WO2020007468A1

WO2020007468A1 - A method for synchronizing programs for simulation of a technical system

Info

Publication number: WO2020007468A1
Application number: PCT/EP2018/068200
Authority: WO
Inventors: Martin Witte
Original assignee: Siemens Aktiengesellschaft
Priority date: 2018-07-05
Filing date: 2018-07-05
Publication date: 2020-01-09

Abstract

The invention refers to a method for synchronizing programs for simulation of a technical system (1, 2, 3), the programs comprising a first simulation program (SP1) and a second simulation program (SP2), wherein the first simulation program (SP1) simulates a number of state variables of the technical system (1, 2, 3) and runs in cycles (cy, cy1, …, cy7), where values of the number of state variables are provided as out-put at the end of a respective cycle (cy, cy1, …, cy7), and wherein the second simulation program (SP2) simulates successive processes for controlling the technical system (1, 2, 3), where the processes (pr, pr1, …, pr7) have variable length in virtual simulated time and where values of a number of control parameters (cp) are provided as output at the end of a respective process (pr, pr1, …, pr7). In the method of the invention, the length of a current process (pr, pr1, …, pr7) at the start of a respective cycle (cy, cy1, …, cy7) is estimated and the length of the respective cycle (cy, cy1, …, cy7) is set to this estimated length. In case that the absolute value of the difference between the set length of the respective cycle (cy, cy1, …, cy7) and the actual length of the current process (pr, pr1, …, pr7) does not fulfill a threshold criterion, the output of the respective cycle (cy, cy1, …, cy7) is discarded for use as an input of a next process (pr, pr1, …, pr7) if the next process (pr, pr1, …, pr7) immediately starts after the current process (pr, pr1, …, pr7).

Description

A method for synchronizing programs for simulation of a tech nical system

The invention refers to a method for synchronizing programs for simulation of a technical system. Furthermore, the inven tion refers to a corresponding computer program product and a corresponding computer program.

When simulating a technical system, there are often used a first simulation program for simulating the state variables of the technical system in successive cycles and a second simulation program for simulating processes for controlling the technical system. The outputs of those simulation pro grams are exchanged between the programs. The simulated pro cesses of the second simulation program are usually of varia ble length in virtual simulated time making it difficult to synchronize the second simulation program with the cycles of the first simulation program.

According to the prior art, a synchronization of the first and second simulation programs is achieved by running the simulations of both programs sequentially. However, this re sults in very long simulation runs. Alternatively, the first and second simulation programs are running parallel and are synchronized at control points of the second simulation pro gram. In case of multiple processes of the second simulation program within one cycle of the first simulation program, the same output of the first simulation program may be read by the second simulation program multiple times. This might lead to unwanted actions of the second simulation program. Fur thermore, events of the first simulation program can get un noticed if the lifetime is only one cycle of the first simu lation program. To address those problems, the simulation programs may have the ability to perform a rollback in virtu al time. This implies that intermediate states have to be saved resulting in a long runtime and high memory resources. It is an object of the invention to provide a method for syn chronizing programs for simulation of a technical system providing a fast simulation with low simulation errors.

This object is achieved by a method according to claim 1. Preferred embodiments are defined in the dependent claims.

The method of the invention synchronizes programs for simula tion of a technical system. In a preferred embodiment, the technical system comprises an automation system, e.g. a pro duction system or a logistic system. Furthermore, the tech nical system may also refer to at least one automatically op erating machine of an automation system.

The programs synchronized by the method of the invention com prise a first simulation program and a second simulation pro gram. The first simulation program simulates a number of state variables of the technical system (e.g. position values of a robot) and runs in cycles, where values of the number of state variables are provided as output at the end of a re spective cycle.

The second simulation program simulates successive processes for controlling the technical system, where the processes have variable length in virtual simulated time and where val ues of a number of control parameters are provided as output at the end of a respective process. The first and the second simulation program interact such that an output of the first simulation program is used as an input for the second simula tion program when starting a new process and that an output of the second simulation program is used as an input for the first simulation program when starting in a cycle.

