System and method for solenoid valve optimization and response degradation measurement

Abstract

A system and method for detecting faults and optimizing power usage of a solenoid valve. The method includes obtaining current signatures of a solenoid coil, detecting various signatures using a dedicated circuit, and optimizing power output of the system using a pulse width modulation controller. Furthermore, by using machine learning, data from the dedicated circuit can be used to optimize the system.

Description

[0001] Cross Reference to Related Applications

[0002] This application claims the benefit of Indian Patent Application No. 201911051558, filed December 12, 2019, the disclosure of which is incorporated by reference herein in its entirety. BACKGROUND

[0003] Many fluid power systems, such as hydraulic systems, include valves that regulate fluid flow. There are various types of valves used for different purposes, such as directional control, pressure control, on / off flow control, and proportional flow control. Valves are often incorporated in machines used for various industrial and mobile applications, including injection molding machines, high-pressure processing machines, lathes, and mobile machines. The number of valves used in a given machine can vary widely.

[0004] Some fluid power systems include spool valves. Spool valves include a regulating member in the form of a spool that moves linearly within a bore or passage defined by a valve body. The spool can include one or more shoulders that control fluid communication between ports defined by the valve body based on the linear position of the spool. In some systems, the regulating member is driven by a solenoid linear actuator. It is not uncommon for a single system to contain as many as 50 or more valves.

[0005] In example systems, multiple valves are connected in series or parallel combinations. Failure of even a single valve can impede proper operation of the entire system. Valve failure due to spool failure can result in problems such as insufficient pressure or insufficient expected cylinder displacement. Two common types of spool failure include the spool becoming completely stuck (i.e., the spool does not move), or the spool moving less or being limited. Some common causes of spool failure are fluid contamination or part wear.

[0006] Valve failure can result in many problems that require time and money to repair. If spool failure can be detected and located, then valve failure due to spool failure can be avoided. SUMMARY

[0007] In general, the present disclosure relates to systems and methods that provide more cost-effective and / or otherwise improved solenoid valve operation. Certain aspects relate to systems and methods that provide enhanced solenoid valve diagnostics (e.g., failure detection). Other aspects relate to systems and methods that control valve power consumption to allow solenoid valves to operate more efficiently.

[0008] One example is a solenoid operated valve comprising: at least one coil and at least one adjustment member; a controller interfaced with a current meter to monitor a current signature of the coil when actuating the solenoid operated valve by operating the solenoid operated valve in an actuation mode in which a first power level is used to drive current through the coil, and the controller includes a processor and a memory in electronic communication with the processor for executing an adjustment member power optimization algorithm operable to: detect when the adjustment member has begun to displace based on sensed current of the current signature sensed by the current meter; detect when the adjustment member has reached a final position based on sensed current of the current signature sensed by the current meter; and displace the solenoid operated valve from the actuation mode to a hold mode once the adjustment member has been determined to be in the final position. A second power level is used to drive current through the coil when the solenoid operated valve is operated in the hold mode, and the second power level is lower than the first power level. The second power level of the hold mode can be controlled by a pulse width modulation controller. In other examples, the controller includes an integrated circuit having a solenoid coil. The controller can detect when the adjustment member has begun to displace by detecting when the current transitions from a positive slope to a negative slope. The controller can then detect that the adjustment member has reached its final position by detecting that the current has transitioned from a positive slope to a negative slope and then back to a positive slope. The controller can use a first latch that is set to a high output when the system detects a negative slope, and a second latch that the controller uses when the controller detects a positive slope after the output state of the first latch has been set to high. Once both the first latch and the second latch are set to high, the controller transitions the current to a hold state.

[0009] A different example solenoid operated valve includes at least one coil and at least one adjustment member, and a controller interfaced with a current meter to monitor a current signature of the coil when actuating the solenoid operated valve, and the controller monitors measurement data related to the current signature from the current meter, the measurement data including measured operational values including: a time to reach a first peak value of current, a time to reach a first valley value of current, a time to reach a maximum current output, a ratio of the time to reach the first valley value to the time to reach the first peak value, and the controller compares the measured operational values to baseline operational values stored in a memory to monitor a health of the solenoid operated valve.

[0010] A method for reducing unexpected downtime of a solenoid operated spool valve is disclosed, the method comprising: determining a response time of a spool of the valve; determining a position of the spool of the valve; calculating a spool response time error value; calculating a spool position error value; comparing one or both of the spool response time error value and the spool position error value to a threshold value; and generating an error signal when one or both of the spool response time error value and the spool position error value exceeds the threshold value.

[0011] In some examples, the step of determining the response time of the valve comprises calculating the response time based on one or more of: a time to reach a first peak current;

[0012] a time to reach a last valley current; a time to reach 90% of a maximum current; a number of low points; and a minimum point of proximity to zero of an ideal current profile.

[0013] In some examples, the step of calculating the response time is performed using a regression model.

[0014] In some examples, calculating the spool response time error value comprises comparing the valve response time to a baseline response time.

[0015] In some examples, the baseline response time is determined during a training of the spool valve.

[0016] In some examples, the spool response time error is calculated as a percentage change relative to the baseline response time.

[0017] In some examples, the step of determining the position of the spool of the valve comprises calculating the response time based on one or more of: a difference in current at a first valley and a steady state current; a Euclidean distance between a reference sticking curve and a latest recorded current profile; a time to reach a first peak current; a time to reach a last valley current; a time to reach 90% of a maximum current; and a ratio of a square of a current at a first valley and a current at the first peak.

[0018] In some examples, the step of calculating the position is performed using a regression model.

[0019] In some examples, calculating the position error value comprises comparing the valve position to a baseline response time.

[0020] In some examples, the baseline response time is determined during a training of the spool valve.

[0021] In some examples, the spool response time error is calculated as a percentage change relative to the baseline response time.

[0022] In the following description, various additional aspects will be described. These aspects may relate to individual features as well as combinations of features. It will be understood that both the foregoing general description and the following detailed description are merely exemplary and explanatory and do not limit the broad inventive concepts on which the examples disclosed herein are based. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Several aspects of the present disclosure are illustrated in the accompanying drawings, which are incorporated in and constitute a part of this specification. A brief description of the drawings is as follows:

[0024] Figure 1 is a schematic representation of an example system including a solenoid valve having features according to the present disclosure.

[0025] Figure 2 This is a conventional solenoid valve using a linear variable differential transformer.

[0026] Figure 2A yes Figure 2 A cross section of a prior art solenoid valve is shown.

[0027] Figure 3 is Figure 1 Example current signature produced by a solenoid valve.

[0028] Figure 4 is a graph illustrating current characteristics and also depicting where pulse width modulation control can be implemented to modify the current characteristics.

[0029] Figure 5 Describes the Figure 1 The solenoid coil of the solenoid valve is wired to the valve core shift detection circuit system.

[0030] Figure 6 Describes the Figure 1 Another spool shift detection arrangement of the solenoid valve's solenoid coil wiring.

