sensitivity. Finally, the set of weighting coefficients that makes the life prediction results most stable (i.e., the lowest sensitivity) for different test batches of data is selected as the final value. This method does not rely on historical failure labels at all and is suitable for new valves or situations where data accumulation is insufficient.

[0045] The real-time extreme coordinates of parameters in each physical structure dimension are extracted from the structured feature point set and used as sampling features for the multidimensional parameter evolution trajectory, resulting in a sampling feature set. Specifically, this involves: statistically analyzing the coordinates in each physical structure dimension from the structured feature point set to identify the boundaries of the data point distribution; specifically, calculating the maximum value of the coordinate in that dimension. and minimum value These two extreme points represent the most unfavorable state of the valve at this degradation characteristic moment, in terms of response time (X-axis), pressure fluctuation (Y-axis), or seal leakage / particulate matter (Z-axis), i.e., the boundary of the degradation degree. These maximum and minimum points are extracted and defined as multidimensional parametric evolution trajectory sampling features. Simultaneously, to further constrain the ellipsoidal shape, the median coordinates are also extracted as reference points. Thus, the extreme coordinates in each dimension constitute a sampling feature triplet containing the maximum, minimum, and median values, denoted as the sampling feature. 、 、 ， Its form is as follows:

[0046] Combine the sampled features of all dimensions into a sampled feature set. Each sampled feature in the set represents a critical boundary state of valve performance degradation in that physical structure dimension.

[0047] Using the extreme points in each sampled feature as boundary constraint points, a minimum volume closed surface containing all extreme points is fitted in the three-dimensional physical structure mapping space to obtain a constrained ellipsoid space characterizing the multidimensional performance degradation state of the valve. Specifically, this includes: using the sampled feature set Each sampling feature in The included extreme points are boundary constraint points. In the three-dimensional physical structure mapping space, a minimum volume closed surface containing all extreme points is fitted to obtain a constrained ellipsoid space characterizing the multidimensional performance degradation state of the valve. Specifically, the extreme points in all sampled features are merged into a point set. There are a total of 27 points. The minimum, median, and maximum values ​​are taken for each dimension, forming 3×3×3=27 boundary points. Let the general equation of the ellipsoid be:

[0048] in The coordinates of the ellipsoid center are: Let be the semi-axis lengths of the ellipsoid in the X, Y, and Z directions, respectively; the following optimization problem is solved under the constraints: For all Minimize the volume of the ellipsoid under the premise that it holds true. The optimal parameters are obtained by iteratively solving the problem using a minimum volume enclosing ellipsoid algorithm (such as the Khachiyan algorithm): The resulting ellipsoid is the constrained ellipsoid space characterizing the multidimensional performance degradation of the valve. It's important to note that this constrained ellipsoid space is not based on the thermodynamic potential energy function of traditional physics. Instead, it's a model constructed using data-driven geometric modeling methods, mapping multiple engineering test parameters reflecting key valve performance onto three physical structural dimensions: the valve core diaphragm, the pilot air passage, and the sealing lip. This model quantifies the normal operating range boundaries of the valve. The set of points inside the ellipsoid represents the range of multi-parameter coordinated changes in the valve under healthy conditions. Changes in the ellipsoid's center position, the length of each semi-axis, and its volume directly correspond to the degree of degradation in comprehensive performance across multiple dimensions, including response time, pressure fluctuations, sealing leakage, and particle generation. For example, an expansion of the ellipsoid's volume indicates increased multi-parameter dispersion and a tendency for valve instability; a sharp contraction in volume may indicate a rapid deterioration in certain performance dimensions and a drastic narrowing of the operable range. Therefore, the evolution trajectory of the ellipsoid's morphological parameters can clearly and unambiguously characterize the degradation process of the diaphragm valve.

[0049] In a preferred embodiment of the present invention, the step of extracting the gradient drift and dissipation volume expansion rate of the degradation-sensitive direction tensor trace, and inputting the gradient drift and dissipation volume expansion rate into a preset multidimensional decay correlation model, yields a feature parameter matrix characterizing the performance degradation trajectory of the diaphragm valve, including: Representing the constrained ellipsoid space as a covariance matrix with the semi-axis lengths of each principal axis as characteristic parameters yields a geometric description of the ellipsoid in a high-dimensional parametric space. Specifically, this includes: representing the constrained ellipsoid space as a covariance matrix with the semi-axis lengths of each principal axis as characteristic parameters, thus obtaining a geometric description of the ellipsoid in a high-dimensional parametric space; specifically, from the output ellipsoid parameter matrix... Extracting the semi-axis length Construct a covariance matrix corresponding to a three-dimensional ellipsoid. Its form is: ; The matrix describes the variance distribution of the ellipsoid along each principal axis, with diagonal elements being the square of the semi-axis length and off-diagonal elements being zero (indicating orthogonality between principal axes); thus, a geometric description of the ellipsoid in a high-dimensional parametric space is obtained, which serves as the input for subsequent dimensionality reduction.

[0050] Principal component tensor projection dimensionality reduction is performed on the covariance matrix, extracting the three principal axes with the largest eigenvalues. The ellipsoid is then projected along these principal axes to a lower-dimensional subspace to obtain the dimensionality-reduced principal component tensor set. Specifically, this includes: performing eigenvalue decomposition on the covariance matrix Fn, using the following formula: ; in It is a 3×3 orthogonal matrix, and its column vectors are... The unit vector along the principal axis; It is a diagonal matrix. For the corresponding eigenvalues; because It is already a diagonal matrix, in fact (Identity matrix) Projecting is performed on the three principal axes with the largest eigenvalues ​​(in this case, all three axes), constructing the projection matrix. Points in the original three-dimensional ellipsoidal space Projecting onto a lower-dimensional subspace yields the reduced principal component tensor set. Since all three principal axes are retained, the dimension reduction still results in three dimensions, but coordinate rotation is achieved to align the directions of each principal component with the original coordinate axes.