According to the method of the invention, the length of a current process in virtual simulated time at the start of a respective cycle is estimated and the length of the respec tive cycle in virtual simulated time is set to the estimated length of the current process. If not indicated to the con- trary, indications of time used hereinafter refer to virtual simulated time and not to real time, where real time is the time the simulation programs actually need for their calcula tions .

In case that the absolute value of the difference between the above set length of the respective cycle and the actual length of the current process does not fulfill a threshold criterion, the output of the respective cycle is discarded for use as an input of the next process if the next process immediately starts in virtual simulated time after the cur rent process. The term "immediate start" means that there is no delay between the current process and the next process in virtual simulated time. However, there may be a delay in real time between those processes. The threshold criterion is de fined such that this criterion is fulfilled in case that the absolute value of the difference as defined above is less than a predetermined threshold value. In one variant of the invention, the threshold criterion is also fulfilled when the difference is identical to the threshold value. In all vari ants of the invention, the threshold criterion is not ful filled in case that the absolute value of the difference is higher than the predetermined threshold value.

The method of the invention has the advantage that a substan tially parallel run of the first and the second simulation programs is performed due to the adaptation of the cycle length to the expected process length. Hence, the simulation has a fast runtime. Nevertheless, in case that there is a high difference between the estimated and the current process length, simulation errors are avoided by discarding the out put of the respective cycle.

In a particularly preferred embodiment, the respective cycle is stopped in real time and resumed again after the end of the current process until the length of the respective cycle reaches the length of the current process. Furthermore, the simulation of the first simulation program is suspended in real time from the end of the current process until the end of the respective cycle which is resumed again, whereupon (i.e. after the end of the suspension) the next cycle starts with the output of the current process as input and the next process starts with the output of the respective cycle having been resumed again as input.

The above embodiment enables a temporary sequential pro cessing for synchronizing the simulation programs in case that the actual length of a current process is much higher than the corresponding estimated length.

In another embodiment of the invention, the next process starts immediately in virtual simulated time with the output of the cycle preceding the respective cycle as input in case that the threshold criterion is not fulfilled and the current process ends before the respective cycle. This avoids simula tion errors in case that the actual length of the current process is much shorter than its estimated length.

In an another variant of the invention, the second simulation program simulates a virtual idling time interval (i.e. a time interval where no process runs in virtual simulated time) un til the end of the respective cycle in case that the thresh old criterion is not fulfilled and the current process ends before the respective cycle, whereupon (i.e. at the end of the respective cycle) the next cycle is started with the out put of the current process as input and the next process is started with the output of the respective cycle as input.

In another preferred embodiment of the invention, the length of the current process in virtual simulated time at the start of the respective cycle is estimated as the mean (i.e. the average) of the lengths of a plurality of processes before the current process.

In another variant of the invention, the threshold value is a fixed value throughout the method resulting in a straightfor- ward implementation of the threshold criterion. Nevertheless, the threshold value may also vary at least sometimes from one process to the next.

In a particularly preferred embodiment, the frequency distri bution of the lengths of a plurality of processes before the current process is calculated, where the threshold value is the maximum value of a first value and the second value, the first value being the mean (i.e. the average) of the frequen cy distribution minus a p-quantile of the frequency distribu tion with p < 50%, preferably with p < 10%, and the second value being a q-quantile of the frequency distribution with q > 50%, preferably with q > 90%, minus the mean of the fre quency distribution.

In another preferred embodiment, a lower bound for the set length for the respective cycle is defined, where the lower bound is used for the set length if the estimated length of the current process is lower than the lower bound, and/or an upper bound for the set length of the respective cycle is de fined, where the upper bound is used for the set length if the estimated length of the current process is higher than the upper bound. Those bounds provide an efficient use of computation resources for the simulation.

In another preferred embodiment of the invention, the first simulation program simulates the kinematics of one or more components of the technical system and/or the second simula tion program simulates a programmable logic controller for controlling the operation of the technical system.

