AUTOMATED DIAGNOSTIC METHOD FOR A PRODUCTION MACHINE

Abstract

AUTOMATED DIAGNOSTIC METHOD FOR A PRODUCTION MACHINE The invention relates to a computer-implemented method for the automated diagnosis of a critical path of a production machine configured to implement a production cycle and comprising a plurality of components capable of emitting or receiving a signal, including a plurality of actuators and a plurality of sensors, the method comprising the following steps: S1: Acquisition of all signals generated and / or received by the machine components; S2: Identification, among all acquired signals, of each signal corresponding to a setpoint signal injected into an actuator or a status signal emitted by a sensor; S3: Selection of the signals associated with a first group of actuators and sensors from among the plurality of actuators and sensors, the first group of actuators and sensors defining the production cycle of the machine;S4: Association of signals from the first group of actuators and sensors to establish signal pairs, each signal pair corresponding to a setpoint signal from a predetermined actuator and a feedback signal from a predetermined sensor associated with said predetermined actuator; S5: Identification of a critical path of the production cycle from the actuator-sensor signal pairs, the critical path corresponding to an uninterrupted sequence of consecutive movements of a subgroup of actuators from the first group of actuators and sensors during the entire production cycle of the machine.

Description

Title of the invention: AUTOMATED DIAGNOSTIC METHOD FOR A PRODUCTION MACHINE TECHNICAL FIELD OF THE INVENTION

[0001] The technical field of the invention is that of the identification of a fault on a machine, more particularly the detection and diagnosis of the non-optimal operation of a production machine performing cyclic operations. TECHNOLOGICAL BACKGROUND OF THE INVENTION

[0002] Production machines conventionally used in industrial production such as multi-axis machining centers, manufacturing lines equipped with operating stations and a conveyor, or a raw material processing line, conventionally include a multitude of actuators controlled by a control chain.

[0003] This type of machine is generally installed for a long period of time, in order to make the investment necessary for its installation profitable, and during its lifetime carries out large numbers of production cycles, that is to say a cyclical repetition of the same successive operations.

[0004] Due to wear and tear of parts, actuators, sensors, and maintenance operations performed on the machine, it is common for the duration of a machine's production cycle to deviate significantly over time, even without changing the control program governing its operation, thereby affecting the machine's production capacity.

[0005] Such deviations are not detected as a classic defect of a machine component because they do not critically affect the operation of a machine component, and are therefore generally not detectable by means of defect detection methods classically known in the art, such as the defect detection method presented in document EP 1 403 437.

[0006] Some analytical diagnostic methods are classically carried out in a "manual" manner, by means of technical experts carrying out a study of the operation of the machine in order to identify the actuators involved in the production cycle, an analysis of the structure of the machine control system in order to identify the ports associated with the actuators to be monitored, an acquisition of signals and an analysis of the information collected, a step of reconstructing the critical path of the production cycle, and an identification of a corrective means in order to improve the operation of the machine.

[0007] A production machine generally comprises a plurality of actuators, each associated with a pre-actuator activated by a control system or a control unit, one or more sensors allowing the actuator's state to be returned continuously or an analog signal corresponding to the attainment of a particular situation or not, which may condition the activation of another actuator.

[0008] Each of these elements can be associated in parallel with other elements performing a similar function, for reasons of redundancy, reliability, or safety. The number of components emitting and receiving signals can then become very large depending on the size and complexity of the machine, and the complexity of the analysis by human means becomes a limiting factor due to the investment in time, and therefore in cost, that it entails.

[0009] Moreover, since each machine has developed its own wear during its lifetime, the diagnosis made on one machine would not be applicable to another machine of the same type, and it would therefore be necessary to repeat the process.

[0010] Consequently, the investment representing the carrying out of such a diagnosis on a production unit whose production capacity is to be increased may become greater than that required for the addition of a new device or a new machine, which entails pollution in view of the associated energy consumption but also in view of the energy and materials required for the manufacture of this additional machine.