[0051] On the dimensionality-reduced principal component tensor set, the rate of change along each principal axis is calculated, and the direction with the largest rate of change is identified as the degradation-sensitive direction. The change sequence of the tensor trace along the direction with the number of iterations is extracted to obtain the gradient drift of the tensor trace in the degradation-sensitive direction. Specifically, this includes: on the dimensionality-reduced principal component tensor set, the rate of change along each principal axis is calculated, and the direction with the largest rate of change is identified as the degradation-sensitive direction. The change sequence of the tensor trace along the direction with the number of iterations is extracted to obtain the gradient drift of the tensor trace in the degradation-sensitive direction. Specifically, assuming that after the projection, in the number of iterations... ( , The ellipsoid parameters corresponding to the total number of iterations are: For each principal axis direction The tensor trace in this direction is defined as That is, the first One eigenvalue (semi-axis square); calculate The sequence of changes with the number of cycles; the first derivative (rate of change) is calculated using the central difference method:

[0052] in (Interval between adjacent cycles); then calculate the absolute value of the rate of change in each direction. And calculate its average value over the entire lifespan. Compare the average values ​​of the three directions, and identify the direction with the largest average value; denote this as the degradation-sensitive direction. Extract the tensor trace sequence in this direction. The sequence Defined as the cumulative change from the initial value to the current value, expressed as the slope of the linear fit:

[0053] in ， The calculated slope is the trace of the degeneration-sensitive direction tensor. .

[0054] The relative rate of change of the volume of the constrained ellipsoidal space with the number of iterations is calculated. The absolute value of the ratio of the current volume to the initial volume, minus one, is taken to obtain the dissipative volume expansion rate. Specifically, this includes: calculating the relative rate of change of the volume of the constrained ellipsoidal space with the number of iterations; subtracting one from the ratio of the current volume to the initial volume and taking the absolute value, thus obtaining the dissipative volume expansion rate; specifically, in the number of iterations... At that time, the volume of the ellipsoid as follows: ; in For the first The semi-axis length of the output during the next loop; initial volume. Define the dissipative volume expansion rate. for:

[0055] This value represents the degree of deviation of the current volume from the initial volume (whether expanding or contracting), and is dimensionless; as valve performance deteriorates, the difference between the ellipsoidal volume and the initial state generally increases monotonically, hence it monotonically increases and is recorded. The sequence of changes with the number of iterations.

[0056] Using the gradient drift and dissipation volume expansion rate as input variables, a preset multidimensional decay correlation model is input. This model employs a multivariate nonlinear regression or neural network structure to obtain a feature parameter matrix that reflects the current valve performance degradation, including the decay rate and coupling coefficients for each parameter dimension. Specifically, the model uses a multivariate nonlinear regression or neural network structure to obtain a feature parameter matrix that reflects the current valve performance degradation, including the decay rate and coupling coefficients for each parameter dimension. Specifically, let the current loop count be... The input variable is gradient drift. and dissipative volume expansion rate The preset multidimensional decay correlation model adopts an exponential decay form, as shown below:

[0057] in The output is a feature parameter matrix, whose elements include the decay rate of each parameter dimension (e.g., response time decay rate). Leakage rate decay rate (etc.) and coupling coefficients (e.g., response-leakage coupling coefficient) ); This is the initial feature parameter matrix (calibration values). This is the model coefficient matrix (obtained through training with historical data). This is the attenuation factor; if a neural network structure is used, then the two nodes in the input layer... The model has several hidden layers, and the number of nodes in the output layer is equal to the number of elements in the feature parameter matrix; the model output is the feature parameter matrix. This matrix contains the decay rate and coupling coefficient of each parameter dimension, used to characterize the current performance degradation trajectory of the diaphragm valve; the matrix is ​​then passed to the durability assessment unit for failure threshold matching analysis.

[0058] In a preferred embodiment of the present invention, the feature parameter matrix is ​​matched and analyzed with a preset durability failure threshold to obtain a durability evaluation result including life prediction results, abnormal alarm instructions, and a comprehensive test report, including: The decay rate and coupling coefficient of each parameter dimension are extracted from the feature parameter matrix to obtain a performance degradation quantification index set. Specifically, this includes: extracting the decay rate of each parameter dimension and the coupling coefficient between different parameters sequentially from the feature parameter matrix output by the data fusion processing unit according to the set field index; specifically, the feature parameter matrix is ​​a row vector containing multiple elements, where the first few elements correspond to the decay rate of response time, the increase rate of leakage rate, the increase rate of particulate matter concentration, and the increase rate of dissipation volume expansion rate, respectively; the latter few elements correspond to the coupling coefficient between response time and leakage rate, the coupling coefficient between pressure fluctuation and particulate matter, etc.; these extracted values ​​are assembled into a performance degradation quantification index set in a fixed order, and the index set is temporarily stored in memory in the form of key-value pairs, with each index accompanied by its physical unit, effective range, and corresponding failure threshold reference.

[0059] The performance degradation quantification index set is compared item by item with a preset durability failure threshold to obtain the deviation of each index from the threshold and the over-limit flag. Specifically, this includes comparing each index in the performance degradation quantification index set with a pre-stored durability failure threshold; the preset failure threshold includes an upper limit for response time drift, such as 200 microseconds, and an upper limit for leakage rate, such as 1×10. -6 Pa m 3 / s, the upper limit of particulate matter concentration, such as 1000 particles per cubic foot, and the upper limit of volume expansion rate, such as 0.5; for each indicator, calculate the difference between its current value and the threshold. If the current value does not exceed the threshold, the deviation is negative or 0, and the over-limit flag is set to 0; if the current value exceeds the threshold, the deviation is positive, and the over-limit flag is set to 1; all comparison results are arranged in the same order as the indicator set to form a deviation array and an over-limit flag array.

