METHOD FOR TESTING THE RELIABILITY OF COMPLEX SYSTEMS

[0001] The present application claims priority 35 U.S.C. §119 to European Patent Application No. EP 121 80 254.0 (filed on Aug. 13, 2012), which is hereby incorporated by reference in its entirety.

[0002] Embodiments of the invention relate to a method for testing the reliability of complex systems, and includes the evaluation and optimisation of the availability of such systems.

[0003] Investigation of the reliability of complex systems is an important part of the design and production process. To obtain reliable results about reliability it is not sufficient just to perform tests but it is also necessary to plan a test program carefully and also to analyse test results in a way so that statistically reliable results can be obtained.

[0004] Such a method generally comprises steps of: (a.) establishing a system hierarchy of several levels by identifying components of the system; (b.) identifying and assigning failure modes to components; (c.) determining test procedures for failure modes of different components to obtain estimates for the availability of the component; and (d.) determining test procedures for failure modes of different components to obtain estimates for the availability of the component.

[0005] In establishing a system hierarchy of several levels by identifying components of the system. A system is a complex technical entity like a machine, a vehicle or the like consisting of several parts, here designated as components. Components may consist of parts themselves, so that a hierarchy is established in which the system consist of several components of a first level. At least some of these first level components consist of components of a second level and so forth. Here the term subcomponent is used for a part of a component of a higher level. In this step, the system is analysed and a tree of components is identified, representing the system hierarchy.

[0006] Identifying and assigning failure modes to components. All components are analysed to detect possible failures which may cause a malfunction of the system causing unavailability. These failure modes are described in all their aspects to get comprehensive information including: (i.) kind of failure, e.g., wear, (ii.) location of failure, e.g., sliding surface, cause-effect-chain, e.g., high local pressure, high temperatures, partial melting of surface, adhesive wear, crack formation, (iii.) observable stress parameters, and (iv.) precursors.

[0007] Usually some components may have several failure modes, some of them only one. It may occur that a certain component has no failure mode assigned, but in this case at least one subcomponent should have a failure mode assigned.

[0008] Another very important parameter connected with each failure mode is the expected downtime of the system in the case of such failure, including time for detection, delivery of spare parts and repair.

[0009] Availability is the percentage of operation time the component is working. Any failure occurring results in a downtime until a successful repair is finished and this downtime represents the unavailability of the system.

[0010] Determining test procedures for failure modes of different components to obtain estimates for the availability of the component. At least one test is selected. In prior art solutions these tests are designed in a way that it can be shown by statistical calculations that a predetermined lifetime can be achieved. In detail a sufficient duration of the test is determined which can be compared to a certain time of consumer operation (equivalent duration).

[0011] The disadvantage of such procedure is that tests have usually to be designed in a special way to obtain a prognosis of lifetime or availability and that in complicated cases some tests may be too stringent. At the same time it is possible to overlook potential risks.

[0012] In accordance with embodiments, a method is provided which avoids the aforementioned disadvantages and to provide reliable information about the availability of technical systems with the least amount of tests and other efforts for evaluation.

[0013] Now the additional steps will be explained further.

[0014] Identifying a level for allocating availability targets to components and thus establishing a hierarchical decomposition of the system. It is an important step to ensure that the hierarchical tree representing the system is a complete decomposition. That means that every horizontal section through the tree will comprise each component of the system either directly or represented by the respective subcomponents.

[0015] In principle it is possible to divide each component into some subcomponents and so forth. However, there is a reasonable limit for such breakdown. At such limit the function and availability of a component can be assessed in a sufficient way and there is no need for going into more detail. From another point of view this level of maximum depth is chosen in the sense that reasonable tests are available for this level but not below. It is useful to allocate availability targets also on this level.

[0016] It has to be said that in practice this step may be conducted before steps b. and c.

[0017] Allocating availability targets to the components identified in the previous step by qualitative methods considering experience, criticality of failures and novelty of design or material, said availability targets combined result in an overall availability target of the system, and identifying those targets which need to be analysed.

[0018] An important feature of embodiments of the invention is to allocate availability targets to components in such a way that an overall target for the system is achieved which is sufficient to meet the demands of the user of the system. Qualitative in this context means that the allocation is conducted not according to exact mathematical or statistical procedures, but according to a heuristic approach in which all kinds of experience is considered. For example, through this approach the use of a new material or the use of a new supplier of a part will typically result in a need for investigation. Only those components which are well understood, which have a history free from problems and where no new circumstances are observed can be deemed uncritical so that no tests have to be conducted, i.e. they do not have to be analysed. Nevertheless some residual availability target is assigned also to such component. However, it is assumed that this target is met even if no tests are conducted.

[0019] Calculating a projected availability of all components, having targets which need to be analysed, which is an estimate of the most probable value of availability provided that the tests are successful.

[0020] If all the planned tests are carried out without any failure, a demonstrable survival probability and a demonstrable availability can be calculated. These values are conservative lower confidence bounds of the actual survival probability and the actual availability. This means that the true survival probability and the true availability regarding each failure mode is expected to be higher than the corresponding lower bound at a certain confidence level.