[0011] There is therefore a need for a solution enabling a diagnostic of the operation of a production machine to be established more efficiently and more quickly than prior art methods, in particular in order to allow an increase in the lifespan of production machines. Summary of the invention

[0012] The invention proposes a computer-implemented method for the automated diagnosis of a critical path of a production machine configured to implement a production cycle and comprising a plurality of components capable of emitting or receiving a signal, including a plurality of actuators and a plurality of sensors, the method comprising the following steps: SI: Acquisition of all signals generated and / or received by the machine components; S2: Identification, among all the acquired signals, of each signal corresponding to a setpoint signal injected into an actuator or a status signal emitted by a sensor; S3: Selection of signals associated with a first group of actuators and sensors from among the plurality of actuators and sensors, the first group of actuators and sensors defining the production cycle of the machine; S4: Association of signals from the first group of actuators and sensors to establish signal pairs, each signal pair corresponding to a setpoint signal from a predetermined actuator and a feedback signal from a predetermined sensor associated with said predetermined actuator. S5: Identification of a critical path of the production cycle from the actuator-sensor signal pairs, the critical path corresponding to an uninterrupted sequence of consecutive movements of a subgroup of actuators from the first group of actuators and sensors during the entire production cycle of the machine.

[0013] Such a process makes it possible to identify in an automated manner the actuators whose operation defines the cycle time and on which a modification of operation will have an equivalent impact on the cycle time, thus making it possible to optimize production.

[0014] Advantageously, such a method is complemented by the following features, taken alone or in combination:

[0015] - acquisition of signals over several cycles S5

[0016] - continuous monitoring of the signals from each of the actuator-sensor pairs critical analysis is carried out following the calculation step S6, the process further comprising an anomaly detection step S7, during which the actuators for which the cycle time presents an anomaly are identified;

[0017] - the method further includes a step S8 for identifying the type of operation to to be carried out to compensate for the identified anomaly;

[0018] - the method further comprises a step S0 of connection to the machine of production in such a way as to allow the collection of signals passing through the machine;

[0019] - the acquisition SI step comprises the following sub-steps: SU: Detection of actuator and sensor control systems S12: Collection of signals emitted and received by the control systems

[0020] - the selection step S3 includes the following sub-steps: S31: Observation of the signals sorted during step S2 during a plurality of operating cycles; S32: Elimination of setpoint signals not associated with a state feedback signal; S33: Elimination of signals not occurring for each machine operating cycle;

[0021] - Step S4 comprises the following sub-steps: S41: Observation of selected signals during a plurality of cycles; S42: Identification of cause-and-effect relationships between setpoint and state feedback signals to identify state feedback correlated with an actuator setpoint and / or actuator setpoints following a state feedback;

[0022] - The S5 identification step comprises the following sub-steps: S51: Identification of the setpoint signals that determine the critical path of the cycle; S52: Identification of a critical path;

[0023] - the method includes a step of continuous monitoring S6 of each couple actuator-detector of the critical path during a plurality of cycles;

[0024] - the method includes an anomaly detection step S7 carried out continuously in parallel to step S6, which includes the following sub-steps: S71: Calculation of the time of the critical path steps at each iteration; S72: Calculate at least one characteristic metric of step times, for example a mean, a standard deviation, a variability; S73: Compare the calculated metrics to reference values ​​established during process initialization; S74: Issue an alert if a threshold is crossed;

[0025] - the steps are carried out during the operation of the machine without interfering with its operation. BRIEF DESCRIPTION OF THE FIGURES

[0026] The figures are presented for illustrative purposes only and are in no way limiting of the invention.

[0027] [Fig.1] is a schematic representation of the steps of the process according to the invention. DETAILED DESCRIPTION

[0028] The invention relates to a method, implemented by computer or by a dedicated processing module comprising a communication unit capable of transmitting and receiving signals so as to implement the method, for the automated diagnosis of a production machine configured to implement a production cycle and comprising a plurality of active components capable of performing an action in response to a received signal, including a plurality of actuators capable of changing state in response to a received instruction, and a plurality of sensors capable of acquiring the state of an actuator or of a state of the production cycle, the method comprising the following steps: SI: Acquisition of all signals generated and / or received by the machine components; S2: Identification, among all the acquired signals, of each signal corresponding to a setpoint signal injected into an actuator or a status signal emitted by a sensor; S3: Selection of signals associated with a first group of actuators and sensors from among the plurality of actuators and sensors, the first group of actuators and sensors defining the production cycle of the machine, the first group thus including all the actuators and sensors involved in steps defining the production cycle time, and no other actuator or sensor of the machine; S4: Association of the signals of the first group of actuators and sensors so as to establish pairs of signals, each pair of signals corresponding to a setpoint signal of a predetermined actuator and a feedback signal of a predetermined sensor associated with said predetermined actuator, a sensor being considered as associated with an actuator if said sensor is configured to acquire the state of said actuator; S5: Identification of a critical path of the production cycle from the actuator-sensor signal pairs, the critical path corresponding to an uninterrupted sequence of consecutive movements of a subgroup of actuators from the first group of actuators and sensors during the entire production cycle of the machine.