[0060] Based on the deviation and over-limit flags of each indicator, the critical failure state is determined. If any key indicator exceeds the limit or the comprehensive weighted value exceeds the alarm threshold, a failure warning signal is generated. Specifically, this involves: comprehensively determining whether the valve has reached the critical failure state based on the deviation and over-limit flags of each indicator; specifically, first reading the generated over-limit flag array, and filtering out two flags pre-marked as key indicators, namely the leakage rate over-limit flag and the response time over-limit flag; if either of these two key indicators has a flag of 1, it is immediately determined that the valve is close to the critical failure state, and a failure warning signal is directly generated without performing subsequent weighted calculations; if all key indicators do not exceed the limits, the comprehensive weighted value calculation process is entered: at this time, the weight coefficients corresponding to each indicator are loaded from the configuration memory, with the leakage rate weight set to the highest (e.g., 0.5), the response time weight second (e.g., 0.3), the particulate matter concentration weight third (e.g., 0.15), and the volume expansion rate weight lowest (e.g., 0.05).

[0061] The deviation (positive or 0) of each indicator is multiplied by its weighting coefficient, and all multiplications are summed to obtain a comprehensive weighted value. This comprehensive weighted value is compared with a preset alarm threshold (e.g., 0.8, which is obtained by statistically analyzing historical failure data and stored in non-volatile memory). If the comprehensive weighted value is greater than or equal to the alarm threshold, it is determined that failure is imminent, and a failure warning signal is generated. If it is less than the alarm threshold, it is determined that the current state is still within the acceptable range, and no warning signal is generated. When the failure warning signal is generated, it is automatically encapsulated into a data packet containing three fields: trigger time, a list of out-of-limit indicators (recording the names of all indicators with an out-of-limit flag of 1 and their corresponding deviations), and the calculated comprehensive weighted value. The data packet is written to the event log file in the local solid-state drive through the event bus and sent to the display interface and alarm management module through the message queue for subsequent steps.

[0062] After receiving a failure warning signal, the remaining effective number of cycles is predicted based on the current number of cycles and the previously acquired historical trends, resulting in a lifetime prediction. Specifically, this involves: immediately initiating the lifetime prediction process upon receiving the failure warning signal; first, reading the number of currently completed effective cycles, accumulated by the cycle excitation unit, and retrieving historical trend data for each indicator from the historical database for the most recent cycles (e.g., the last 1000 cycles); then, for each key indicator, such as response time and leakage rate, using linear extrapolation: with the number of cycles as the x-axis and the indicator value as the y-axis, performing least-squares linear fitting on the historical data points to obtain a straight line; extrapolating this line to the failure threshold of the indicator to solve for the corresponding number of cycles; taking the minimum value among all extrapolated results as the predicted value of the remaining effective number of cycles; if the historical changes show a clear exponential growth trend, then using exponential regression to fit an exponential curve before extrapolation; finally, obtaining the lifetime prediction result, including the remaining effective number of cycles and the expected failure time, calculated based on the current testing frequency.

[0063] The process involves analyzing the combinations of over-limit flag bits to generate abnormal alarm commands that differentiate alarm levels and associate fault types. Specifically, this includes: analyzing the generated combinations of over-limit flag bits to generate abnormal alarm commands that differentiate alarm levels and associate fault types; first, identifying the fault type based on the bits that are 1 in the over-limit flag bits and their combination patterns: if only the response time exceeds the limit, the fault type is response lag; if only the leakage rate exceeds the limit, the fault type is seal failure; if only particulate matter exceeds the limit, the fault type is particulate matter exceeding the limit; if multiple indicators exceed the limit simultaneously, the fault type is comprehensive degradation; further, determining the alarm level based on the number of over-limit indicators and the magnitude of the deviation: if only one minor indicator slightly exceeds the limit, it is a minor alarm; if one critical indicator exceeds the limit or multiple minor indicators exceed the limit, it is a severe alarm; if the leakage rate or response time severely exceeds the limit and is accompanied by a surge in particulate matter, it is a critical alarm; each alarm command includes a timestamp, fault type, alarm level, and suggested handling measures, such as stopping the test and checking the seals, and is sent to the display interface and alarm indicator driver module in the form of a structured data frame.

[0064] The life prediction results, abnormal alarm commands, and pre-stored raw degradation data are summarized to form a comprehensive test report, resulting in a durability assessment. Specifically, this includes: summarizing and organizing the obtained life prediction results, generated abnormal alarm commands, and pre-stored raw degradation data in the local database, including complete response time series, pressure fluctuation series, leakage rate series, particulate matter concentration series, and ellipsoidal volume change records for each cycle; automatically filling in data according to a preset report template, including sections such as test overview, key performance indicator curves, failure determination details, life prediction curves, and alarm event list; exporting the report in both PDF and CSV formats to a designated folder, and simultaneously displaying the durability assessment results (including one of three states: qualified, warning, or failure) on the system's main interface; thus, the entire evaluation process for the multi-parameter collaborative durability test is completed.

[0065] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the system as described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0066] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the system as described above. All implementations in the above system embodiments are applicable to this embodiment and can achieve the same technical effects.

[0067] Experimental example: To ensure the validity of test results in ultra-clean environments such as atomic layer deposition (ALD), the system background particle concentration was calibrated according to the contamination source elimination mechanism before installing the diaphragm valve under test. Specifically, without installing the diaphragm valve, the drive positioning unit and cyclic excitation unit were fully operated to ensure all mechanical moving parts, such as the precision linear sliding stage and high-speed solenoid valve, were in normal working condition for 30 minutes. Simultaneously, the particulate matter analyzer was started to continuously collect the particulate matter concentration at the test pipeline outlet at a polling interval of 0.5 seconds, recording the number of particles with a diameter greater than or equal to 0.5 micrometers. The average background particle concentration was found to be 15 particles / ft³, with a standard deviation of 5 particles / ft³. This background baseline data was stored in the system database. In subsequent formal tests, all real-time collected raw particulate matter concentration data were automatically subtracted from this baseline to obtain corrected particulate matter concentration data. This corrected data was then used for subsequent multi-parameter fusion analysis and the construction of the constrained ellipsoid space.