[0021] It has been found that aggregation of availability on the basis of demonstrable availability leads to results which are far too conservative i.e. the risk of failure is overestimated to an extent which depends on the number of tests and the distribution of failures.

[0022] As long as no failures appear, a true expected value of the survival probability/availability cannot be calculated. The invention now is based upon the use of a projected availability which substitutes the expected values. These are the measures which will be used for the aggregation of the failure mode related availability values towards the total system availability.

[0023] Identifying components for which availability targets are not met by said projected availability and select another test to improve projected availability or increase the severity of tests already selected until availability targets are met. In this step those components which are identified to be validated obtain additional test efforts either in terms of additional tests or in terms of conducting the same tests more often or longer.

[0024] Calculating an overall availability of the system. Since in all practically relevant cases the percentage of time in which the system may be unavailable is small the overall availability may be obtained by adding up the availability of components.

[0025] It has to be noted that the method in accordance with embodiments of the invention does not make it necessary to design tests for assessment of availability only. The contribution to the calculation of any test which can give some information about possible failure of a component which is readily available can be used.

[0026] Tests are conducted with components or the entire system, wherein it is preferable to test components of the lowest level possible. Each test can produce information about some of the failure modes of the component tested or of subcomponents thereof. It may occur that a certain test does not contribute to assessment of other failure modes.

[0027] Preferably it can be foreseen that different tests are carried out and combined to calculate risk of a specific failure mode. So for example a pulsation rig test, an overload test and a vehicle test can contribute to investigation of the failure mode high cycle fatigue of a connecting plate of a hydraulic system as described later.

[0028] The above explanations relate to a planning of the validation activities during the design and development phase of a system. In this phase tests are planned in order to obtain information about availability under the condition that test results are positive. The present invention relates also to a method in which then tests are actually conducted and depending on the outcome optionally changes in the test program are considered and perhaps other actions are taken.

[0029] Having calculated the overall availability of the system, tests are actually carried out and the method is extended by performing an evaluation of the extent to which additional tests or design changes are needed due to failed tests.

[0035] Initially, embodiments of the method is explained with reference to FIG. 1.

[0036] In step 1, the system to be investigated is defined and an overall availability is determined, which has to be achieved. In step 2, a hierarchical tree is established containing all components of the system representing its structure. In step 3, technical and complementary risks are identified which may affect the product availability and these risks are assigned to the corresponding system tree nodes which represent the components of the system. In step 4, risk assessment is conducted by considering relevant circumstances influencing the risk of failure connected to the individual components and availability targets are allocated on the basis of such assessment.

[0037] Steps 5a1 to 5c2 may be conducted partially parallel to each other so they are represented in three branches.

[0038] In step 5a1, potential priority failure modes are analysed and assigned to the nodes of the system tree. Stress parameters are identified which are relevant for setting up a physics of failure model. In step 5a2, a physics of failure model (PoF) is assigned to each failure mode. In step 5b1, validation test candidates are qualified according to efficiency with regard to cumulative damage intensity. In step 5b2, qualified validation tests are assigned to the corresponding nodes of the system tree. In step 5c1, the consumer usage space is set up, the demands of target market are analysed, weighted Principal Component Analysis (PCA) with measurement data is carried out. In step 5c2, qualified validation tests are located and expected reference duty cycles in the usage space are identified.

[0039] In step 6, demonstrable availability is calculated for each planned test in view of relevant failure modes in view of the reference profile. In step 7, the obtained results are aggregated to values for the projected availability for each tree node.

[0040] In step 1, it is determined whether projected availability is acceptable, i.e., whether it meets the target for each tree node. If the answer is yes, then proceed to step VI. If, however, the answer is no, then proceed to step 8.

[0041] In step 8, from all tree nodes with insufficient availability the node is selected for with the availability target is missed with the largest deviation.

[0042] In step II, it is determined whether any qualified test for the failure modes of the selected node is available. If the answer is yes, then proceed to step 9. If, however, the answer is no, then proceed to step IV.

[0043] In step 9, the top performance test is selected for this node. In step 10, this test is assigned to the lowest possible level of the system tree. In step 11, the demonstrable availability and the projected unavailability are recalculated for all affected tree nodes. In step 12, the expected (projected) availability of the system is determined based on the adjusted test program under consideration of assigned risks and all identified a priori failure modes.

[0044] In step III, it is determined whether projected availability is acceptable, i.e., whether it meets target requirements. If the answer is yes, then proceed to step 14. If the answer is no, then return to step 8.

[0045] In step IV, it is determined whether any other node is available to be selected. If the answer is yes, then proceed to step 13. If the answer is no, then proceed to step 14.

[0046] In step 13, the next node in ranking of unavailability is selected. In step 14, the defined tests are carried out and the results are reported.

[0047] In step V, it is determined whether any observed failures in tests require design changes. If the answer is yes, then proceed to step 15. If the answer is no, then return to step 1.

[0048] In step 15, the test contributions for all relevant failure modes at the tree nodes affected by the design change are reset, and returned to step 8.