Besides the above method, the invention refers to a computer program product with program code, which is stored on a ma chine-readable carrier, for carrying out the method according to the invention or one or more preferred embodiments thereof when the program code is executed on a computer. Furthermore, the invention refers to a computer program with program code for carrying out the method according to the in vention or one or more preferred embodiments thereof when the program code is executed on a computer.

Embodiments of the invention will be described hereinafter in detail with respect to the accompanying drawings wherein

Fig. 1 is a schematic drawing showing the interaction of two simulation programs which are synchronized based on an embodiment of the invention; and

Fig. 2 and Fig. 3 are time charts illustrating two variants of the method according to the invention.

In the following, the invention will be described based on the interaction of two simulation programs designated as SP1 and SP2 in Fig. 1. It is the purpose of those simulation pro grams to simulate a technical system. In the embodiment of Fig. 1, this technical system comprises two robots 1 and 2 used to manipulate an object 0 processed in an automation system as well as a PLC 3 (PLC = Programmable Logic Control ler) used for controlling the robots 1 and 2. This is a very simple example of a technical system. Usually, the technical system to be simulated has a much higher complexity.

The simulation program SP1 is a first simulation program in the sense of the patent claims. In the embodiment of Fig. 1, this simulation program is a rigid body simulator for simu lating the kinematics of the robots 1 and 2. The simulator may be the Mechatronic Concept Designer or the SimCenter Mo tion Software developed by Siemens or any simulator having a controllable cycle (stepsize) . The simulation program SP1 simulates values of state variables sv (particularly position and velocity values) referring to the kinematic state of the robots 1 and 2 and outputs those values to an interface IF for use by the PLC 3 which is simulated by the simulation program SP2. As the states variables sv are simulated, the values of the state variables need to be calculated. There fore, those values are not available at all time. Instead, the values are output in successive cycles cy.

The simulation program SP2 is a second simulation program in the sense of the patent claims. As mentioned above, the pro gram SP2 simulates the (virtual) PLC 3 which successively runs processes pr which perform control tasks of the tech nical system. The processes may refer to the main control task of a PLC. E.g., the processes may be the well-known OBI processes. At the end of a respective process, values of con trol parameters cp are output via the interface IF for use by the first simulation program SP1. Contrary to the cycles cy, the length of the processes pr is not predictable and depends on the current state of the PLC and particularly on the cur rent processing load of the PLC.

The exchange of the outputs (i.e. the values of the state variables sv and the control parameters cp) between the simu lation programs SP1 and SP2 can only take place at the end of respective processes and cycles. Due to the variable length of the processes pr, a problem arises with respect to the proper exchange of the outputs because an output of the state variables sv may occur with a considerable time offset to the end of a process of the PLC resulting in simulation errors. According to the prior art, this problem is overcome by run ning the simulation program sequentially leading to long sim ulation times. Alternatively, it is known to construct the simulation programs such that a rollback from a given time can be performed, resulting in a considerable runtime and memory penalty.

Contrary to the prior art approaches, the method of the in vention implemented in the simulation of Fig. 1 estimates the time length of a current process pr and adapts the length of a cycle cy to this estimation. Furthermore, as indicated in Fig. 1, a time criterion TC based on a threshold value At is used. The threshold criterion TC is fulfilled in case that the difference (i.e. its absolute value) between the length of a respective cycle cy (corresponding to the estimated length of the current process) and the actual length of the current process is smaller than the threshold value At.

In case that the threshold criterion is fulfilled, the simu lation error is low and can be accepted. Contrary to that, in case that the time criterion is not fulfilled, countermeas ures as described below are taken in order to avoid the simu lation errors.

Fig. 2 shows a first variant of the method according to the invention. This figure is a time chart where the horizontal axis represents the real time t, i.e. the time consumed by the simulations. The upper horizontal line of Fig. 2 refers to the time advancement of the simulation program SP1 wherein the lower horizontal line of Fig. 2 as well as in Fig. 3 re fers to the time advancement of the simulation program SP2. The diagonally hatched horizontal bars in Fig. 2 refer to time periods where the corresponding simulation program is running whereas sections of the horizontal lines without bars refer to the advancement of real time with the corresponding simulation being stopped. The time advancement within the horizontal bars is associated with virtual simulated time, i.e. with the time modelled by the simulation programs. Con trary to the diagonally hatched bars, the dotted bar only shown at the lower line of Fig. 3 refers to a virtual time period where the processes are suspended within the simula tion .