[0031] Figure 7 It shows Figure 6 A more detailed schematic diagram of the spool shift detection arrangement.

[0032] Figure 8 Is about Figure 7 A flow chart showing details of how the spool detection circuit arrangement operates.

[0033] Figure 9 is a graph of the current data from a healthy spool.

[0034] Figures 10A-10Cis a graph of the current profile of a solenoid valve showing the effect of different supply voltages, the tested supply voltages being 28.8V, 24V and 19.2V.

[0035] Figures 11A-11C is a graph showing how the current profile of a solenoid valve can vary with temperature. The tests were conducted at 0C, 25C and 55C respectively.

[0036] Figure 12 is a graph showing how the current profile of a solenoid valve varies based on the viscosity of the fluid flowing through the solenoid valve.

[0037] Figure 13 is a graph showing how the current profile of a solenoid valve varies based on the level of contamination of the fluid flowing through the valve.

[0038] Figure 14 is a flowchart showing how a linear regression method can be used to train a solenoid valve.

[0039] Figure 15 is a flowchart showing how a solenoid valve can use information from a linear regression to determine a valve core response time. Figure 14

[0040] Figure 16A and Figure 16B are graphs comparing a healthy valve current profile directly to a valve with a lower voltage to simulate a late spool (as shown in Figure 16A ) or a higher viscosity oil to simulate a late spool (as shown in Figure 16B ).

[0041] Figure 17 is a schematic example current profile produced by a solenoid valve of Figure 1 superimposed on an ideal profile line.

[0042] Figure 17A shows a schematic example of Figure 17 with annotations showing how to determine an ideal profile line.

[0043] Figure 18 is a schematic regression logic model to calculate an ideal spool response time.

[0044] Figure 19 is an example flowchart with process details about how a spool core detection circuit arrangement disclosed herein can operate to detect spool core response time degradation.

[0045] Figure 20 is an example flowchart with process details about real-time assessment of spool core response time degradation identified through the process shown in Figure 19 . ​

[0046] Figure 21 is an exemplary flowchart with process details regarding how the valve trim detection circuit arrangement disclosed herein can operate to detect valve trim position degradation when a pre-trained model is available.

[0047] Figure 22 is an exemplary flowchart with process details regarding real-time assessment of valve trim position degradation identified by the process shown in Figure 20

[0048] Figure 23 is an exemplary flowchart with process details regarding real-time assessment of valve trim position degradation using a regression model obtained from an online learning phase. DETAILED DESCRIPTION

[0049] Various examples will be described in detail with reference to the drawings, wherein like reference numerals represent like parts and assemblies throughout the several views. Further, any examples set forth in this specification are not meant to be limiting and are merely set forth for illustrative purposes in accordance with the principles of the present disclosure. As such, the examples described herein are merely representative of the many possible examples that could be contemplated in accordance with the principles of the present disclosure.

[0050] Reference throughout this specification to "one embodiment", "an embodiment", "certain embodiments", "certain implementations" or other similar terms means that a described embodiment might include a particular feature, structure, or characteristic, but each embodiment can not necessarily include the particular feature, structure, or characteristic. Furthermore, these terms can not necessarily refer to the same embodiment. Additionally, when a particular feature, structure, or characteristic is described in connection with an embodiment, it is submitted that it is within the knowledge of those skilled in the art to effect such feature, structure, or characteristic in connection with other embodiments whether or not explicitly described.

[0051] In the drawings, some of the structural and / or methodological features can be shown in particular arrangements and / or orders. It should be appreciated, however, that such specific arrangements and / or orders can not be required. Instead, in some embodiments, such features can be arranged in a different manner and / or order than shown in the illustrative drawings. Further, inclusion of a structural and / or methodological feature in a particular drawing is not meant to imply that such feature is required in all embodiments, and in some embodiments, such feature can not be included or can be combined with other features.

[0052] ​When valves in mechanical systems degrade or wear, the position of the flow or pressure regulating components of these valves (such as the spool of a solenoid-operated valve) may deviate from the position expected by a given operating command on the system, resulting in, for example, excessive or insufficient flow, an undesirable pressure differential across the valve, etc. Such abnormal spool operation may be caused by fluid contamination or wear of valve components. Therefore, it is beneficial to detect such deviations during system operation so that the command input can be adjusted to achieve the desired flow / pressure and to prevent system failures and their consequences, such as mechanical / equipment failure.

[0053] Certain aspects of the present disclosure relate to monitoring a spool of a solenoid-operated valve. In some embodiments, and by way of non-limiting example, systems and methods for monitoring a spool of a solenoid-operated valve include a solenoid coil and a spool. In some embodiments, these systems and methods incorporate an ammeter and a processor and memory in electronic communication with the processor for executing a spool fault detection algorithm.

[0054] In a typical hydraulic spool valve assembly, a valve is directly coupled to the Figure 2 and Figure 2A The linear variable differential transformer (LVDT) 12 of the spool 2 shown in the figure is used to detect the spool position. The LVDT 12 is coupled adjacent to the valve body 4, which is adjacent to the solenoid 6 of the spool 2. However, LVDTs are expensive and can deteriorate over time due to the effects of high-pressure hydraulic fluid in the flow channel in which they are located. The systems and methods of monitoring the spool of a solenoid-operated valve of the present disclosure can be implemented without directly measuring the spool position or requiring additional sensors (such as LVDTs), but instead the spool position is inferred from measurements of other measured parameters of the valve (e.g., measured solenoid coil current). In addition, the disclosed systems and methods can be performed in "real time" where detections can be reported nearly instantly and in parallel, allowing continuous monitoring with little to no noticeable delay between detection and reporting of results.

[0055] Figure 1A mechanical system 10 is shown that is operated at least in part by using a hydraulic device. The hydraulic device includes a non-limiting embodiment of a valve assembly 100 for illustrating the principles of the present disclosure. In some examples, the valve assembly 100 is an on-off valve. The valve assembly 100 includes a housing 103 (e.g., a valve block or valve body) that houses a valve core 112, which is mounted in a valve core bore 114 defined by the housing 103. In this example, the spool valve is a three-way spool valve. However, the principles of the present disclosure are readily applicable to other spool valves (e.g., two-way spool valves) and other fluid control valves, such as flow control valves having on / off or poppet valves. The spool 112 includes a shaft 126 coupled to a pair of metering lands 122 and 124 at opposite ends of the shaft 126, each of which provides a flow regulation function for the valve assembly 100. The solenoid linear actuator 130 is coupled to the valve spool 112 and is adapted to drive axial linear movement of the valve spool 112 within the valve spool bore 114, the linear movement being along the central axis A of the valve spool bore 114. The solenoid linear actuator 130 houses a coil 132 for generating a controlled magnetic field by applying a control signal or command signal that generates a current in the coil 132.