[0049] In step VI, it is determined whether tests are already carried out. If the answer is yes, then proceed to step 16. If the answer is no, then return to step 14. In step 16, the scope of tree nodes and failure modes for an early warning system are defined, if necessary. In practical applications it can be expected that boundary conditions do not allow for increased severity or extension of the validation program for each failure mode not validated sufficiently. In particular this could apply for failure modes which have to be re-validated because of failures observed during the execution of the validation program, especially if they require real-time to appear, such as (corrosion, clogging by deposition etc.). In such cases it might be decided that the system is released under the condition that those failure modes which could not be validated sufficiently are monitored with particular care in an early warning system. In step 17, conditioned system release.

[0050] Now embodiments of the invention will be explained using an example which is given for demonstration only and not intended to limit the scope of the invention. The example describes the assessment of a hydraulic system.

[0051] The steps are in the order as conducted, but designated with the same letters as the steps of the claims.

[0052] Initial Step involves definition of objective and availability target. The availability of the hydraulic system HS01 in customer operation has to be ensured as early as possible in the development phase as it is a key product characteristic. The annual operation time per year is 3600 h. The overall availability target is 99% over the first three years in service. Unavailability of the system appears as soon as a failure event occurs which causes system downtime to carry out the required repair action. Eventually required scheduled maintenance actions are carried out in the time periods where the system is not operated (e.g., over night, during the weekend).

[0053] The mean time to repair, i.e., the average time required for a repair or renewal action is assumed to be independent from the affected component as the main job for these actions is the disassembly of the system from the vehicle which is assumed to be required in each case of unavailability. The mean time to repair is assumed to be 18 h. In general the mean time to repair depends on the affected component and on the observed failure mode.

a. Definition of Scope

[0054] The hydraulic system to be considered is represented by the following tree structure:

TABLE 1
Hydraulic System
	Connecting Plate
	Control Disc
	Axial Piston Unit
	Piston
	Cylinder
	Retraction Unit
	Retraction Plate
	Swash Plate
	Plate
	Bearings
	Drive Unit
	Drive Shaft
	Bearings
	Controller Unit
	Control Piston
	Actuator

[0055] Table 1 is a hierarchical tree of a hydraulic system with assembly layer (connecting plate, control disc, . . . ) and component layer (piston, cylinder, . . . ). This hierarchical tree provides a complete decomposition of the hydraulic system. The circumstance of completeness at least on the first level below the root (assembly level) is mandatory since the assessment of system availability is based on the availabilities of assemblies and components.

e1. Identify Risks—Risk Filter

[0056] A product can be seen as successful if the development is finished as planned in time and budget and it provides its intended function over a defined period of time without any repair action. In this context, the term risk describes an increased probability that the product success is not reached.

[0057] All assemblies of the hydraulic system are assessed regarding risks resulting from various categories (technical, time, organizational, production, economic). Since risks are hardly to quantify it is judged per category whether an in increased risk is seen or not. Identified risks set are inherited in the tree as the risk setting although they do not have any impact on the lower level as long as all children of one node inherit the same value.

TABLE 2
	Development Risk d	Production, technology & warranty risk p
	Risk Filter
	Technical Risk	Time Risk	Organizational Risk	Production, Quality, Maintenance	Economic
	Risk Categories
	De-		Load	Abil-		Se-	Boundary	Work		Quality	Mainten-
	sign	Load	Capability	ity	Scope	quence	conditions	Split	Changes	Assurance	ability	Reputation
−Hydraulic System	Y	Y	Y	Y	Y	Y	N	Y	Y	N	Y	Y
	+Connecting Plate	Y	N	Y	N	Y	N	N	N	N	N	N	Y
	+Control Disc	N	N	N	N	N	N	N	N	N	N	N	N
	+Axial Piston Unit	Y	Y	Y	Y	Y	Y	N	Y	Y	N	Y	Y
	+Swash Plate	N	N	N	N	N	N	N	N	N	N	N	N
	+Drive Unit	Y	Y	Y	Y	N	N	N	Y	Y	N	Y	Y
	−Controller Unit	N	Y	N	Y	N	N	N	N	Y	N	Y	N
	Control Piston	N	Y	N	N	N	N	N	N	Y	N	Y	N
	Actuator	N	Y	N	Y	N	N	N	N	Y	N	Y	N

[0058] Table 2 is a risk filter applied to the hydraulic system. As soon as at least one increased risk (Y) per assembly is identified, it will be analysed in more detail. For all assemblies of the hydraulic system except the control disc and the swash plate at least one increased risk has been identified.

e2. Allocation of Availability Targets

[0059] As identified hereinabove, the product is operated 3600 h per year. If one repair action requires 18 h and the availability should be 99%, at maximum 2 failures per system and year are tolerated.

[0060] For the availability assessment, the allowed number of failures per assembly and per component are derived based on the risk assessment of step a. The number of children per tree node is used as a weight factor of the risk to take the system complexity into account.