[0068] The system parameters are configured as follows: the high-speed solenoid valve drive voltage is DC24V; the precision linear sliding stage positioning accuracy is ±0.01mm; the radial deviation resolution of the centering detection module is no greater than 0.01mm; the pilot gas path drive pressure is set to 0.45MPa; the excitation frequency range for cyclic testing is 0.5Hz to 5.0Hz; and the duty cycle is set to 50%. The sampling rates of the multi-parameter acquisition unit are 10kHz for the response time channel, 1kHz for the pressure channel, 100Hz for the leakage rate channel, and 50Hz for the particulate counting channel. The dimensionality reduction dimension of the principal component tensor projection of the constrained ellipsoidal space is 3; and the weight allocation of the attenuation correlation model is: leakage rate weight 0.50, response time weight 0.30, particulate concentration weight 0.15, and volume expansion rate weight 0.05. The failure thresholds are: a response time drift upper limit of 200 microseconds and a leakage rate upper limit of 1×10⁻⁶. -6 Standard cubic centimeters per second, maximum particulate matter concentration of 1000 particles per cubic foot, and maximum volume expansion rate of 0.5.

[0069] The experiment consisted of 500 complete data acquisition cycles, and after multi-level screening, 256 valid events were finally detected. During each cycle, parameter data of four dimensions were collected simultaneously: valve response time, pipeline pressure fluctuation, helium leakage rate, and particulate matter concentration. All data were aligned with a unified timestamp and stored in the original test dataset.

[0070] Step 1: Generating positioning reference signals for the driving positioning unit.

[0071] This step examines the process by which the drive positioning unit adjusts the physical gap between the high-speed solenoid valve and the drive end of the diaphragm valve under test according to preset assembly parameters. The preset target gap value in the experiment was 4.0 mm, with a centering deviation tolerance of no more than 0.20 mm radial offset and no more than 0.5° angular skew. After receiving the target gap value, the precision linear sliding stage begins to drive the high-speed solenoid valve to move, acquiring the current physical gap value in real time for closed-loop control. The sliding stage's movement speed is 0.5 mm / s.

[0072] When the current gap value falls within the target range of 4.0mm ± 0.01mm, the sliding stage locks its position and outputs a preliminary positioning completion signal. Subsequently, the alignment detection module is activated to measure the radial offset and angular skew between the solenoid valve's axis and the diaphragm valve's drive end axis. If the deviation exceeds the tolerance, the fine-tuning actuator is triggered for compensation, with a fine-tuning step of 0.02mm. After 1 to 3 fine-tunings, all deviation indicators meet the requirements, and an alignment confirmation signal is output.

[0073] The distribution of drive clearance values ​​over 500 cycles is as follows: clearance values ​​between 3.0 and 4.0 mm occurred 215 times, and clearance values ​​between 4.0 and 5.0 mm occurred 156 times, with the two accounting for approximately 74% in total; extreme cases with clearance values ​​less than 2.0 mm or greater than 6.0 mm occurred 24 times, accounting for approximately 4.8%. The distribution of radial alignment deviation is as follows: deviations between 0.05 and 0.10 mm occurred most frequently, 142 times, and deviations between 0.10 and 0.15 mm occurred 118 times, with the two accounting for approximately 52% in total; deviations exceeding 0.25 mm occurred 18 times, all of which were compensated for by fine-tuning the actuator.

[0074] Step 2: Generation of cyclic excitation unit and alternating on / off action sequence

[0075] This step examines the process by which the cyclic excitation unit generates periodic control commands based on the positioning reference signal and applies them to the diaphragm valve under test. After obtaining the locked drive distance parameter (4.0 mm) and the alignment confirmation flag, the system loads a preset cyclic test formula, which includes a target total number of cycles of 500, an operating frequency of 2.0 Hz, a duty cycle of 50%, and a test pressure of 0.45 MPa, forming a waveform parameter set.

[0076] The programmable logic controller (PLC) generates a raw pulse control sequence based on a waveform parameter set. Each pulse corresponds to one on / off command, and the pulse width is 250ms. The PLC sends the raw pulse control sequence to the power drive module of the high-speed solenoid valve, where it is amplified into a solenoid valve excitation signal with sufficient drive current. The amplified excitation pulse sequence controls the high-speed solenoid valve to alternately turn on and off at a frequency of 2.0Hz and a duty cycle of 50%. This drives the diaphragm valve under test to perform the corresponding opening and closing actions through the pilot air path, forming an alternating opening and closing action sequence.

[0077] During 500 cycles of execution, the excitation signal integrity detection statistics are as follows: 418 cycles (approximately 83.6%) had normal drive pulse generation and the amplified excitation pulse amplitude was stable within ±5% of the nominal value; 52 cycles (approximately 10.4%) had excitation pulse amplitude deviating from the nominal value by 5% to 10% due to temperature drift of the power drive module; and 30 cycles (approximately 6.0%) had excitation pulse amplitude deviating by more than 10% due to power fluctuations. The excitation pulse sequences of these cycles were marked as abnormal and recorded, but the test was not interrupted.

[0078] Step 3: Generation of multi-parameter acquisition unit and original test dataset

[0079] This step examines the process by which the multi-parameter acquisition unit receives the alternating opening and closing sequence and simultaneously monitors four-dimensional parameters. Using the alternating opening and closing sequence as the acquisition trigger, four monitoring channels—response time, pressure fluctuation, leakage rate, and particulate matter concentration—are activated to achieve synchronous acquisition triggering.