[0056] A fluid supply source 101 (e.g., a pump) supplies hydraulic fluid to a work port 104 via a supply line 102 through a supply port 105. The work port 104 is connected to a hydraulic cylinder 106, which drives a load, i.e., a hydraulic device or machine. The fluid from the work port is discharged into a tank 108 via a tank port 107 and a tank line 110.

[0057] The control unit 170 is configured to provide control signals or command signals that generate current in the coils 132 to drive the valve spool 112 to move linearly along the axial direction of the axis A. The control unit also includes a current meter 173 (e.g., an ammeter) that is suitable for measuring the current in the one or more coils 132 of the solenoid linear actuator 130. The valve spool 112 is in a closed position (e.g., a closed position in which flow is blocked, such as Figure 1 The spool 112 is configured to move between a first open position and a second open position (e.g., an open position that allows flow through the valve body). In the closed position, the spool 112 blocks fluid communication between the supply port 105, the work port 104, and the tank port 107. In the first open position, the spool 112 is moved to open fluid communication between the supply port 105 and the work port 104, allowing pressurized hydraulic fluid from the supply port 105 to flow into the hydraulic cylinder 106 to drive the hydraulic cylinder 106 from the first rod position to the second rod position. In the second open position, the spool 112 is moved to open fluid communication between the work port 104 and the tank port 107, allowing hydraulic fluid from the hydraulic cylinder 106 to flow to the tank, allowing the hydraulic cylinder 106 to move from the second rod position back to the first rod position.

[0058] Measurements from the galvanometer 173 are fed to an operational subsystem 174 of the mechanical system 10, which is operatively coupled to the control unit 170. The operational subsystem 174 includes one or more processors 180 adapted to execute computer-readable instructions and process signals received from the control unit. The operational subsystem 174 also includes a memory 178 storing computer-readable instructions and a command interface 176, both of which are operatively coupled to the one or more processors 180.

[0059] When the solenoid linear actuator 130 receives current to drive the spool 112 to axially linearly move within the spool bore along the axis A, a portion 113 of the spool 112 (e.g., a ferromagnetic portion, an armature portion, etc.) or a portion of the spool assembly that includes the spool 112 and is fixedly coupled to the spool 112 moves relative to the one or more coils of the solenoid linear actuator 130, causing a change in magnetic flux through the one or more coils 132, which likewise generates an inductance in the one or more coils. The inductance generated in the coil due to the interaction of these magnetic fields with the spool 112 or the portion 113 causes a change in current in the one or more coils 132. The current in the one or more coils 132 is different depending on whether the spool 112 has actually moved, has not moved, or has completed movement. The current in the one or more coils 132 can be measured by the galvanometer 173 as a function of time. Such measurements of the current in the one or more coils 132 can be visualized as a plot of the change in coil current over time over a period of time and is referred to as a “current signature.” One current signature can correspond to movement of the spool 112 from the closed position to the first open position, and another current signature can correspond to movement of the spool 112 from the closed position to the second open position.

[0060] The control unit 170 can generate a spool failure condition to indicate whether the spool 112 is moving normally in response to a control signal or command signal, e.g., as expected and intended. In the case that the spool 112 is moving normally in response to a control signal or command signal through its full stroke length (e.g., the full length between one of the closed and open positions), the control unit can indicate a negative spool failure condition, i.e., no spool failure. In the case that the spool 112 is not moving normally in response to a control signal or command signal, the control unit 170 can indicate a positive spool failure condition, i.e., there is a spool failure. When the spool 112 is not moving normally, it can be partially moving in response to a control signal or command signal through its intended stroke length, or it can not be moving at all, and the resulting spool failure condition indicated by the control unit 170 can also indicate whether the spool 112 is moving at all and how much it is moving. The spool failure condition reported by the control unit 170 (whether negative or positive, and what type of positive spool failure it is (e.g., no movement at all or partial movement)) is all based on the current signature measured by the current meter 173.

[0061] Figure 3 An example current signature is shown in FIG. 2. The current signature 200 is characterized by a current rise 210 as a magnetic field forms in the center of the coil 132. When the magnetic field is strong enough to push the spool and overcome the spring force, the spool begins to move through the center of the coil 132, which causes a back EMF 132 to be generated in the coil 132 due to the sudden change in magnetic field or coil inductance, which causes the current to decrease. Once the spool is fully displaced, the current will continue to rise 212 to a continuous peak 216 level, completing the current signature. There are a variety of ways that data from the current signature can be used to detect failures in the spool, and to optimize power usage to extend the life of the solenoid valve.

[0062] One embodiment uses the collected data to optimize power usage by using a pulse width modulation controller (PWM). This embodiment is in the Figure 4current 216 is not needed to keep the spool in place. The current can be reduced to a hold level 218 that is sufficient to keep the spool in place. Reducing the current through the solenoid coil 132 can reduce the power consumption of the valve system. The reduction in power consumption keeps the temperature of the coil at a lower level. When the temperature is kept at a lower level, the life cycle of the coil is significantly increased. The time required to displace the spool and the current profile depend on many factors. Some of the factors include the coil gauge, the load on the spool, the spool friction, and the power supply voltage. Valve spool displacement detection and current reduction can be achieved by detecting the current and using PWM to reduce the coil current to a hold level. The PWM voltage control can be done by a microcontroller or other means. However, using a dedicated hardware circuit to detect the valve spool displacement and enable the PWM controller can achieve accurate detection of the valve spool displacement.

[0063] One embodiment to implement a dedicated circuit is to use a current sense block to detect the current when the spool is displaced, as shown in Figure 5 By using the current sense block 414, the current flowing through the coil when the spool is displaced can be sensed and read. The coil current rate is di / dt = V / L (di / dt is the instantaneous rate of change of current (ampere / second), V is the voltage across the solenoid, and L is the inductance in henry). V can also be calculated as V-V 反电磁力 , where V is equal to the power supply voltage and V 反电磁力 is equal to the back electromotive force of the circuit. The rise of the current is limited by the DC resistance of the coil path. Once the solenoid is energized, the current increases, which causes the magnetic field to expand until the force is strong enough to move the armature. The movement of the armature increases the concentration of the magnetic field. As the magnetic mass of the armature itself moves further into the magnetic field, the magnetic field induces a back voltage in the windings of the solenoid. Because the magnetic field expands rapidly when the armature strokes, the magnetic field causes the current

[0064]

[0065] to momentarily decrease. Where di / dt is equal to the instantaneous rate of change of current, V is equal to the power supply voltage, and V 反电磁力Is equal to the decrease caused by the magnetic field expansion, and L is the inductance. After the armature has moved, the current continues to rise to its maximum peak level. The current signature can be tracked, and the valley point can be used as an indication that the spool has moved to completion. This will be used to provide a position indication. In addition, the signature will be used to provide a trigger for the coil to switch to a pulse width modulation (hold) mode or a fully open (peak) mode. Some benefits of using integrated circuit detection include: a simpler way to switch to peak and hold modes without user configuration. In some embodiments, there will be no need for sensors, which will reduce cost, increase ease of installation, assembly, and the use of signature knowledge will simplify fault detection.