In Fig. 2, processes of the simulated PLC 3 corresponding to the process pr in Fig. 1 are designated as prl, pr2, ..., pr7. Analogously, respective cycles corresponding to the cycle cy in Fig. 1 are designated as cyl, cy2, ..., cy6. Furthermore, vertical lines B indicate the ends of cycles and processes. Only some of those lines are designated by the reference nu meral B. Moreover, arrows from top to down in Figs. 2 and 3 refer to outputs of the simulation program SP1 whereas arrows from bottom to top refer to outputs of the simulation program SP2.

In order to implement the method as shown in Fig. 2 as well as in Fig. 3, an estimation of the length of the current pro cess of simulation program SP2 is made. This estimation is used for adjusting the cycle times of the first simulation program SP1. I.e., the cycle times correspond to the estima tions and are variable throughout the simulation. In the em bodiment described herein, the estimation of a current pro cess is based on a moving average calculation of the lengths of several processes preceding the current process. The num ber of processes used for calculating the average can be set to different values, e.g. to ten or more. This moving average is designated in the following as t_pr_mean. Based on the frequency distribution of the lengths of the processes pre ceding the current process, the above mentioned threshold criterion TC is defined based on the following threshold val ue At :

At = max (t_pr_mean-t_pr_5, t_pr_95-t_pr_mean)

In the above formula, the term t_pr_x (0 < x < 100) refers to the x%-quantile of the frequency distribution of the lengths of the processes. This x%-quantile defines a process length t_pr_x where x% of the processes are shorter than t_pr_x. Ac cording to the above formula, a quantile with x = 5 and a quantile with x = 95 is used. However, also other values for the quantiles can be used as long as the lower value is lower 50% and the upper value is higher than 50%.

In the scenario of Fig. 2, the threshold criterion TC is ful filled for the estimated lengths of the processes prl to pr4, where those lengths correspond to the lengths of the respec tive cycles cyl to cy4. As a consequence, the cycles and pro cesses corresponding to each other exchange their outputs.

Due to low deviations between the cycle and process lengths, only short simulation stops between successive cycles or pro- cesses occur. Those simulation stops enable the provision of the output of the slower process/cycle to the process/cycle succeeding the faster process/cycle.

In case of process pr5, the actual process length is much larger than the estimated length corresponding to the cycle length cy5. In other words, the threshold criterion TC is not fulfilled for process pr5. In order to avoid larger errors, the output of cycle cy5 is discarded for use as input of the next process pr6, as indicated by the cross C. As a conse quence, the cycle cy5 is stopped without producing an output and waits until the end of the process pr5. Thereafter, the cycle is resumed as indicated by reference numeral cy5 until the overall cycle length of cy5 and cy5 ' corresponds to the length of the current process pr5. Thereafter, the output of the process pr5 is used as input for the next cycle cy6.

Hence, a sequential processing is done due to the fact that the threshold criterion is not fulfilled for the estimation of the process pr5. Due to this sequential processing, timing errors are avoided.

In the case of process pr6 shown in Fig. 2, the actual length of this process is much shorter than the estimated process length corresponding to the cycle length cy6. I.e., the threshold criterion TC is not fulfilled for the process pr6, either. This also has the consequence that the output of the cycle cy6 is discarded for use as an input of the next pro cess pr7. Instead, the output of the cycle cy5 is used for the process pr7.

The effect of the embodiment of Fig. 2 is that the simulation programs SP1 and SP2 run mostly parallel to each other and that an optimal cycle length is achieved, minimizing the run time requirements of the simulation program SP1. In case that the threshold criterion is fulfilled, the error is at most At which can be chosen appropriately and is small in comparison to the response delay that the processes have anyway. In case that a process is shorter than t_pr_mean-At, the average tim- ing error is smaller than t_pr_mean. If the process length is beyond the upper limit of t_pr_mean+At, the error is 0 be cause of the sequential processing, but performance degrades.