[0029] Such a process therefore makes it possible to identify in a fully automated way the critical path of the production cycle of a production machine, and thus to determine a level of machine performance which can be used as a reference level.

[0030] The reference level can thus be compared to a known previous performance level, which is subsequently used to restore a performance level close to the initial performance level, for example, by comparing actuator data with the specifications declared during machine installation. Alternatively, the reference level can be used to prevent subsequent drift in machine operation by continuously monitoring its operating data and detecting any drift in the operating data relative to the reference level. This makes it possible to maintain machine performance at a level close to the reference level.

[0031] In particular, cycle drift, reflecting a deterioration in the execution time of movements in the critical path, has a direct impact on line efficiency. Performing a diagnosis on the movements constituting the critical path allows for a correction that has a direct impact on the productivity of the production line.

[0032] Motion drift, that is, a degradation in the execution time of a movement, makes it possible to identify wear on a specific component of the line. Alternatively, the reference level can be used to optimize the operation of the machine by comparing the operating data of the actuators with the physical capabilities of the actuators, in order to identify a path of optimization of the control of the actuators towards a chosen performance point.

[0033] Thus, the invention makes it possible to produce quickly and efficiently a functional diagnosis which can be used as an optimization, maintenance or preventive maintenance tool for any machine allowing a connection and exchange of information.

[0034] The method is implemented by means of a computing unit equipped with a processor and a memory comprising code data which, when processed by the processor, allows the steps of the diagnostic process to be carried out.

[0035] The computing unit is connected directly or indirectly to the machine for which a diagnosis is to be carried out, and the process is implemented automatically, or as a result of a user command.

[0036] The process can be carried out during the normal operation of the machine, without interfering with its operation, in particular without requiring any modification of the PLC programs. Alternatively, the process can be carried out by subjecting the machine to sequencing in groups of cycles.

[0037] Carrying out the process continuously without interfering with the operation of the machine has the advantage of greater consistency of the information collected with the normal conditions of use of the machine and of establishing a diagnosis from data more faithful to the classic operating data of the machine.

[0038] The diagnostic process advantageously includes a connection step S0 configured to allow the computing unit to be connected to a production machine and to execute a data collection process in order to acquire the signals passing between the different functional elements of the machine.

[0039] The S0 connection step advantageously includes a substep for establishing SOI communication between the processing unit and the machine.

[0040] Advantageously, the acquisition step SI of the signals generated by each of the machine components includes a detection substep SU configured to detect the signals emitted and received by the actuator and detector control units, and a collection substep S12 configured to allow the computing unit to collect the signals emitted and received by the control units.

[0041] The identification step S2 is configured to identify each signal corresponding to a setpoint signal injected into an actuator or a status feedback signal emitted by a sensor or detector. Boolean values ​​identified as inputs or outputs in the PLC are identified and stored. Other Boolean signals from the PLC are filtered and eliminated, as are all PLC variables.

[0042] The selection step S3 is configured to remove from the observed signals the state signals not relating to actuators defining the cycle time, so as to select the signals associated with actuators or sensors defining the cycle time. The selection step S3 advantageously includes the substeps S31 for identifying all setpoint signals that are associated with a state feedback signal, and S32 for eliminating "isolated" signals, that is, signals that are not associated with a "counterpart," a state feedback signal, or a setpoint.

[0043] In one embodiment, during the selection step S3, a calibration period is initiated, corresponding to an acquisition over several dozen cycles for the production of a part or a range of parts. The occurrence distribution of all the selected movements is then observed. The mode of this distribution is then determined, the mode being defined as the number of cycles, denoted n, performed during this calibration period. Signals with a number of occurrences corresponding to k*n±l, k€N are considered to define the cycle time, since they are performed in each production cycle. All signals associated with movements having a number of occurrences other than k*n±l, k€N are eliminated. In other words, during this step, all signals are observed over N cycles, and if a signal does not have an activation number that corresponds to a multiple of N, it is not correlated with the key steps of the cycle.The signal is therefore isolated from the observation selection, which has the effect of reducing the observation selection to signals that are of interest in the diagnosis, thereby reducing the calculation or processing time of the following steps.