[0066] In Figure 6 A different embodiment of spool shift detection using dedicated circuitry is shown in the block diagram of FIG. 14. First, the current through the coil 132 is detected by the current sense block 414. The output of the current sense block is provided to the negative slope detector block 416 and the positive slope detector block 422 in parallel. Normally, the current will rise until the magnetic field generated is sufficient to start moving the spool. The positive slope will be detected, but the circuit will wait if the negative slope detector latch output 418 has not been set high. Once the spool starts to move, the current will start to decrease, which will cause the negative slope detector to be latched high. Once the spool has moved to its final position, the slope of the current will again start to rise. After the output state of the negative slope detector is set high and the current starts to rise, the positive slope detector 422 can be latched 424 high. After both the negative slope detector and the positive slope detector are set high, the PWM controller 426 is enabled and sets the current to the hold position. This can result in reduced power consumption and lower operating temperatures. Figure 4 The negative slope and positive slope current patterns are shown when the PWM is in place.

[0067] Figure 7 A specific example of the circuit using hardware fabricated in LTSpice is shown. LTSpice is free software that implements a SPICE (Simulation Program with Integrated Circuit Emphasis) circuit simulator. It is manufactured by component manufacturer Linear Technology, which allows for the addition of specific components. These components include a solenoid coil 132 wired in series with the PWM driver 426, which is then attached to the current sense block 414 wired in parallel with the negative slope detector 416 and the positive slope detector 422. As described above, once both the negative slope detector 416 and the positive slope detector 422 are latched high, the PWM enable block wired in series with the negative slope detector block is enabled. This enables the PWM and reduces the current output to the hold position shown. Figure 4 Figure 8 The flowchart of FIG. 15 shows an additional explanation of how the system 1500 will work. ​

[0068] System 1500 starts at start block 1502 where system 1500 determines power input 1504. If the power output is low, system 1500 does nothing 1542. If the power output is high, system goes to spool displacement block 1506. After spool displacement block 1506, there is a decision of negative or positive slope current sense block, then to negative slope block 1510 or positive slope block 1516. If system 1500 goes to positive slope block 1510, system determines if negative slope output state 1514 is set high. If not, system returns to current sense block 1508. If negative slope output 1514 is set high, system 1500 trips latch positive slope output state 1528 high. Then there is an internal delay 1530 and enable pulse width modulation control 1532 which results in coil current reduction to hold level 1534. System then determines if current coil is high or low, if high, system 1500 is turned off 1540 and system 1500 returns to determine power input 1504. If positive slope detection 1510 is negative, system does nothing. Returning to current sense block 1508, if negative slope is detected, system goes to negative slope detection block 1516. From negative slope detection block 1516, if there is a negative response, system 1500 waits 1518, after waiting 1508, system determines if positive slope latch 1522 is set high, if not, a spool failure 1524 is detected and a failure LED 1526 is turned on. If there is a positive response at negative slope detection block 1516, system 1500 does nothing 1520. Returning to negative slope detection block 1516, if there is a positive slope response, system 1500 goes to determine negative slope output state 1514. After positive slope detection block 1510 is positive, system 1500 follows the aforementioned path from negative slope output state 1514.

[0069] An additional benefit of using indirect methods such as dedicated circuitry is that the circuitry is able to monitor the peak and valley of the current signature valve. They also monitor the peak, valley, slope, and amplitude of the current. This is what generates the current signature which can then be compared to previous current signatures to help detect problems that are currently detected using direct non-contact position switches and proximity sensors. Time-based techniques can fail in situations where the spool movement is limited due to increased friction between the spool and the valve body, but the movement time still meets the spool specification.

[0070] Figure 9 A current profile from a healthy coil 1600 is shown. Normal current signatures depend on several variables including, but not limited to: temperature, coil condition, pressure, flow conditions, different sizes of valves (flow), and so on.

[0071] To determine which variables affect the current signature, a number of studies were conducted to show that the signature pattern is different when operating under various conditions. Each of these tests included a healthy spool as a control. Different voltages provided to the system were used to test Figures 10A-10C the plots shown in FIGS. 1700a-1700c. In Figure 10A the current signature 1700a of a solenoid valve with a 28.8V power supply voltage is shown, the current signature 1700a has a current amplitude 1702a of about 1.4A, Figure 10B the current signature 1700b of a solenoid valve with a 24V power supply voltage and a current amplitude 1702b of 1.15A is shown, Figure 10C the current signature 1700c with a power supply voltage of 19.2V and a current amplitude 1702c of 0.9A is shown. The changes in power supply voltage and current also changed the time it took for the spool of the solenoid valve to shift from Figure 10A 55ms 1704a shown to Figure 10B 70ms 1704b shown to Figure 10C 85ms 1704b shown. These plots show that the voltage is the cause of the spool hysteresis and can cause the solenoid valve to fail or have issues.

[0072] One study testing the effect of different temperatures on the solenoid valve and its current signature is shown in Figures 11A-11C FIG. 1800. Figure 11A the current signature 1800a of a solenoid at 0 degrees Celsius is shown, Figure 11B the current signature 1800b at 25 degrees Celsius is shown, and Figure 11C the current signature 1800b at 55 degrees Celsius is shown. Due to the significant differences exhibited, temperature can affect various operating parameters that measure signal levels, and therefore, methods based on stored levels or stored signatures can be ineffective.

[0073] Figure 12 One different study exhibited by the plots in FIGS. 1900a-1900c exhibits a hysteresis spool movement caused by different viscosity levels of liquids. Five different levels in the International Organization for Standardization Viscosity Grades (ISO-3448) were used; ISO-VG5, ISO-VG46, ISO-VG150, ISO-VG320, ISO-VG680. These plots are shown in Figure 12The results shown in FIG. 11 A-11E show the normalized peak current amplitude and the initial slope is different in each case, indicating that different electromagnetic forces are required to move the spool from its position. This is shown in FIG. 11 A, where the initial slope 1102A is different from the initial slope 1102B. The peak current 1103A has a different amplitude than the peak current 1103B, and the time to reach the peak current is different. The same is true for the valley current 1104A and 1104B. The final slope 1105A is the same as the final slope 1105B, because the spool completes its full stroke in each case.