[0080] The response time acquisition channel measures the time difference from the issuance of the excitation pulse to the actual position of the diaphragm valve, based on the trigger edge of each opening and closing action. In 500 cycles, the initial baseline response time was 3.2 ms, exhibiting an exponential upward trend with the number of cycles: the average response time for the first 100 cycles was 3.4 ms, for cycles 101 to 250 it was 4.3 ms, for cycles 251 to 400 it was 5.5 ms, and for cycles 401 to 500 it was 7.1 ms. A response time exceeding the failure threshold of 6.0 ms was considered a response lag event.

[0081] The pressure fluctuation acquisition channel continuously records pipeline pressure values ​​at a sampling rate of 1000Hz. Each opening and closing action triggers a pressure waveform recording, and the pressure fluctuation amplitude is statistically analyzed. Before correction, the average pressure fluctuation amplitude of 500 cycles was 0.35MPa, with a fluctuation range of 0.15 to 0.55MPa. After correction by the pressure compensation algorithm, the average fluctuation amplitude was reduced to approximately 0.06MPa.

[0082] The helium leakage rate acquisition channel continuously reads the leakage rate value at a sampling rate of 100Hz and correlates the leakage rate value with the current number of actions. Over 500 cycles, the leakage rate increases from approximately 1×10⁻⁶ initially. -8 The Pa·m³ / s increases exponentially, approaching the failure threshold of 1×10⁻⁶ at approximately 430 cycles. -6 Pa·m³ / s. The particulate matter concentration acquisition channel recorded the exit particle count at a sampling rate of 50 Hz. The concentration increased from an initial of about 50 particles / ft³ to about 1650 particles / ft³, and also exceeded the failure threshold of 1000 particles / ft³ at about 430 cycles.

[0083] Step 4: The data fusion processing unit and the constrained ellipsoidal space characterizing the multidimensional performance degradation of the valve.

[0084] This step examines the process by which the data fusion processing unit parses, fits, and extracts feature parameter matrices from the original test dataset. First, it extracts four independent parameter time series arrays: response time series, pressure fluctuation series, leakage rate series, and particulate matter concentration series. Second derivative operations are performed on each parameter time series array, and the point with the largest curvature is identified as the performance degradation feature moment. Parameter values ​​near each feature moment are then extracted to construct a multidimensional parameter feature point cloud.

[0085] Each point in the multidimensional parametric feature point cloud is spatially calibrated according to three physical structure mapping dimensions: the valve core diaphragm micro-variation zone, the pilot air passage inlet section, and the sealing lip contact zone. This yields the three-dimensional spatial coordinates of each feature point in the three-dimensional physical structure coordinate system, forming a structured feature point set. Real-time parameter extreme value coordinates on each physical structure dimension are extracted from the structured feature point set as sampling features. Using the extreme value points in each sampling feature as boundary constraint points, a minimum volume closed surface containing all extreme value points is fitted in the three-dimensional physical structure mapping space to obtain the constrained ellipsoid space characterizing the multidimensional performance degradation state of the valve. The major axis of the ellipsoid corresponds to the valve core diaphragm micro-variation zone dimension (nominal length 4.0 mm), the middle axis corresponds to the pilot air passage inlet section dimension (nominal length 3.0 mm), and the minor axis corresponds to the sealing lip contact zone dimension (nominal length 2.5 mm).

[0086] Principal component tensor projection dimensionality reduction was performed on the constrained ellipsoid space characterizing the multidimensional performance degradation of the valve, and the gradient drift and dissipative volume expansion rate of the tensor trace in the degradation-sensitive direction were extracted. In 500 cycles, the ellipsoid volume gradually shrank from an initial of about 113 mm³ to about 85 mm³, with a shrinkage rate of about 24.7%. The gradient drift and dissipative volume expansion rate were input into a preset multidimensional degradation correlation model to obtain the feature parameter matrix characterizing the performance degradation trajectory of the diaphragm valve.

[0087] like Figure 3 The figure shows a three-dimensional structural simulation of the constrained ellipsoidal space characterizing the multidimensional performance degradation of the valve. The three coordinate axes in the figure correspond to three physical structural dimensions and their nominal lengths, namely: The dimensions of the valve core diaphragm (corresponding to the lower right axis of the image), the dimensions of the pilot airway inlet section (corresponding to the pilot airway dimensions in the image), and the dimensions of the sealing lip contact zone (corresponding to the vertical upward axis of the image). The three dots are red, yellow, and blue-green. Figure 3 These special points on the ellipsoid correspond to the real-time parameter extremum coordinates extracted from three physical dimensions: the valve core diaphragm micro-variation zone, the pilot air inlet section, and the sealing lip contact zone. The blue ellipsoid in the figure is the fitted minimum volume closed surface (constrained ellipsoidal space), which intuitively defines the safe / operational boundary of the valve's performance degradation in the current state. The red spiral / winding curve is a time-series tracking of the degradation process.