Fig. 3 shows another embodiment of the invention. A time chart analogously to Fig. 2 is shown in Fig. 3. The same es timation of the current process lengths and the same thresh old criterion are used in Fig. 2 and Fig. 3. In case that the threshold criterion is fulfilled and in case that the thresh old criterion is not fulfilled with the estimated process length being shorter than the current process length, the processing of Fig. 3 is identical to Fig. 2. Hence, the time chart of Fig. 3 corresponds to the time chart of Fig. 2 until process pr6. For process pr6, the threshold criterion TC is not fulfilled and the estimated process length is much larger than the actual length of process pr6. Contrary to Fig. 2, the output of cy6 is not discarded for use in process pr7. Instead, the simulation program SP2 simulates a virtual idling time interval id until the end of the cycle cy6 where upon the process pr7 is started with the output of the cycle cy6. I.e., the process pr7 is delayed in virtual time until the process cy6 ends in order to receive the output of the process of cy6. The effect is that for short processes, no additional timing error is created, just as in sequential processing .

In another improvement of the above methods, a lower bound and an upper bound for the estimated process lengths and thus the cycle lengths can be defined. The result is that an over- sampling simulation program SP1 will not slow down too much and that the lower bound of SP1 will create idle times in the simulation program SP2, lowering the CPU usage of the program SP2 and stopping oversampling.

The invention as described in the foregoing has several ad vantageous. Particularly, a parallel co-simulation of two simulation programs is enabled, thus speeding up the simula tion. Furthermore, the cycle length of the simulation program SP1 adapts to the need of the simulation program SP2, improv ing precision and performance. An oversampling of the simula tion program SP2 is avoided, which further improves speed and precision of the parallel co-simulation. No rollback capabil- ities need to be implemented in the simulation programs SP1 and SP2, which saves time and memory and reduces the comple xity.

Claims

Patent Claims

1. A method for synchronizing programs for simulation of a technical system (1, 2, 3), the programs comprising a first simulation program (SP1) and a second simulation program (SP2 ) , wherein

the first simulation program (SP1) simulates a number of state variables (sv) of the technical system (1, 2, 3) and runs in cycles (cy, cyl, ..., cy7), where values of the number of state variables are provided as output at the end of a re spective cycle (cy, cyl, ..., cy7);

the second simulation program (SP2) simulates successive pro cesses for controlling the technical system (1, 2, 3), where the processes (pr, prl, ..., pr7) have variable length in vir tual simulated time and where values of a number of control parameters (cp) are provided as output at the end of a re spective process (pr, prl, ..., pr7);

an output of the first simulation program (SP1) is used as an input for the second simulation program (SP2) when starting a new process (pr, prl, ..., pr7) and an output of the second simulation program (SP2) is used as an input for the first simulation program (SP1) when starting a new cycle (cy, cyl, .··, cyl) ;

characterized in that

the length of a current process (pr, prl, ..., pr7) in virtual simulated time at the start of a respective cycle (cy, cyl,

..., cy7) is estimated and the length of the respective cycle (cy, cyl, ..., cy7) in virtual simulated time is set to the es timated length of the current process (pr, prl, ..., pr7);

in case that the absolute value of the difference between the set length of the respective cycle (cy, cyl, ..., cy7) and the actual length of the current process (pr, prl, ..., pr7) does not fulfill a threshold criterion (TC) , the output of the re spective cycle (cy, cyl, ..., cy7) is discarded for use as an input of a next process (pr, prl, ..., pr7) if the next process (pr, prl, ..., pr7) immediately starts after the current pro cess (pr, prl, ..., pr7), where the threshold criterion (TC) is fulfilled in case that the absolute value of the difference is less than a predetermined threshold value (At) .