[0074] In the final test, the oil was contaminated with iron powder and white grease with a particle size of 6 to 10 microns. The concentration of particles in the oil was gradually increased from level 1 to level 4. This is shown in FIG. 10A-10E, respectively, where 10A is the oil without contamination. The following changes can be observed in the current signature: the normalized peak current amplitude is different in the oil contaminated state than in the healthy state, the time to reach the peak current increases as the level of oil contamination increases, the normalized valley current amplitude increases as the level of contaminant concentration in the oil increases, the time to reach the valley current in the oil contaminated state is different than in the healthy state. Greater force is required to move the spool if the gap between the spool and the bore is filled with particles. However, the current signature produced is different depending on where the particles fall on the spool. Figure 13 The results shown in FIG. 11 A-11E show the normalized peak current amplitude and the initial slope is different in each case, indicating that different electromagnetic forces are required to move the spool from its position. This is shown in FIG. 11 A, where the initial slope 1102A is different from the initial slope 1102B. The peak current 1103A has a different amplitude than the peak current 1103B, and the time to reach the peak current is different. The same is true for the valley current 1104A and 1104B. The final slope 1105A is the same as the final slope 1105B, because the spool completes its full stroke in each case. Figure 13 The results shown in FIG. 11 A-11E show the normalized peak current amplitude and the initial slope is different in each case, indicating that different electromagnetic forces are required to move the spool from its position. This is shown in FIG. 11 A, where the initial slope 1102A is different from the initial slope 1102B. The peak current 1103A has a different amplitude than the peak current 1103B, and the time to reach the peak current is different. The same is true for the valley current 1104A and 1104B. The final slope 1105A is the same as the final slope 1105B, because the spool completes its full stroke in each case. Figure 13 The results shown in FIG. 11 A-11E show the normalized peak current amplitude and the initial slope is different in each case, indicating that different electromagnetic forces are required to move the spool from its position. This is shown in FIG. 11 A, where the initial slope 1102A is different from the initial slope 1102B. The peak current 1103A has a different amplitude than the peak current 1103B, and the time to reach the peak current is different. The same is true for the valley current 1104A and 1104B. The final slope 1105A is the same as the final slope 1105B, because the spool completes its full stroke in each case.

[0075] As shown in FIG. 12, the normalized peak current amplitude is different in each case, indicating that different electromagnetic forces are required to move the spool from its position. This is shown in FIG. 12A, where the initial slope 1202A is different from the initial slope 1202B. The peak current 1203A has a different amplitude than the peak current 1203B, and the time to reach the peak current is different. The same is true for the valley current 1204A and 1204B. The final slope 1205A is the same as the final slope 1205B, because the spool completes its full stroke in each case. Figure 12 and Figure 13As shown, the spool response degradation mainly occurs in the region between the peak and valley of the current signature. The following key features can be used to monitor the spool response degradation using the current signature: the ratio of "normalized valley current amplitude" to "normalized peak current amplitude", the time to reach peak current, the time to reach valley current, and the time to reach maximum current (steady current). To determine the effect of the above listed features on the spool performance, supervised machine learning techniques (linear regression) can be used. Using linear regression can give the causal relationship between the hysteresis spool movement (dependent variable) and the features extracted from the current signature (independent / explanatory variables). The ratio of "normalized valley current amplitude" to "normalized peak current amplitude" is important as this ratio increases with the increase in the degree of spool response degradation. The time to reach peak current is important as it is an important indication of the start of spool movement and can be affected in case of spool movement being limited due to contaminant particles. The time to reach valley current and the time to reach maximum current are important as they are greatly affected by different voltages. With these features, an optimal fit prediction model can be developed. Each solenoid valve in a healthy state can be trained using this model and get the corresponding values from the regression model which are used as a reference to predict the degradation of the spool response over time due to the above mentioned factors. This helps in predicting the life of the valve and also helps in predicting the life of the valve coil.

[0076] The present disclosure can use statistical process control (SPC) methods along with the regression model to monitor the behavior change of each key feature individually and in turn monitor the degradation of the spool movement (response). As more data is acquired, the algorithm will become more robust. To generate this algorithm, the following approach is followed: The feature values (data) are arranged in chronological order: The mean of the feature values obtained from the training is taken as the mean line, 2 standard deviations of the mean are set as the upper and lower control limits, the current in the oil contaminated state is different from the current in the oil healthy state. The features are listed in the previous paragraph.

[0077] By using the linear regression model, it can be easier to detect when a failure can occur, the cause of the failure, and potentially predict when a failure can occur. Figure 14A trained solenoid valve of one example is shown. The process 1900 of real-time valve core response time degradation algorithm is presented. At step 1902, the process is started automatically or by command. At step 1904, an evaluation is performed as to whether training has been completed (e.g., via process 1000), whereby if training has not been completed, the process stops at step 1924, or if training has been completed, the process proceeds to step 1106. At step 1106, the stored trained data (e.g., from process 1000) is read. At step 1908, the process will dwell in the hold loop of step 1108 until a valve open command is validated, after which the process proceeds to step 1110, where the current signature of the valve is sampled. At steps 1912-1920, the valve core response time is evaluated. At step 1914, the measured current signature is normalized and features are computed. At step 1916, the features are scaled by using standardization techniques to transform the features to a common range by using the mean and standard deviation stored during the learning phase. Step 1920 can be implemented to evaluate SPC (Statistical Process Control) rule violations. The computed response time is used to compute the valve core response time degradation percentage as the percentage change of the measured response time relative to the stored trained response time from process 1000. The process can terminate at step 1924.

[0078] A possible way of real-time system evaluation is shown by system 2000 in Figure 15 The process starts at step 2002, where it checks if the system has been trained, if not, the real-time evaluation stops 2006. If system 2000 has been trained, the system will read the stored trained data at step 2008 and evaluate if the valve has been opened by the system or otherwise at step 2010. If not, system 2000 will check again. If the valve has been opened, system 2000 will sample the current signature at step 2012 and compute the features at step 2014, then evaluate if system 2000 has violated SPC rules at step 2016, followed by measuring the response as a regression output at step 2018, then compute the response degradation change percentage at step 2020 from the measured response and the response from the trained response.

[0079] Figure 16A And Figure 16B A specific example of how a linear regression equation based on collected data can be used to determine the percentage of degradation is shown in Figure 16AIn the figure, two current signatures are shown, where the healthy current signature is 610A with a voltage of 12V across the solenoid. The hysteresis spool current signature 610B has a voltage of 9V across it. For the healthy spool, the valley to peak ratio is 0.786, the time to reach the first peak is 16ms, and the time to reach the first valley is 24ms, and the time to reach the maximum current is 35ms. In order to create a linear regression equation, statistical analysis software is typically used. An example is MiniTab. MiniTab is software developed by Pennsylvania State University and distributed by MiniTab LLC. Minitab can perform automatic calculations and create equations (such as linear regression equations) using data. Linear regression equations are typically in the form of Y=a+bX, where X is the explanatory variable and Y is the dependent variable. For this spool, after using the software to calculate the linear regression equation using the above data, Y equals 53.532. The explanatory variables are the peak to valley ratio (I 谷值 / I 峰值 ), time to reach the first peak (time to reach I 峰值 time to reach the first valley value (time to reach I 谷值 time), and the time to reach the maximum current (the time to reach I 最大 The hysteresis spool (9V) has a peak-to-valley ratio of 0.851, a time to first peak of 24ms, a time to first valley of 31ms, and a time to maximum current of 60ms. Using the equation generated from the healthy spool:

[0080] Y=(86.24*I 谷值 / I 峰值 )--(0.093*(reach I 峰值 Time))+(0.831*(reaching I 谷值 Time))-(0.2019*(reach I 最大 Time))-25.61

[0081] For the lower voltage spool, Y is equal to 59.193.