[0088] like Figure 4The figure shows the trajectory of the shrinkage evolution of the ellipsoid volume with the number of cycles. The three axes in the figure correspond to the three physical structure mapping dimensions of the valve core diaphragm micro-variation zone, the pilot air inlet section, and the sealing lip contact zone, respectively. The Z-axis in the figure, i.e., the ellipsoid volume-normalized, is used as the main perspective and directly shows the shrinkage process of the ellipsoid volume with the increase of the number of cycles. This is the most core quantitative indicator of valve performance degradation. The X-axis is the normalized X and the Y-axis is the normalized Y. These two axes are the mapping projections of the three physical dimensions (valve core diaphragm, pilot air inlet, and sealing lip) in the degradation feature space. They do not directly represent physical dimensions, but rather capture the drift and coupling relationship of characteristic parameters in the degradation process. The gradient of red, yellow, and green in the diagram (representing time / cycle number) indicates that the red dotted area (upper left / high position) represents the early cycle, when the ellipsoid volume is at its maximum (around 120 on the Z-axis), indicating that the valve is in a relatively healthy state with sufficient performance margin. The green dotted area (lower right / lower position) represents the late cycle, when the ellipsoid volume has shrunk to a minimum, indicating that the valve performance has severely deteriorated and is approaching or reaching the failure threshold. The trajectory is not a straight downward sloping line, but a winding and undulating path in space, which shows that the degradation of the valve in multiple dimensions (seal, diaphragm, air passage) does not occur synchronously, but is a nonlinear and complex evolutionary process in which various physical characteristics are coupled with each other.

[0089] Step 5: Durability assessment unit and life prediction results

[0090] The decay rate and coupling coefficient of each parameter dimension were extracted from the feature parameter matrix: the response time decay rate is approximately 8.1 microseconds / 100 cycles, and the leakage rate increase rate is approximately 2.2 × 10⁻⁶. -7 The particulate matter concentration growth rate is approximately 35.0 particles / ft³ / hundred cycles, and the volume expansion rate growth rate is approximately 0.09 / hundred cycles. The coupling coefficient between response time and leakage rate is approximately 0.78, and the coupling coefficient between pressure fluctuation and particulate matter is approximately 0.65.

[0091] Each indicator was compared with a preset durability failure threshold: response time drift upper limit 200 microseconds, leakage rate upper limit 1×10 -6 The maximum values ​​for Pa·m³ / s, particulate matter concentration (maximum 1000 particles / ft³), and volume expansion rate (maximum 0.5) were set. In 500 cycles, at least one indicator exceeded the limit in 298 cycles, accounting for approximately 59.6%. Of these, 89 cycles exceeded the limit only for response time, 62 only for leakage rate, 48 only for particulate matter, and 99 cycles exceeded the limit simultaneously for multiple indicators.

[0092] The alarm level is determined based on the combination of the over-limit flags: a slight over-limit of a single primary indicator is considered a minor alarm, totaling 87 times; an over-limit of any critical indicator (leakage rate or response time) or multiple over-limits of secondary indicators is considered a serious alarm, totaling 156 times; a serious over-limit of a critical indicator accompanied by a sharp increase in particulate matter is considered a critical alarm, totaling 55 times.

[0093] For critical and emergency alarm cycles, the remaining effective cycle count is predicted based on the number of completed cycles and historical trends. When a key indicator exceeds its limit, historical trend data for each indicator from the last 1000 cycles is read, and a linear extrapolation method is used: least-squares linear fitting is performed with the cycle count as the x-axis and the indicator value as the y-axis, extrapolating to the failure threshold to obtain the predicted failure point. In 500 cycles, the predicted remaining lifetime gradually decreases from approximately 500 cycles initially to approximately 70 cycles. When the predicted remaining lifetime falls below 50 cycles, the system generates a failure warning signal.

[0094] Figure 5 The diagram illustrates the simulation process of the decay trajectory of the multidimensional decay correlation model in a three-dimensional feature space. The cyan dot (initial state), located in the upper left corner of the diagram, represents the first cycle of valve operation, at which point all indicators are at healthy baseline values ​​and no significant decay has occurred. The failure threshold, located at the end of the trajectory, indicates that after hundreds of cycles, at least one key indicator of the valve has reached the failure threshold. The continuous colored trajectory from the cyan dot to the red cross in the diagram represents the evolution path of the multidimensional decay correlation model. The trajectory is not a straight diagonal line, but is full of zigzag turns and complex entanglements, which confirms the existence of the coupling coefficient. This means that when one indicator deteriorates, it will pull another indicator to deteriorate together, making its degradation path particularly chaotic and complex in three-dimensional space. The color also changes gradually, with the color change representing the progression of the number of cycles / time, transitioning from cyan (early stage, relatively healthy) to red-orange (late stage, near failure). The end of the trajectory becomes extremely dense, and the color turns red and yellow, indicating that a severe alarm will be triggered when near failure. At this stage, the predicted remaining lifespan will decrease sharply.

[0095] A total of 500 complete cycles of durability multi-parameter collaborative testing were conducted on the diaphragm valve. The evolution data of four-dimensional performance parameters could be obtained in real time without disassembling the valve. The experimental results show that the drive positioning unit achieved an average radial centering deviation accuracy of about 0.06 mm in 500 cycles, meeting the centering tolerance requirement of ±0.20 mm, and the stability of the positioning reference signal is good. The alternating opening and closing action sequence generated by the cyclic excitation unit drove the diaphragm valve to perform a complete 500 opening and closing actions, and the excitation pulse amplitude remained stable within the nominal value ±5% in 83.6% of the cycles.

[0096] The multi-parameter collaborative acquisition unit realizes synchronous triggering and data fusion of four monitoring channels. The original test dataset contains a total of approximately 2.0 × 10⁻⁶ four-dimensional parameter time series. 6 The data fusion processing unit successfully constructed a constrained ellipsoidal space characterizing the multidimensional performance degradation of the valve. The ellipsoidal volume shrinkage rate was approximately 24.7%, effectively reflecting the performance degradation trend of the diaphragm valve throughout its entire life cycle. Combined with the background particle concentration correction mechanism, the particulate matter concentration data eliminated interference from the test system itself, making the failure determination more accurate. At the same time, the durability determination unit output a complete comprehensive test report, including a test overview, key performance index curves, failure determination details, life prediction curves, and an alarm event list, with a prediction accuracy within ±8%.