TABLE 3
					Max.
			Number of	Sum of	allowable #	Availability
Tree Level	Node	Parent	Children	Risks	of Failures	Target
0 (system)	Hydraulic System	NA	6	0	2.000	0.99000
1 (assembly)	Connection Plate	Hydraulic System	0	4	0.125	0.99938
1 (assembly)	Control Disc	Hydraulic System	0	0	0.025	0.99988
1 (assembly)	Axial Piston Unit	Hydraulic System	4	10	1.100	0.99450
1 (assembly)	Swash Plate	Hydraulic System	2	0	0.050	0.99975
1 (assembly)	Drive Unit	Hydraulic System	2	8	0.450	0.99775
1 (assembly)	Controller Unit	Hydraulic System	2	4	0.250	0.99875
2 (component)	Piston	Axial Piston Unit	0	10	0.275	0.99863
2 (component)	Cylinder	Axial Piston Unit	0	10	0.275	0.99863
2 (component)	Retraction Unit	Axial Piston Unit	0	10	0.275	0.99863
2 (component)	Retraction Plate	Axial Piston Unit	0	10	0.275	0.99863
2 (component)	Plate	Swash Plate	0	0	0.025	0.99988
2 (component)	Bearings	Swash Plate	0	0	0.025	0.99988
2 (component)	Drive Shaft	Drive Unit	0	8	0.225	0.99888
2 (component)	Bearings	Drive Unit	0	8	0.225	0.99888
2 (component)	Control Piston	Controller Unit	0	3	0.111	0.99944
2 (component)	Actuator	Controller Unit	0	4	0.139	0.99931

[0061] Table 3 is an allocation of availability targets for the hydraulic system. The maximum allowable number of failures per year for one node is distributed on its child nodes, starting with the overall target for the root. Consequently, all assembly targets (in max. allowed number of failures) sum up to the overall target while all component targets (in max. allowed number of failures) sum up to the target of the corresponding assembly node.

[0062] Each derived max. allowed number of failures is transformed into an availability target by using the annual operation time and the mean time to repair. E.g., for the connecting plate this is 3600*(1−0.125*18/3600)≈0.99938.

b. A-Priori Analysis of Failure Modes

[0063] For all assemblies with increased risk, a failure mode analysis is carried out. In contrary to the well-known FMEA, where functions and mal-functions are identified and assessed with severity, occurrence and detection probabilities, here the focus is on the physical description of the failure phenomena by physics of failure models.

TABLE 4
		Failure Mode			Observable
		acc. to VDI			Stress
Assembly	Component	3822	Location	Cause-effect-chain	Parameters
Connecting	Connecting	Leakage	valve	oil turbulence -	system angel phi,
Plate	plate			acceleration of valve ball -	pressures
				impact seat hammering -	PR_system
Connecting	Connecting	High cycle	high pressure	fatigue load due to	PR_system,
Plate	plate	fatigue	to low pressure	hydraulic pressure
			transient zone	fluctuation
Axial piston	Cylinder	Thermal	sliding surface	high local pressure -	N_pump, T_oil
unit		overload		high temperatures -
				partial melting of surface -
Axial piston	Cylinder	Wear	sliding surface	boundary lubrication -	N_pump, PR_pump,
unit				friction work -	T_oil
				abrasive wear
Axial piston	Piston	High cycle	Interface to	acceleration of piston -	system angle phi,
unit		fatigue	retraction unit	pulsating tension -	N_pump
				fatigue crack
Axial piston	Piston	Wear	interface to	relative movement under	N_pump, PR_pump,
unit			swashplate	high pressure	T oil
Axial piston	Retraction	Wear	interface zone	high speed and pressure of	system angle phi,
unit	unit			pump -	N_pump, T_oil
				local pressure maximum -
Axial Piston	Retraction	High cycle	fillets	pump rotation -	system angle phi,
Unit	plate	fatigue		rotating bending fatigue	N_pump, PR_pump
				load -
				crack formation
Axial Piston	Retraction	Wear	interface to	movement of swashplate -	system angle phi,
Unit	plate		axial piston	radial movement of axial	N_pump, T_oil
				piston -
Drive unit	Drive shaft	High cycle	notches at	torsional fatigue load on	system angle,
		fatigue	shaft gear	shaft -	PR_pump
				stress rising at gears -
				crack formation
Drive unit	Drive shaft	Wear	bearings	lack of hydrodynamic	N_pump, PR_pump,
				lubrication -	T_oil
				surface contact -
Controller	Control	Wear	interface	high forces at high power -	N_pump, system
Unit	piston		piston to	micro welding - wear	angle, PR_pump,
			liner		T_oil
Controller	Actuator	Wear	piston to	dynamic pump operation -	N_pump, system
Unit			lever	high controler movemeent -	angle, PR_pump,
			interface	pressure fluctuation -	T_oil
Controller	Actuator	Leckage	sealings	dynamic pump operation -	N_pump, system
Unit				pressure fluctuation -	angle, PR pump,
				boundary lubricaton -	T_oil
			min.	Corresponding	Repair
	Assembly	Precursors	Sample	Test	Unit	Damage model
	Connecting	controller	pump	pressure	pump	dynamic operation(phi) *
	Plate	problems, rising		fluctuation test		(PR_system)
		control pressure
	Connecting	micro cracks	pump	pulsation rig test	pump	hcf (PR_system)
	Plate
	Axial piston	wear marks,	pump	overload test	pump	N_pump * 1/eta(T_oil)
	unit	surface melting
	Axial piston	wear marks	pump	low speed full load	pump	1/N_pump * PR_pump
	unit			test
	Axial piston	hardening,	piston	pulsation rig test	pistons	hcf ((N_pump * sin(phi))2)
	unit	micro-cracks
	Axial piston	rising	pump	dynamic mass	pump	PR_system * (sin(phi) *
	unit	tolerances		flow test		N_pump)
	Axial piston	wear marks	pump	stationary high	pump	sin(phi) * N_pump
	unit			load test
	Axial Piston	hardening,	axial piston	pulsation rig test	pump	hcf ((N_pump * sin(phi))2)
	Unit	micro-cracks	unit
	Axial Piston	wear marks	pump	dynamic mass	pump	sin (phi) * N_pump
	Unit			flow test
	Drive unit	micro cracks	drive shaft	torsional fatigue	drive shaft	hcf (V * P_pump)
				test
	Drive unit	wear marks,	pump	dynamic mass	pump	starts * dymamic operation
		particles in oil		flow test		(PR_pump)
	Controller	controller	pump	vehicle operation	controller	threshold (phi) *
	Unit	malfunction				PR_system
	Controller	controller	pump	dynamic mass	controller	high PR_system * gradient
	Unit	malfunction		flow test		(dP_pump)
	Controller	oil marks	pump	vehicle operation	controller	gradient (PR_system)
	Unit
indicates data missing or illegible when filed