[0082] To calculate the percent response degradation, use the following equation, using the dependent variable difference (including the most recent calculation y 最近 , the average result of training y 训练 The regression equation output y in the case of valve sticking 最差情况 ), we get the following equation:

[0083]

[0084] Using this calculation, there is a 17.307% degradation in response for the hysteretic spool simulated with the lower voltage. Similarly, Figure 16BIn the middle, an example is shown using oil with a thicker viscosity to exhibit a hysteresis spool. The oil used for the healthy spool 611A is VG-46, with a valley-to-peak ratio of 0.772, a time to first peak of 17 ms, a time to first valley of 23 ms, and a time to maximum current of 33 ms. In the hysteresis spool 611B, VG680 oil is introduced. The time to first peak is 20 ms, the time to first valley is 29 ms, and the time to maximum current is 37 ms. Using a linear regression equation on the data, Y equals 51.869 for the healthy spool and Y equals 60.397 for the hysteresis spool. The response degradation and response degradation percentage calculated using the equation are 0.001 and 24.811, respectively. Using a linear regression model created from the data helps prevent errors and can help predict errors when the degradation percentage exceeds a threshold. As previously mentioned, all spools will produce different results regardless of age, so each spool will have a different healthy linear regression equation.

[0085] Reference is made to Figures 17-22 , a variation of the above-described method for detecting spool performance degradation is presented, in which spool response time degradation and spool position degradation are evaluated to detect spool performance degradation. By detecting spool performance degradation, potential spool failure can be predicted before it occurs, which in turn allows valve failure to be predicted before it occurs. By using such a method, unexpected downtime of the valve and the machine using the valve caused by unexpected failure of the solenoid-operated spool can be reduced. Since most of the description related to the method shown in Figures 1-1 6 applies fully to this example, the same concepts will not be repeated here when such overlap occurs. In addition, the concepts presented for this example can be combined with the concepts presented in Figures 1-1 6.

[0086] In the examples presented in Figures 17-22 , the disclosed method involves predicting the progression of the failure occurring in the spool, which in turn involves predicting the valve failure due to the spool failure. The degree of predicting the failure in advance depends on the accuracy of the detection system. Therefore, it is necessary to detect the spool failure as early as possible. The spool failure has two aspects: (1) spool position degradation (spool movement is limited), resulting in reduced flow output; and (2) spool response time degradation, resulting in the spool completing the movement to produce full flow, but the time to reach the final position of the spool exceeds the specified time. Either of these two cases can occur alone, or both can coexist at a given moment. In addition, either case alone can cause the spool to fail or operate below the desired performance level. Therefore, by monitoring both cases simultaneously, an improved assessment of the overall health of the spool valve can be achieved.

[0087] Reference is made to Figure 17, showing the actual current signature 300 of the coil 132 and the ideal current line 400, where the current is plotted against time. In one aspect, the current signature 300 includes a first peak 302, a first valley 304, a last valley 306, a minimum point 308 from the ideal line, and a 90% point 310 of the maximum current value.

[0088] Many useful features can be extracted from the current signature 300 and the ideal current line 400 for detecting spool response time and position degradation. For example, the features in the following paragraphs can be extracted:

[0089] Time to reach first peak current: This is the difference between the time the current command was given and the time the first peak 302 is observed on the current signature 300. This feature indicates the start of the spool movement.

[0090] Time to reach last valley current: This is the difference between the time the current command was given and the time the last valley 306 is observed on the current signature. This feature indicates the end of the spool movement.

[0091] Time to reach 90% of the maximum / steady state current: This is the difference between the time the current command was given and the time 90% current 310 of the steady state value is observed on the current signature. This feature shows the change in behavior when the spool response time and spool position degrade.

[0092] Minimum point close to zero from the ideal line: The ideal line 400 is plotted as shown in Figure 17 , and the shortest distance between this line and the current signature (the closest point 308 on the current signature below the line) is used as a feature. It also serves as a marker for the completion of the spool movement.

[0093] Number of low points in the current signature: This feature indirectly counts all the small negative slopes in the current signature, which are an indication of mechanical movement in the magnetic field.

[0094] Difference of 'first valley current' and'steady state current': This is the difference in magnitude of the current seen at the first valley 306 and the steady state current. Depending on the distance the spool travels, the magnitude of the valley is affected, and so is this difference.

[0095] Ratio of'square of current at first valley' and 'current at first peak': This derived feature monitors the changes happening in the peak valley region of the current signature.

[0096] Euclidean distance between the reference stuck curve and the latest recorded current signature: This feature is used to monitor the health of the valve by comparing the latest recorded signature with the current signature under the worst case (completely stuck).

[0097] ReferenceFigure 17 The determination of the ideal characteristic line 400 is presented in more detail. In one aspect, the ideal characteristic 400 line extends between a start point 400a and an end point 400b with a plurality of steps or samples 400c in between. In one example, the ideal current characteristic 400 of 150 mS is captured in 150 instantaneous samples or steps 400c. To produce the ideal characteristic line 400, the current is first normalized. Subsequently, the steps 400c are calculated using the following equation:

[0098] Calculation: Step 400c = (1 - first point of current characteristic 400d) / 150.

[0099] With such step points, the ideal line is drawn by cumulatively adding 150 steps. Thus, the ideal line 400 will also have the same number (150) of points 400c. Once the ideal characteristic line 400 is produced, the aforementioned ‘minimum distance from ideal line approaching zero’ feature 308 can be calculated.

[0100] In one aspect, the aforementioned features can be used in an algorithm to detect spool response time degradation. For example, the following features can be used: time to reach first peak current, time to reach last valley current, time to reach 90% of maximum current, number of low points, and minimum point from ideal line approaching zero. In one aspect, a variety of supervised machine learning techniques can be used to derive spool response time (dependent variable) from the extracted features (independent / explanatory variables) from the current characteristic. As Figure 18 Illustratively, a linear regression, polynomial regression model(s), or module(s) 500 can be used to predict the spool response time using the aforementioned features as inputs. Other methods can be used. In one aspect, the strength of the impact of these features on the spool performance is found using linear / polynomial regression. Regression analysis can derive the causal relationship between the time taken for the spool to move (dependent variable) and the features extracted from the current characteristic (independent / explanatory variables). With this knowledge, a best-fit predictive model can be developed with the observed value dataset. Thus, each solenoid valve in a healthy state can be trained with this model and the corresponding Y value can be obtained from the regression model, which is used as a reference to predict the degradation of the spool response time over time.

[0101] As mentioned above, the regression model predicts the spool response time. In one example, the following equation shows one of the models obtained after regression, which mainly indicates the feature coefficients.

[0102] Valve core response time = X + (0.2X) * 'time to reach last trough' + (0.022X) * 'time to reach first peak' + (0.081X) * 'time to reach 90% steady state current' + (0.014) * 'number of low points in current signature' + (0.053X) *'minimum point where proximity to ideal line is zero'.