[0097] Based on the above experimental design and verification, the following conclusions can be drawn: The drive positioning unit achieves a positioning accuracy of approximately 0.06 mm in average radial centering deviation in 500 cycles through closed-loop collaborative control of a precision linear sliding stage and an alignment detection module, meeting the preset tolerance requirements and providing a reliable physical reference for subsequent cyclic excitation.

[0098] The multi-parameter collaborative acquisition unit enables simultaneous monitoring of four dimensions: response time, pipeline pressure, helium leakage rate, and particulate matter concentration. After correction, the pressure fluctuation amplitude decreased from 0.35 MPa to 0.06 MPa, and the rate of increase of leakage rate and particulate matter concentration decreased by approximately 55% and 68%, respectively, effectively demonstrating the role of collaborative calibration in improving the quality of multi-parameter data.

[0099] The data fusion processing unit successfully extracted two core degradation features, gradient drift and dissipative volume expansion rate, based on the principal component tensor projection dimensionality reduction method of the constrained ellipsoid space that characterizes the multidimensional performance degradation of valves. The ellipsoid volume shrinkage rate was about 24.7%, which is in good agreement with the actual failure process.

[0100] The weight allocation scheme of the multidimensional attenuation correlation model (leakage rate 0.50, response time 0.30, particulate matter 0.15, volume expansion rate 0.05) can effectively distinguish between the over-limit of key indicators and the fluctuation of secondary indicators, with an alarm accuracy of 87.3% and a false alarm rate controlled within 5%.

[0101] The remaining life prediction based on the linear extrapolation method has an error within ±8% throughout 500 cycles. When the predicted remaining life is less than 50 cycles, the system can issue a failure warning signal in a timely manner, realizing early warning of diaphragm valve durability failure.

[0102] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A multi-parameter collaborative testing system for the durability of a diaphragm valve, characterized in that, include: The drive positioning unit is used to adjust the physical gap between the high-speed solenoid valve and the drive end of the diaphragm valve under test according to the preset assembly parameters, so as to obtain a positioning reference signal containing the real-time drive distance and the centering status. The cyclic excitation unit is used to generate periodic control commands based on the positioning reference signal to obtain an alternating opening and closing action sequence acting on the diaphragm valve under test. The multi-parameter acquisition unit is used to receive the alternating opening and closing action sequence, and simultaneously monitor the valve response time, pipeline pressure fluctuation, helium leakage rate and particulate matter concentration triggered by this sequence, forming a raw test dataset containing multi-parameter time series. The data fusion processing unit is used to analyze the temporal dimension and physical structure mapping dimension of the original test dataset, dynamically lock the multidimensional parameter evolution trajectory sampling features, and fit and construct a constrained ellipsoid space characterizing the multidimensional performance degradation state of the valve based on the multidimensional parameter evolution trajectory sampling features. Principal component tensor projection dimensionality reduction is performed on the constrained ellipsoid space to extract the gradient drift and dissipation volume expansion rate of the degradation-sensitive direction tensor trace. The gradient drift and dissipation volume expansion rate are then input into a preset multidimensional decay correlation model to obtain the feature parameter matrix characterizing the performance degradation trajectory of the diaphragm valve. The durability assessment unit is used to match and analyze the feature parameter matrix with a preset durability failure threshold to obtain a durability assessment result that includes life prediction results, abnormal alarm instructions and comprehensive test reports.

2. The multi-parameter collaborative testing system for the durability of a diaphragm valve according to claim 1, characterized in that, Adjusting the physical gap between the high-speed solenoid valve and the drive end of the diaphragm valve under test according to preset assembly parameters, a positioning reference signal containing real-time drive distance and alignment status is obtained, including: Read the preset assembly parameters, parse out the target gap value and centering deviation tolerance, and obtain the gap adjustment target; According to the gap adjustment target, the precision linear sliding stage is driven to move the high-speed solenoid valve. At the same time, the current physical gap value is collected in real time, and closed-loop control is implemented until the current gap value falls within the error range of the target gap value. Then the position of the sliding stage is locked, and a preliminary positioning completion signal is obtained. After receiving the initial positioning completion signal, the centering detection module is activated to measure the radial offset and angular deviation between the centerline of the high-speed solenoid valve and the centerline of the drive end of the diaphragm valve under test, and to obtain the centering status data. The centering status data is compared with the centering deviation tolerance. If the tolerance is exceeded, the fine-tuning actuator is triggered to compensate until all deviation indicators meet the tolerance requirements, and a centering confirmation signal is output. Based on the locked sliding stage position, the current physical gap value, and the alignment confirmation signal, a positioning reference signal containing the real-time driving distance and alignment status is formed, and the positioning reference signal is transmitted to the cyclic excitation unit.

3. The multi-parameter collaborative testing system for the durability of a diaphragm valve according to claim 2, characterized in that, Based on the positioning reference signal, periodic control commands are generated to obtain an alternating opening and closing sequence acting on the diaphragm valve under test, including: The locked drive distance parameters and alignment status confirmation flag are parsed from the positioning reference signal to obtain the excitation parameter initialization conditions; Based on the initialization conditions of the excitation parameters, a preset cyclic test formula is loaded, which includes the target total number of cycles, action frequency, duty cycle and test pressure value, forming a waveform parameter set; The waveform parameter set is input into the programmable logic controller (PLC), which generates a corresponding high-speed solenoid valve drive pulse sequence. Each pulse of the drive pulse sequence corresponds to an opening or closing command of the diaphragm valve under test, thus obtaining the original pulse control sequence. The original pulse control sequence is sent to the power drive module of the high-speed solenoid valve, which amplifies it into a solenoid valve excitation signal with sufficient drive current, thus obtaining the amplified excitation pulse sequence. Based on the amplified excitation pulse sequence, the high-speed solenoid valve is controlled to alternately turn on and off according to the set frequency and duty cycle. The pilot air path drives the diaphragm valve under test to perform the corresponding opening and closing actions, forming an alternating opening and closing action sequence acting on the diaphragm valve under test.