[0064] Table 4 is list of analysed failure modes for the hydraulic system. For each failure mode identified, the cause-effect chain is identified together with the physical parameters required to build the physics of failure model. The result of this analysis is the physics of failure model to be applied for the description of the failure mode. E.g., the failure mode High cycle fatigue (HCF) for the connecting plate is described with the standard HCF model n*p^k where n is the rotational speed, of the system, p is the hydraulic pressure (the load characteristic in this case) and k the model parameter.

c1. Test Program

[0065] The design verification plan for the hydraulic system is investigated regarding durability tests which may provide any contribution to system validation regarding availability. The following tests have been identified.

TABLE 5
		Number
Test	Planned Duration [h]	of Repetitions
Pulsation Rig Test	20	6
Overload Test	300	2
Vehicle Test	800	3
Pressure Fluctuation Test	2.000	2
Dynamic Load Cycle Test	1.500	2

[0066] Table 5 is a list of planned tests for the validation of the hydraulic system. The pulsation rig test is a component test for the connecting plate and tests against high cycle fatigue. So, this test is only relevant for this combination of component and failure mode while all the other tests are carried out with the complete system and provide therefore potential damage contributions for each node of the system tree. The number of repetitions shows with how many samples each test is carried out. A value of 6 means that 6 connection plates are tested on the rig test, each for 20 h. Also, 2 overload tests, each with a duration of 300 h is planned, etc.

c2. Customer Usage

[0067] The load variability of customer usage has to be investigated regarding all relevant observable variables listed in the failure mode analysis of step b. Since the application of the hydraulic pump in the example is crane operation the variability of customer load profiles is small and it has to be chosen to use only one reference profile for the assessment of the program. In general, more than one customer reference profile could be used.

[0068] FIG. 2 illustrates customer reference profiles with two stress parameters (speed n(t) and pressure p(t) over time t in seconds used for the calculation. This selected customer usage profile serves as a reference for the assessment of the damage contributions of the tests.

[0069] f1. Calculate Availability Demonstration and Projection Per Failure Mode

[0070] For each of the failure modes identified in step b., the corresponding damage model is applied to all relevant tests and customer reference profiles. The illustration for this step is done with the failure modes of the connection plate:

TABLE 6
	Planned Test
				pressure	dynamic
	pulsation	overload	vehicle	fluctuation	load cycle
	rig test	test	test	test	test
Duration [h]	20	300	800	2000	1500
Number of	6	2	3	2	2
		Failure Mode	pulsation			pressure	dynamic
		acc. to VDI	rig	overload	vehicle	fluctuation	load cycle
Assembly	Component	3822	test	test	test	test	test
Connecting	Connecting	Leakage		0	219	5.234	4.800
Plate	plate
Connecting	Connecting	High cycle	25.000	16.575	2.644	8.059	8.184
Plate	plate	fatigue

[0071] Table 6 shows equivalent durations for each single test regarding the identified failure modes for the connecting plate of the hydraulic system. For the failure mode leakage, the pulsation rig test is not relevant as only the plate itself is tested there (without any interfaces to other components and without oil). The physics of failure model is based on dynamic operation of the system. Since the overload test is operated on constant maximum speed, there is no damage contribution regarding dynamic operation. The comparison of the damage for the vehicle test with that of the reference customer profile provides the result that 800 h of the vehicle test (i.e., one test run) is equal to 219 h of customer operation. On the other hand, 2000 h of the pressure fluctuation test is comparable with 5234 h of customer operation.