[0103] In one particular example, X is approximately 76.1. In one aspect, the above will yield the valve core response time. Thus, the system can measure the valve core response time as a regression output and then calculate the response degradation percentage change from the measured response time and the response time from the trained response as follows:

[0104] Valve core response time degradation percentage at time x = 100 * abs((response time calculated at training - response time calculated at time x) / (response time calculated at training))

[0105] The prediction model has shown to achieve an R2 score of over 98% and a root mean square error of less than 3.0 by using the root mean square error method to evaluate the goodness of fit metric of the linear regression model, where the prediction model uses the inputs and calculations described above.

[0106] Referring to Figure 19 , an online learning flowchart of the process 1000 of the valve core response time degradation detection algorithm is presented. In step 1002, the process is initiated either automatically or by command. In step 1004, it is evaluated whether the valve has been trained. If so, the process terminates at step 1022. If the valve has not been trained, the process proceeds to step 1006 where it is evaluated whether a training command has been received. If not, the process terminates at step 1022. If so, the process proceeds to step 1008 where it is determined whether a minimum number of training cycles has been reached where the valve current signatures are stored. In one example, at least 30 training cycles must be received before proceeding. If the minimum number of training cycles has not been reached, the process proceeds to step 1010 where the current signature is recorded in temporary memory. In step 1012, the aforementioned features of the current signature are calculated and the process loops back to step 1008 until the minimum number of training cycles is reached. At each step 1012, the current signature is also normalized in step 1014 and the identified features are calculated in step 1016. Once the minimum number of training cycles is reached, the process proceeds to step 1018 where various further calculations are performed. For example, step 1018 can calculate the mean value of the calculated features, the value of the standard deviation, the variance, and the regression results. In step 1020, the learned parameters / features are stored in memory (e.g., permanent memory) and thereafter the process terminates at step 1022.

[0107] Referring Figure 20 Another process 1100 that presents a real-time spool response time degradation algorithm is presented. At step 1102, the process is started automatically or by command. At step 1104, an evaluation is performed as to whether training has been completed (e.g., via process 1000), whereby if training has not been completed, the process stops at step 1124, or if training has been completed, the process proceeds to step 1106. At step 1106, the stored trained data (e.g., from process 1000) is read. At step 1108, the process will dwell in a hold loop at step 1108 until a valve open command is validated, after which the process proceeds to step 1110, where the current signature of the valve is sampled. At steps 1112 through 1120, the spool response time is evaluated. At step 1114, the measured current signature is normalized and features are computed. At step 1116, the features are scaled by using standardization techniques to transform the features to a common range by using the mean and standard deviation stored at the learning phase. At step 1118, the information is used in a regression model to compute the response time. Step 1120 can be implemented to evaluate SPC (Statistical Process Control) rule violations. The computed response time is used to compute the spool response time degradation percentage as the percentage change in the measured response time relative to the stored trained response time from process 1000. The process can terminate at step 1124.

[0108] Linear regression and polynomial regression logic models or modules of Figure 18 can also be used with different inputs and using the following features to predict the spool position: the difference of the "first valley current" and the "steady state current"; the Euclidean distance between the reference stick curve and the latest recorded current signature; the time to reach the first peak current; the time to reach 90% of the maximum current; the time to reach the last valley current; the ratio of the "current square at the first valley" and the "current at the first peak".

[0109] Various methods can be used to detect the spool position achieved by the valve. In one example, the real-time position is predicted by using a pre-trained model. For different configurations of the valve (e.g., Eaton Corporation size 3 single solenoid spool valve, Eaton Corporation size 5 single solenoid spool valve, Eaton Corporation size 5 dual solenoid spool valve, etc.), a pre-trained prediction model can be obtained from experimental data. In one aspect, a variety of supervised machine learning techniques can be used to obtain the completed spool movement (dependent variable) based on the features (independent variables / explanatory variables) extracted from the current characteristics. In this example, linear regression is used. Linear regression gives a causal relationship between the completed spool movement (dependent variable) and the features (independent variables / explanatory variables) extracted from the current characteristics. With this knowledge, a best-fit prediction model for the observed value data set can be developed based on the configuration of the selected valve. Each solenoid valve in a healthy state is trained with the model, and the corresponding Y is obtained from the regression model, which is used as a reference for predicting the degradation of the spool position over time. In one aspect, the advantage of this approach is that the features used to predict the completed spool movement remain constant regardless of the configuration of the solenoid-operated spool valve, and only the feature coefficients in the machine learning model change. For more details on the learning and testing process, please refer to Figure 21 and Figure 22 .

[0110] The 'completed spool position' of this method can be predicted using a regression model approach. The complete spool position can be calculated by the following equation:

[0111] Completed spool position = X + (0.575X) * 'first valley value and I 稳定 Difference' + (0.601X) * 'Euclidean distance' + (0.222X) * 'Time to reach the first peak' - (0.584X) * 'Time to reach 90% stable current' - (0.117X) * 'Time to reach the last valley' + (0.236X) * 'Ratio of the square of the first valley to the first peak'

[0112] The above equation outputs the completed spool movement. In a specific example, X is approximately 2.229. Accordingly, the system measures the completed spool movement as the regression output and then calculates the percentage change in position degradation based on the measured completed position and the position from the trained response using the following equation:

[0113] Spool position degradation percentage at time x = 100 * abs((spool movement calculated during training - spool movement calculated at time x) / (spool movement calculated during training))

[0114] The prediction model has shown to achieve an R2 score of over 99% and a root mean square error of less than 0.1 by using the root mean square error method to evaluate the goodness of fit measure of the linear regression model, where the prediction model uses the inputs and calculations described above.

[0115] Referring Figure 21 An online learning flowchart of the process 1200 of the spool position degradation detection algorithm is presented. In the case where the process 1000 is also implemented, note that the process 1000 and the process 1200 can be merged into a single process while recording and calculating the necessary information and features for determining the spool response time and position degradation. In step 1202, the process is initiated automatically or by command. In step 1203, the valve configuration is selected or identified for training. Note that similar steps can be incorporated into the process 1000. In step 1204, it is evaluated whether the valve has already been trained. If yes, the process terminates in step 1222. If the valve has not been trained, the process proceeds to step 1206 where it is evaluated whether a training command has been received. If no, the process terminates in step 1222. If yes, the process proceeds to step 1208 where it is determined whether a minimum number of training cycles in which the valve current features are stored has been reached. In one example, at least 30 training cycles must be received before proceeding. The minimum number of cycles for the processes 1000 and 1200 can be the same or different. If the minimum number of training cycles has not been reached, the process proceeds to step 1210 where the current features are recorded in temporary memory. In step 1212, the aforementioned features of the current features are calculated and the process loops back to step 1208 until the minimum number of training cycles is reached. At each step 1212, the current features are also normalized in step 1214 and the identified features are calculated in step 1216. Once the minimum number of training cycles is reached, the process proceeds to step 1218 where various further calculations are performed. For example, step 1218 can calculate the mean value, the value of the standard deviation, the variance of the calculated features, and the regression results. In step 1220, the learned parameters / features are stored in memory (e.g., permanent memory) after which the process terminates in step 1222.