2. The method according to claim 1, characterized in that in case that the threshold criterion (TC) is not fulfilled and the current process (pr, prl, ..., pr7) ends after the respec tive cycle (cy, cyl, ..., cy7), the respective cycle (cy, cyl, ..., cy7) is stopped in real time and resumed again at the end of the current process (pr, prl, ..., pr7) until the length of the respective cycle (cy, cyl, ..., cy7) reaches the length of the current process (pr, prl, ..., pr7), where the simulation of the second simulation program (SP2) is suspended in real time from the end of the current process (pr, prl, ..., pr7) until the end of the respective cycle (cy, cyl, ..., cy7) which is resumed again, whereupon the next cycle (cy, cyl, ..., cy7) starts with the output of the current process (pr, prl, ..., pr7) as input and the next process (pr, prl, ..., pr7) starts with the output of the respective cycle (cy, cyl, ..., cy7) having been resumed again as input.

3. The method according to claim 1 or 2, characterized in that in case that the threshold criterion (TC) is not ful filled and the current process (pr, prl, ..., pr7) ends before the respective cycle (cy, cyl, ..., cy7), the next process (pr, prl, ..., pr7) starts immediately with the output of the cycle (cy, cyl, ..., cy7) preceding the respective cycle (cy, cyl, ..., cy7) as input.

4. The method according to claim 1 or 2, where in case that the threshold criterion (TC) is not fulfilled and the current process (pr, prl, ..., pr7) ends before the respective cycle (cy, cyl, ..., cy7), the second simulation program (SP2) simu lates a virtual idling time interval (id) until the end of the respective cycle (cy, cyl, ..., cy7), whereupon the next cycle (cy, cyl, ..., cy7) is started with the output of the current process (pr, prl, ..., pr7) as input and the next pro cess (pr, prl, ..., pr7) is started with the output of the re spective cycle (cy, cyl, ..., cy7) as input.

5. The method according to one of the preceding claims, char acterized in that the length of the current process (pr, prl, ..., pr7) in virtual simulated time at the start of a respec tive cycle (cy, cyl, ..., cy7) is estimated as the mean of the lengths of a plurality of processes (pr, prl, ..., pr7) before the current process (pr, prl, ..., pr7) .

6. The method according to one of the preceding claims, char acterized in that the threshold value (At) is a fixed value throughout the method.

7. The method according to one of claims 1 to 5, character ized in that the threshold value (At) varies at least some times from one process (pr, prl, ..., pr7) to the next.

8. The method according to one of the preceding claims, char acterized in that the frequency distribution of the lengths of a plurality of processes (pr, prl, ..., pr7) before the cur rent process (pr, prl, ..., pr7) is calculated, where the threshold value (At) is the maximum value of a first value and a second value, the first value being the mean of the frequency distribution minus a p-quantile of the frequency distribution with p < 50%, preferably with p < 10%, and the second value being a q-quantile of the frequency distribution with q > 50%, preferably with q > 90%, minus the mean of the frequency distribution.

9. The method according to one of the preceding claims, char acterized in that a lower bound for the set length of the re spective cycle (cy, cyl, ..., cy7) is defined, where the lower bound is used for the set length if the estimated length of the current process (pr, prl, ..., pr7) is lower than the lower bound, and/or an upper bound for the set length of the re spective cycle (cy, cyl, ..., cy7) is defined, where the upper bound is used for the set length if the estimated length of the current process (pr, prl, ..., pr7) is higher than the up per bound.

10. The method according to one of the preceding claims, wherein the technical system (1, 2, 3) comprises an automa tion system or at least one automatically operating machine of an automation system.

11. The method according to one of the preceding claims, wherein the first simulation program (SP1) simulates the kin ematics of one or more components of the technical system (1, 2, 3) and/or the second simulation program (SP2) simulates a programmable logic controller for controlling the operation of the technical system (1, 2, 3) .

12. A computer program product with program code, which is stored on a machine-readable carrier, for carrying out a method according to one of the preceding claims when the pro gram code is executed on a computer.

13. A computer program with program code for carrying out a method according to one of claims 1 to 11 when the program code is executed on a computer.