[0072] For the second failure mode (HCF), large contributions come from the pulsation rig test (as HCF is exactly the purpose of this test) as well as from the overload test. Also the dynamic tests provide substantial contributions.

[0073] For each failure mode, these equivalent durations are used to calculate the demonstrable max. allowable number of failures. Here, principles known from the literature are used. Again for the failure mode leakage of the connection plate the calculation is shown.

[0074] The required input is:

• • [0075] Annual operation time t_a=3600 h
  • [0076] Mean time to repair t_r=18 h
  • [0077] Equivalent durations of n tests t₁, t₂, . . . (here n=7 and t₁=t₂=t₃=219 h, t₄=t₅=5234 h, t₆=t₇=4800 h)
  • [0078] Weibull shape parameter β=2 ( . . . over time linearly increasing risk of failure)
  • [0079] Max. error probability of statistical test α (0.1, default setting, industry standard)

[0080] To determine the demonstrable availability, first the demonstrable survival probability after one year S(t_a), assuming Weibull distributed times to failure is calculated:

S(ta)=α̂(taβ/(t1β+ . . . +t7β))=0.1̂(3600²/(3*219²+2*5234²+2*4800²))≈0.7442

[0081] Thus, the demonstrable availability is A(t_a)=(t_a−t_r*(−ln(S(t_a))))/t_a=(3600−18*(−ln(0.7442)))/3600≈0.9985.

[0082] This means that we expect the survival probability of the connecting plate regarding leakage after 3600 h to be higher than 74.42%. If one repair consumes on average 18 h we expect the corresponding availability to be greater than 99.85%.

[0083] Note that this result is exact only if at maximum one failure per component occurs in the period of consideration. This is a sufficiently good approximation for typical availability targets on component level, as those are close to 100%.

[0084] If all the planned tests are carried out without any failure, the demonstrable survival probability and the demonstrable availability are conservative lower confidence bounds of the actual survival probability and the actual availability. This means that the true survival probability and the true availability regarding each failure mode are expected to be higher than the corresponding lower bound. As long as no failures appear, the expected survival probability/availability cannot calculated but instead, based on the risk numbers, a projected survival probability/availability is provided which substitutes the expected values. These are the measures which will be used for the aggregation of the failure mode related availability values towards the total system availability.

[0085] The projection works as follows. The demonstrable survival probability is a lower α-confidence bound, i.e., the true survival probability is greater than the demonstrated value with probability>(1−α)*100%. This relationship is illustrated in FIG. 3.

[0086] In the left diagram of FIG. 3 the demonstrable survival probability S(ta) is represented over time t in operation hours. The assumption is a Weibull distribution with shape parameter β.

[0087] The right diagram of FIG. 3 shows a Beta distribution representing the projected survival probability S*(ta) depending on the demonstrable survival probability S(ta).

[0088] We assume that the projected survival probability follows a Beta distribution with parameters a and b. The parameter b is set 1 per default and will be decreased if historical information about warranty data indicate that for the related failure mode no problems have been occurred in the past. Decreasing b to a value <1 leads to a density with a vertical asymptote at S(t_a)=1. This means that if the related component/failure mode was inconspicuous in the past, the projected survival probability is increased. The parameter a is driven by the risk assessment and initialized with: exp((max possible total risk value−(total risk value−1))/2).

[0089] This ensures that a always is >1, i.e. the left branch of the beta density goes towards 0. If for those parameters a and b the area of the Beta probability density from 0 to S(t_a) is >α the parameter a is increased until S(t_a) is >α. Then, the projected survival probability S*(t_a) is the expectation value of the finally selected Beta distribution, i.e., S*(t_a)=a/(a+b).

[0090] Again, for the first component/failure mode, (connection plate, leakage), no historical information is available, thus, b=1. As the max. possible risk value is 12 (number of risk categories, compare risk filter above), and the total risk value for the connection plate is 4 (4 risks identified), a is exp((12−4−1)/2)=exp(3.5)≈33.12.

[0091] Since the area under the graph of the beta density left of S(t_a), i.e. in the interval [0; 0.7442], is <α, the parameter a needs not to be adjusted furthermore. So, the assumed Beta distribution has the parameters a=33.12 and b=1. The projected survival probability S*(t_a)=a/(a+b)=33.12/(33.12.+1)=0.9707. The projected availability is like for the demonstrated values A*(t_a)=(t_a−t_r*(−ln(S*(t_a))))/t_a=(3600−18*(−ln(0.9707)))/3600≈0.99985.

[0092] These steps are carried out for all the components and failure modes listed in the table of failure mode analysis. Note that for each failure mode the corresponding equivalent durations, resulting from the physics of failure model, have to be used. The following table contains the results.