[0044] The association step S4 can advantageously be carried out by learning, by observing over a plurality of cycles the signals filtered during the selection step S3. The setpoint and response signals for the actuators are compared so as to identify actuator-sensor sets critical for the cycle time.

[0045] Advantageously, the association step S4 includes an observation substep S41 configured to record the selected signals over a plurality of cycles, and an identification substep S42 during which the observed signals are processed to identify cause-and-effect relationships between the actuator setpoint signals and the sensor or detector state feedback signals in order to identify state feedback correlated with an actuator setpoint and / or actuator setpoints resulting from state feedback. When a causal relationship is established between an actuator and a sensor or detector, they are associated and subsequently treated as an actuator-sensor assembly.

[0046] In one embodiment, the S42 identification substep, the association of inputs and outputs, is performed by means of several processing layers. On an acquisition log, each occurrence of an output signal is recorded and the Potential inputs that are executed when the output is running are identified. Then, the total number of occurrences of an output during acquisition is observed and compared with potential inputs that have the same number of occurrences, within a tolerance. Inputs potentially linked to an output and exhibiting an equivalent number of occurrences are retained. Finally, the output name is compared with the names of the retained potential inputs using an approximate string matching algorithm to determine the corresponding input.

[0047] The number of cycles during which the observation substep S41 is carried out can be between 20 cycles and 100 cycles, depending on the production process, the deviation and variability of the cycle, and the number of repetitions of certain steps in a cycle.

[0048] Once the actuator-sensor assemblies are identified, during the identification step of a critical path S5 of the production cycle is carried out, in order to allow an estimation of the cycle time.

[0049] The critical path is understood to be an uninterrupted sequence of consecutive movements of a group of actuators from among all the actuators of the machine during the entire production cycle of the machine. Any additional time spent on one of the movements of the critical path will impact the entire cycle by the same amount of time.

[0050] Step S5 may advantageously include a monitoring substep S50 in which the signals emitted and received by the cycle-time critical actuator-sensor assemblies are recorded for a plurality of cycles.

[0051] Step S5 advantageously comprises the following substeps: S51: Identification of setpoint signals that appear to be crucial for the critical path of the cycle; S52: Identification of a critical path.

[0052] During substep S51, the setpoint signals present in each cycle are extracted and processed to identify the critical path of the cycle. The critical path is defined as the sequence of movements without any pauses between two consecutive movements. For each signal ending that triggers a first movement, the signal(s) triggered after the end of the first movement are identified. Since the signals are associated with their respective movements, only the longest signal in terms of duration that triggers a signal associated with the following movement is retained and identified as part of the critical path. If several sequences of consecutive movements without pauses are identified, the longest sequence of movements in terms of duration is considered the critical path.

[0053] Substep S52 is then carried out, during which the cycle time of the critical path is calculated. This step can be performed using recordings acquired during steps S3 or S4, or alternatively by monitoring the actuator-sensor assemblies of the critical path.

[0054] The result of this step therefore makes it possible to determine the cycle time of the machine on its critical production path, which makes it possible to establish a model of the operation of the machine at a time T, which can serve as a reference for comparison to establish a diagnosis.

[0055] The comparison reference can then be used in subsequent steps as a comparison element in order to detect an anomaly.

[0056] Advantageously, following the identification of the machine's critical path, a continuous monitoring step S6 of each actuator-detector pair in the critical path can be performed over a plurality of cycles. Such a step is advantageously associated with a detection step S7, configured to determine whether the operation of an actuator is abnormal or not. The actuators can be observed to detect a drift in the duration of a cycle step, or variability in the duration of a step.

[0057] In one embodiment, the detection step S7 comprises the following substeps: a substep S71 in which the durations of the critical path steps are calculated at each iteration of the cycle; a substep S72 in which metrics characteristic of the step times are calculated, for example, a mean, a standard deviation, and variability, in order to monitor performance criteria chosen according to the desired objective; a substep S73 in which the calculated metrics are compared to reference values ​​established in step S6; and a substep S74 in which an alert is issued if a metric differs from its associated reference value. Advantageously, a threshold can be set for each reference value in order to define a tolerance zone that will trigger the issuance of an alert if the metric crosses the threshold.

[0058] In one embodiment, an S8 step is configured to identify the type of operation to be performed to compensate for the identified anomaly.