4. The multi-parameter collaborative testing system for the durability of a diaphragm valve according to claim 3, characterized in that, The alternating opening and closing sequence is received, and the valve response time, pipeline pressure fluctuation, helium leakage rate, and particulate matter concentration triggered by this sequence are monitored simultaneously to form a raw test dataset containing multi-parameter time series, including: Using the alternating opening and closing action sequence as a trigger reference, the monitoring channels for multi-parameter acquisition are activated to obtain a synchronous acquisition trigger signal; Based on the collected synchronous trigger signal, the valve response time corresponding to each alternating opening and closing action is captured in real time to obtain response time sequence data containing timestamps; While capturing the response time, the pipeline pressure value is continuously collected, and the pressure waveform during each opening and closing action is recorded to obtain pressure fluctuation sequence data including timestamps. While collecting pressure data, the helium leakage rate is continuously read, and the leakage rate value is associated with the current number of actions and timestamp to obtain leakage rate sequence data including timestamp. While collecting the leakage rate, the particle count at the gas outlet is continuously read to obtain particulate matter concentration sequence data including timestamps; The response time series data, pressure fluctuation series data, leakage rate series data, and particulate matter concentration series data are aligned and merged in multiple dimensions according to the same timestamp to form an original test dataset containing multiple parameter time series.

5. The multi-parameter collaborative testing system for the durability of a diaphragm valve according to claim 4, characterized in that, Based on the sampling features of the multidimensional parametric evolution trajectory, a constrained ellipsoidal space characterizing the multidimensional performance degradation state of the valve is constructed. The sampling features of the multidimensional parametric evolution trajectory correspond to the real-time parametric extreme coordinates of the valve core diaphragm micro-variation zone, the pilot air passage inlet section, and the sealing lip contact zone, including: The response time series, pressure fluctuation series, leakage rate series, and particulate matter concentration series after co-correction were extracted from the original test dataset to obtain four independent parameter time series arrays. Perform second derivative operations on each parameter time series array to locate the point with the largest curvature, take the time position corresponding to the point as the performance degradation feature moment, and extract the parameter values ​​near each feature moment to form a multi-dimensional parameter feature point cloud. Each point in the multidimensional parameter feature point cloud is spatially calibrated according to the three physical structure mapping dimensions of the valve core diaphragm micro-variation zone, the pilot air passage inlet section, and the sealing lip contact zone, so as to obtain the three-dimensional spatial coordinates of each feature point in the physical structure coordinate system and form a structured feature point set. The real-time extreme value coordinates of the parameters in each physical structure dimension are extracted from the set of structured feature points and used as sampling features of the multidimensional parameter evolution trajectory to obtain a set of sampling features; Using the extreme points in each sampled feature as boundary constraint points, a minimum volume closed surface containing all extreme points is fitted in the three-dimensional physical structure mapping space to obtain a constrained ellipsoid space characterizing the multidimensional performance degradation state of the valve.

6. The multi-parameter collaborative testing system for the durability of a diaphragm valve according to claim 5, characterized in that, The gradient drift and dissipation volume expansion rate of the extracted degradation-sensitive direction tensor trace are then input into a preset multidimensional decay correlation model to obtain a feature parameter matrix characterizing the performance degradation trajectory of the diaphragm valve, including: The constrained ellipsoid space is represented as a covariance matrix with the semi-axis lengths of each principal axis as characteristic parameters, thus obtaining the geometric description of the ellipsoid in the high-dimensional parametric space. Principal component tensor projection dimensionality reduction is performed on the covariance matrix, the three principal axis directions with the largest eigenvalues ​​are extracted, and the ellipsoid is projected along the principal axis directions to a low-dimensional subspace to obtain the dimensionality-reduced principal component tensor set. On the dimensionality-reduced principal component tensor set, the rate of change in each principal axis direction is calculated, the direction with the largest rate of change is identified as the degradation-sensitive direction, and the change sequence of the tensor trace in the direction with the number of iterations is extracted to obtain the gradient drift of the tensor trace in the degradation-sensitive direction. Calculate the relative rate of change of the volume of the constrained ellipsoidal space with the number of cycles, subtract one from the ratio of the current volume to the initial volume and take the absolute value to obtain the dissipative volume expansion rate; The gradient drift and dissipation volume expansion rate are used as input variables and input into a preset multidimensional decay correlation model. The model adopts a multivariate nonlinear regression or neural network structure to obtain a feature parameter matrix that shows the decay rate and coupling coefficient of each parameter dimension of the current valve performance degradation.

7. The multi-parameter collaborative testing system for the durability of a diaphragm valve according to claim 6, characterized in that, The feature parameter matrix is ​​matched and analyzed with a preset durability failure threshold to obtain a durability evaluation result that includes life prediction results, abnormal alarm commands, and a comprehensive test report, including: The decay rate and coupling coefficient of each parameter dimension are extracted from the feature parameter matrix to obtain a set of performance degradation quantification indicators. The set of performance degradation quantification indicators is compared with the preset durability failure threshold item by item to obtain the deviation of each indicator from the threshold and the over-limit flag bit. Based on the deviation of each indicator and the over-limit flag, the critical failure state is determined. If any key indicator exceeds the limit or the comprehensive weighted value exceeds the alarm threshold, a failure warning signal is generated. After receiving the failure warning signal, the remaining effective number of cycles is predicted based on the current number of cycles and the historical change trend obtained in advance, and the lifetime prediction result is obtained. Analyze the combination of over-limit flag bits to generate abnormal alarm commands that distinguish alarm levels and associate fault types; The life prediction results, abnormal alarm commands, and pre-stored raw degradation data are summarized to form a comprehensive test report, and the durability assessment results are obtained.

8. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to perform the system as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, performs the system as described in any one of claims 1 to 7.

Images