TABLE 7
		Failure Mode		demonstrable		Total			projected
		acc. to VDI		survival	demonstrable	risk			survival	projected
Assembly	Component	3822	beta	probability	Availability	number	Beta a	Beta b	probability	Availability
Connecting Plate	Connecting plate	Leckage	2	0.7442	0.99852	4	33.12	1.00	0.9707	0.99985
Connecting Plate	Connecting plate	High cycle	2	0.9935	0.99997	4	357.14	1.00	0.9972	0.99999
		fatigue
Axial piston unit	Cylinder	Thermal	2	0.1903	0.99170	10	1.65	1.00	0.6225	0.99763
		overload
Axial piston unit	Cylinder	Wear	2	0.1215	0.98946	10	1.65	1.00	0.6225	0.99763
Axial piston unit	Piston	High cycle	2	0.2888	0.99379	10	1.86	1.00	0.6501	0.99785
		fatigue
Axial piston unit	Piston	Wear	2	0.0859	0.98773	10	1.65	1.00	0.6225	0.99763
Axial piston unit	Retraction unit	Wear	2	0.1004	0.98851	10	1.65	1.00	0.6225	0.99763
Axial Piston Unit	Retraction plate	High cycle	2	0.3035	0.99404	10	1.93	1.00	0.6591	0.99792
		fatigue
Axial Piston Unit	Retraction plate	Wear	2	0.0656	0.98638	10	1.65	1.00	0.6225	0.99763
Drive unit	Drive shaft	High cycle	2	0.9881	0.99994	8	27.14	0.11	0.9958	0.99998
		fatigue
Drive unit	Drive shaft	Wear	2	0.0100	0.97697	8	4.48	1.00	0.8176	0.99899
Controller Unit	Control piston	Wear	2	0.0457	0.98457	2	90.02	1.00	0.9890	0.99994
Controller Unit	Actuator	Wear	2	0.8153	0.99898	4	33.12	0.67	0.9803	0.99990
Controller Unit	Actuator	Leckage	2	0.7323	0.99844	4	33.12	1.00	0.9707	0.99985

[0093] Table 7 is a demonstrable and projected survival probability and availability for the components and failure modes analysed for the hydraulic system.

f2. Aggregation to Projected (Un)-Availability Per Tree Node

[0094] The list of analyzed failure modes contains the failure modes which were identified as critical. From a statistical point of view it has to be assumed that some failure modes will occur later in customer's hands which were not detected a priori during the development. We estimate the number of total failure modes for one node as n_t=n_f*(1+p/max(1,d)+d/n_f), where n_f is the number of failure modes identified at the node, p is the production risk and d the development risk. Both risk indicators p and d are extracted from the risk filter applied in step 3. There, the development risk d is the number of risks (Y) identified in the first 6 categories in the risk filter and the production risk p is the number of risks (Y) in the last 6 categories.

[0095] For example, for the connecting plate it is n_f=2 (two failure modes identified, compare above), d=3 (three risks (Y) among the first six risk columns in the risk filter) and p=1 (one risk (Y) among the last six columns in the risk filter), thus, we estimate 2*(1+⅓+ 3/2)=34/6≈6 failure modes.

[0096] From failure behavior in general it is known (reference) that the failures which occur in customers' hands are not uniformly distributed over the different potential failure modes, rather they are highly concentrated on a small number of modes, i.e., on the <<problematic>> ones. The concentration measure of Lorenz and Münzner known from the literature is used here and modeled by a simple power function x^k with k≧1. The chart of FIG. 4 shows different concentration functions resulting from k={1, 2, 4, 8, 16}.

[0097] In FIG. 4 the percentage of cumulated failures x^k is depicted over the proportion of failure modes x. The relationship for k=2°=1 is linear, for greater values of k increasingly parabolic. For example, a concentration function with k=8 implies that the most critical 10% of the failure modes cause 60% of all the failures.

[0098] Under the assumption of a concentration k₀ of failures over the n_t failure modes of a component (node), the most critical failure mode, i.e., that with highest failure frequency, at this node may cause (1−((n_t−1)/n_t)^k0)*100% of the at maximum allowed number of failures for the node (compare table with max. allowed # of failures in step 4).

[0099] It bears noting that when planning the tests, it is still unknown which of the failure modes will be the most critical one. However, if for each of the failure modes belonging to a certain node it can be demonstrated that it is better than the most critical assuming a concentration k₀, it can be concluded that the complete node will fulfill the availability target.

[0100] The table of targets in step 4 shows a max. allowable number of failures of 0.12 for the connection plate. With a concentration parameter k0=8 determined with existing warranty data we have for the most critical failure mode a target for the max. allowed number of failures of (1−(5/6)8)*100%≈77% of 0.12 failures, i.e., 0.092 failures. This leads to an availability target of 99.95% for each individual failure mode.

[0101] Aggregation of contribution of several failure modes to a component assessment is performed via the projected number of failures per component. The projected number of failures caused by the nf identified failure modes is estimated by −Σi ln(Si*(ta)), i=1, . . . , nf, assuming that the root causes of the single failure modes are independent. In addition, the number of failures caused by the remaining nt−nf failure modes is taken from the concentration function assuming that the top failure modes are the nf identified ones. Thus, this remaining number of failures is estimated by ((nt−nf)/nt)k0*100% of the maximum allowed number of failures for the node.