[0059] Such a process can be used to perform preventive maintenance, or to maintain a level of productivity similar to the machine's productivity level at the time the cycle time identified in step S5 was established.

[0060] For example, by carrying out the process at the time of machine start-up, and by performing continuous monitoring during the machine's lifetime, it is thus possible to avoid drift in one or more of the actuators and thus maintain a productivity level similar to that observed when the production line was started up.

[0061] Advantageously, step S5 is performed at regular intervals to establish a diagnosis of the machine: if the critical path identified at the end of step S5 remains unchanged, then the machine's operation is essentially nominal. If another critical path is identified, this indicates a non-uniform drift of the actuators of the first critical path and of the actuators outside the first critical path, which impacts the production cycle time. This allows, in addition to the continuous monitoring performed during step S7 and focused on the drift of the critical path, verification that the overall condition of the machine, outside the critical path, does not drift in such a way as to degrade the cycle time.The frequency of this iteration can be defined by the operator according to operational constraints, for example weekly, or monthly, or every N cycles, depending on the duration of the cycle and the criticality of the consequences of a drift.

Claims

Demands

1. A computer-implemented method for the automated diagnosis of a critical path of a production machine configured to implement a production cycle and comprising a plurality of components capable of emitting or receiving a signal, including a plurality of actuators and a plurality of sensors, the method comprising the following steps: S1: Acquisition of all signals generated and / or received by the machine components; S2: Identification, among all acquired signals, of each signal corresponding to a setpoint signal injected into an actuator or a status signal emitted by a sensor; S3: Selection of the signals associated with a first group of actuators and sensors from among the plurality of actuators and sensors, the first group of actuators and sensors defining the production cycle of the machine;S4: Association of signals from the first group of actuators and sensors to establish signal pairs, each signal pair corresponding to a setpoint signal from a predetermined actuator and a feedback signal from a predetermined sensor associated with said predetermined actuator; S5: Identification of a critical path of the production cycle from the actuator-sensor signal pairs, the critical path corresponding to an uninterrupted sequence of consecutive movements of a subgroup of actuators from the first group of actuators and sensors during the entire production cycle of the machine.

2. A method according to claim 1, wherein continuous monitoring S6 of the signals of each of the signal pairs is carried out consecutively to the identification step S5, the method further comprising an anomaly detection step S7, during which the actuators for which the cycle time presents an anomaly are identified.

3. A method according to claim 1 or claim 2, further comprising a step S8 of identifying the type of operation to be performed to compensate for the identified anomaly.

4. A method according to any one of the preceding claims, further comprising a SO step of connection to the production machine so as to allow the collection of signals passing through the machine.

5. A method according to any one of the preceding claims, wherein the acquisition step SI comprises the following substeps: SU: detection of actuator and sensor control PLCs, S12: collection of signals emitted and received by the control PLCs.

6. A method according to any one of the preceding claims, wherein the selection step S3 comprises the following substeps: S31: Observation of the signals sorted during step S2 during a plurality of operating cycles; S32: Elimination of setpoint signals not associated with a status feedback signal; S33: Elimination of signals not having an occurrence for each operating cycle of the machine.

7. A method according to any one of the preceding claims, wherein step S4 comprises the following substeps: S41: Observation of the signals selected during step S3 for a plurality of cycles; S42: Identification of cause-and-effect relationships between the setpoint and state feedback signals to identify state feedback correlated with an actuator setpoint and / or actuator setpoints following state feedback.

8. A method according to any one of the preceding claims, wherein the identification step S5 comprises the following substeps: S51: Identification of setpoint signals that appear to be determining for the critical path of the cycle; S52: Identification of a critical path.

9. A method according to any one of claims 2 to 8, wherein the anomaly detection step S7 comprises the following substeps: S71: Calculation of the time of the critical path steps at each iteration; S72: Calculation of characteristic metrics of the step times, for example a mean, a standard deviation, a variability; S73: Compare the calculated metrics to reference values ​​established during process initialization; S74: Issue an alert if a threshold is crossed.

10. A method according to any one of the preceding claims, wherein the steps are carried out during the operation of the machine without interfering with its operation.

11. Product computer program comprising code data configured to, when executed by a computing unit, implement a method according to any one of claims 1 to 10.

12. Processing module comprising a computing unit capable of executing code data, a memory capable of storing code data, and a communication unit capable of transmitting and receiving signals so as to implement a method according to any one of claims 1 to 10.

Images