Systematic fault diagnosis and tolerant control for station-keeping of a marine vessel was found in [14]。 A structural analysis technique is utilised to generate the residual for the fault diagnosis in [14] and control re-configuration design was applied to alleviate the effects of thruster failure。 For sensor failure, such as temporary loss of sensor data, fault- tolerant sensor fusion was demonstrated in [15]。 These strategies were validated in sea trials。 For PM systems, the tension compensation with feed- forward control in the case of line breakage was suggested in [4]。 The structural analysis was fur- ther applied on the PM problem in [16] and an off-line fault-accommodation strategy was designed with the switch between a bank of controllers in the event of mooring line breakage [7]。 System- wide fault-tolerant control of a PM system was demonstrated in [17] where active fault isolation was used to isolate faults that are otherwise only detectable。 This fault-handling control was based on thruster usage with the reference position main- tained, also when accommodating the effects of line failure。 An optimal set-point chasing algorithm was proposed in [18], where it was suggested to move the reference position to an optimal position where a tension-based cost function achieves a minimum value。 This paper investigates diagnosis of several mooring system failures, including loss of underwa-

ter buoyancy elements, and the paper shows how diagnosis is combined with reliability-based fault- tolerant control。 The diagnosis principles employed in this paper are based on early results reported in [19] which are further refined and combined into a total fault-tolerant control scheme in this paper。

Reliability of a marine structure was initially combined with a control algorithm by [20] for an off- shore drilling platform with a rigid riser。 Extension to a moored structure was dealt with in [21], [22] and [23], where an explicit account of external en- vironment was done through a reliability index。 For mooring systems, [22] utilised the structural reliab- ility index for off-line calculation of thrust needed。 A reliability-based control algorithm was further proposed in [23], making the reliability index an intrinsic part of a controller。 This algorithm could handle one critical mooring line - the one with the smallest reliability index。 The structural reliability index was further proposed into a set-point chasing algorithm in [11] to reduce the risk of mooring line breakage in an extreme environmental situation and fault-tolerance aspects were also considered。 This paper aims to combine the structural reliability in- dex into the fault-accommodation strategy sugges- ted in [18]。 In the case of mooring line breakage, the fault-tolerant control helps to prevent the structural reliability index of the mooring line from a critical index。

Figure 1: Architecture of fault-tolerant control using set- point for fault-handling

Fig。 (1) shows the fault-tolerant control archi- tecture for the PM system。 Once there is a fault in the system, the residual generator diagnoses the fault with the measured signals and control input。 Related residuals for a specific fault would be pro- duced accordingly, which are then delivered to the

change detection algorithm。 The change of the re- siduals is then found by the dedicated detector。 The generated hypothesis flags are supervised to de- cide the remedial actions。 If the remedial decision is made, the design of observer and controller is adjusted based on the system model in the faulty case。 The set-point chasing algorithm further pro- duces the suitable reference position for the PM system to ensure system safety。 This paper con- cerns the change detection and set-point generation illustrated in Fig。 (1)。

The paper is structured as follows。 Residual gen- eration is first obtained through structural analysis of the nonlinear system。 Properties of residuals are then studied and whitening is applied to meet the- oretical pre-conditions。 Section 3 analyses probab- ility distributions for the experimental data and a change detection algorithm is designed。 Fault tol- erant control with an optimal set-point chasing al- gorithm is then described and simulations and ex- perimental results appear in Section 5, showing con- vincing agreement between model and experiments。 Conclusions finish the paper。