[0116] Referring Figure 22, another process 1300 for presenting real-time spool position degradation algorithms is presented. At step 1302, the process 1302 is started automatically or by command. At step 1304, an evaluation is performed as to whether training has been completed (e.g., via process 1200), whereby if training has not been completed, the process stops at step 1324, or if training has been completed, the process proceeds to step 1306. At step 1306, the stored trained data (e.g., from process 1200) is read. At step 1308, the process will dwell in the hold loop at step 1308 until a valve open command is validated, after which the process proceeds to step 1310, where the current signature of the valve is sampled. In steps 1312 through 1320, the spool position movement is evaluated. At step 1314, the measured current signature is normalized and features are calculated. At step 1316, the features are scaled by using standardization techniques to transform the features to a common range by using the mean and standard deviation stored during the learning phase. At step 1318, the information is used in a regression model to calculate the spool movement. Step 1320 can be implemented to evaluate SPC (Statistical Process Control) rule violations. The calculated spool movement is used to calculate the spool position degradation percentage as a percentage change in the measured movement relative to the stored trained movement from process 1200. The process can terminate at step 1324.

[0117] Figure 21 and Figure 22 The flowchart for'spool position degradation detection' is discussed when a pre-trained model is available for the required spool operated solenoid valve configuration. For example, when a valve configuration is selected at step 1203, the algorithm will know which pre-trained model (regression coefficients) should be used as the health baseline for real-time evaluation of spool position degradation. When we do not have a pre-trained model for the required spool operated solenoid valve configuration, the self-learning regression model mentioned is another method that can be used to find the health baseline for'spool position degradation detection'. In this case, the regression model (regression coefficients) can be learned during the online learning phase as shown in Figure 23 . In such a method, once the health baseline is determined for the valve configuration during the online learning phase, the real-time evaluation of'spool position degradation' will be the same as described with respect to Figure 22 .

[0118] Referring to Figure 23the process 1400. In the case where the process 1000 is also implemented, note that the process 1000 and the process 1400 can be merged, in whole or in part, into a single process, while recording and calculating the necessary information and features for determining the spool response time and position degradation. In step 1402, the process is initiated automatically or by command. In step 1403, the valve configuration is selected or identified for training. Note that similar steps can be incorporated into the process 1000. In step 1404, it is evaluated whether the valve has already been trained. If so, the process terminates in step 1430. If the valve has not been trained, the process proceeds to step 1406, where it is evaluated whether a training command has been received. If not, the process terminates in step 1430. If so, the process proceeds to step 1408, where it is determined whether a minimum number of training cycles, in which the valve current features are stored, has been reached. In one example, at least 30 training cycles must be received before proceeding. The minimum cycle number for the processes 1000 and 1400 can be the same or different. If the minimum training cycle number has not been reached, the process proceeds to step 1410, where the current features are recorded in temporary memory. In step 1412, the aforementioned features of the current features are calculated, and the process loops back to step 1408 until the minimum training cycle number is reached. At each step 1412, the current features are also normalized in step 1414, and the identified features are calculated in step 1416. Once the minimum training cycle number is reached, the process proceeds to step 1418, where various further calculations are performed. For example, step 1418 can perform a regression to Y = fn(features), Y = stroke length, and step 1420 can calculate performance metrics of the regression, such as Rsq (R-squared), RMSE (Root Mean Square Error). In step 1422, the Rsq and RMSE are compared to the spool stroke to determine whether the model is good or not. In one example, the model is good if the Rsq is greater than 70% and the RMSE is less than 20% of the spool stroke. In the case where these parameters are not met, the process terminates in step 1430. In the case where these parameters are met, the process moves to step 1426, where further calculations are performed, such as calculating the mean, standard deviation, variance of the calculated features, and the regression results. In step 1428, the learned parameters / features are stored in memory (e.g., permanent memory), after which the process terminates in step 1430.

[0119] By using the above process, both spool response time degradation and spool position degradation can be assessed simultaneously with a comparison of the change value to a baseline (e.g., model value). Both degradations can be expressed as a percent change, percent error, percent difference, and / or an actual or absolute change in value. In some examples, the system monitors the spool response time and position degradation change values and compares them to a threshold. In some examples, an alert or signal is generated when either of the spool position or response time change values exceed a threshold (e.g., predetermined threshold). For example, a signal can be generated and transmitted over a vehicle CAN bus system to indicate that a valve should be evaluated, serviced, or replaced. In some examples, an alert or signal is generated when both the spool position and response time change values exceed respective thresholds. By such a method, a slide valve failure can be prevented from occurring before it happens.

[0120] From the foregoing detailed description, it will be evident that modifications and variations can be made in the aspects of the disclosure without departing from the spirit or scope of the disclosure. While there has been described in detail herein best modes of practicing the teachings of the present disclosure, it will be apparent to those skilled in the art that various alterations, modifications, and improvements can be made without departing from the scope of the appended claims.

Claims

1. A solenoid-operated valve comprising: at least one coil and at least one adjustment member; a controller that interfaces with the current meter to monitor a current characteristic of the coil when the solenoid-operated valve is actuated by operating the solenoid-operated valve in an actuation mode in which current is driven through the coil using a first power level to move the adjusting member; and The controller includes a processor and a memory in electronic communication with the processor for executing a power optimization algorithm operable to: detecting when the adjustment member has begun to shift based on a sensed current of the current characteristic sensed by the ammeter; detecting when the adjusting member has reached a final position based on a sensed current of the current characteristic sensed by the ammeter; as well as Once the regulating member has been determined to be in the final position, the solenoid-operated valve is shifted from the actuation mode to the holding mode, wherein, when the solenoid-operated valve operates in the holding mode, current is driven through the coil using a second power level, and wherein the second power level is lower than the first power level.

2. The valve according to claim 1, wherein The second power level of the hold mode is controlled by a pulse width modulation controller.

3. The valve according to claim 1, wherein The controller includes an integrated circuit having the solenoid coil.

4. The valve according to claim 1, wherein The controller detects whether the adjustment member has begun to shift by detecting when the current changes from a positive slope to a negative slope.

5. The valve according to claim 1, wherein The controller detects that the adjustment member has reached its final position by detecting that the current has transitioned from a positive slope to a negative slope and then back to a positive slope.

6. The valve according to claim 5, wherein The controller uses a first latch that is set to a high output when the system detects a negative slope, and uses a second latch that is then set to high when the controller detects a positive slope after the output state of the first latch has been set to high; once both the first latch and the second latch are set high, the controller transitions the current to a holding state.

7. The valve according to claim 1, wherein The regulating member is a valve core.

Images