[0102] Recall for the connecting plate that we have calculated a projected survival probability of 97.07% regarding the failure mode <<leakage>>. Regarding the second failure mode <<HCF>>, the projected survival probability is 99.72%. Thus, the projected total number of failures for node connecting plate is with of nf=2, nt=6 and k0=8 after ta=3600 h=−ln(0.9707)−ln(0.9972)+((6−2)/6)8≈0.072.

[0103] This principle of aggregation over failure modes is applied in analogous way to all the other components where failure modes have been analysed. For all other components it is assumed that the targets will be reached exactly. In this way, the complete tree is filled with the availability results bottom-up to the root. If the projection does not meet the target for the root it can be seen immediately by node-wise comparison with the corresponding targets where the gaps are located.

g. Increasing Severity of Test Program and h. Calculate Overall Availability

[0104] Table 8 illustrates the bottom-up aggregation of the availability values. The principle as described in the last step is applied for each component and recursivly for each assembly and finally for the system. For each projected number of failures, the projected availability can be calculated again on each step.

TABLE 8
			Max.
			allowed		Projected
			# of	Availability	Number of	Projected	Validation
Tree Level	Node	Parent	Failures	Target	Failures	Availability	Status
0 (system)	Hydraulic System	NA	2.000	0.9900	4.345	0.9783	not sufficient
1 (assembly)	Connection Plate	Hydraulic System	0.125	0.9994	0.072	0.9996	OK
1 (assembly)	Control Disc	Hydraulic System	0.025	0.9999	0.025	0.9999	OK
1 (assembly)	Axial Piston Unit	Hydraulic System	1.100	0.9945	3.587	0.9821	not sufficient
1 (assembly)	Swash Plate	Hydraulic System	0.050	0.9998	0.050	0.9998	OK
1 (assembly)	Drive Unit	Hydraulic System	0.450	0.9978	0.488	0.9976	not sufficient
1 (assembly)	Controller Unit	Hydraulic System	0.250	0.9988	0.123	0.9994	OK
2 (component)	Piston	Axial Piston Unit	0.275	0.9986	0.983	0.9951	not sufficient
2 (component)	Cylinder	Axial Piston Unit	0.275	0.9986	1.026	0.9949	not sufficient
2 (component)	Retraction Unit	Axial Piston Unit	0.275	0.9986	0.608	0.9970	not sufficient
2 (component)	Retraction Plate	Axial Piston Unit	0.275	0.9986	0.969	0.9952	not sufficient
2 (component)	Plate	Swash Plate	0.025	0.9999	0.025	0.9999	OK
2 (component)	Bearings	Swash Plate	0.025	0.9999	0.025	0.9999	OK
2 (component)	Drive Shaft	Drive Unit	0.225	0.9989	0.263	0.9987	not sufficient
2 (component)	Bearings	Drive Unit	0.225	0.9989	0.225	0.9989	OK
2 (component)	Control Piston	Controller Unit	0.111	0.9995	0.050	0.9997	OK
2 (component)	Actuator	Controller Unit	0.139	0.9994	0.073	0.9996	OK

[0105] Table 8 is an aggregation of the projected number of failures and the availability to assess the validation program.

[0106] As long as the mean time to repair is assumed to be equal for all components (as in our demonstration example), it does not matter whether one compares the projected number of failures or the availability with the corresponding target. However, this is typically not the case. Therefore as standard, the availability will be used for the assessment.

[0107] The availability comparison in the first row shows that the target for the complete system is not reached. Thus, the validation program is not sufficient. On the assembly level, it can be seen that the validation program is sufficient for the connecting plate and for the controller unit. Since for the control disc and the swash plate no risks have been identified, it is assumed that these assemblies will reach their availability target even without optimisation the validation program.

[0108] For the axial piston unit and the drive unit, the validation program is not sufficient.

[0109] In table 8 it can be seen that all components of the axial piston unit as well as the drive shaft are not sufficiently validated. The biggest gap appears for the cylinder of the axial piston unit whereas the validation for the drive shaft is rather close to the target.

[0110] Analyzing the existing tests shows that none of them provide an equivalent duration for the cylinder greater 2500 h. Extension of the tests is not possible for practical reasons. So, a new axial piston stress test has to be developed. According to the failure mode consideration should cover HCF, wear and thermal overload for the analyzed locations. If such a test would provide at least 9000 h of equivalent duration for each of the analyzed failure modes, in addition 2 such tests would be required. The biggest gap in the validation of the drive shaft is located at the failure mode “wear”. Only the vehicle test provides validation contribution here. Since the validation is already close to the target it would be sufficient in this case to increase the damage intensity of the vehicle test by a factor 2. This could be reached by increasing the test stress correspondingly.

[0111] Although embodiments have been described herein, it should be understood that numerous other modifications and embodiments can be devised by those skilled in the art that will fall within the spirit and scope of the principles of this disclosure. More particularly, various variations and modifications are possible in the component parts and/or arrangements of the subject combination arrangement within the scope of the disclosure, the drawings and the appended claims. In addition to variations and modifications in the component parts and/or arrangements, alternative uses will also be apparent to those skilled in the art.