# Research on safety modeling and analysis in information fusion system

- 238 Downloads

## Abstract

Avionics system integration is a prominent trend in research and development of civil airplane. It can improve task effectiveness, function efficiency, and resource utilization of system. Information fusion, which includes function information fusion, processing information fusion, and sensor input fusion, is a core process of avionics system integration. Some researches about the impact of information fusion on system safety are done in avionics system which is shown as follows: (1) the concept of Mishap Dilution, Mishap Implication and Mishap Confusion (MD–MI–MC) is first defined in function information fusion of avionics system. (2) The model of multi-source MD–MI–MC is established based on hazard theory. (3) The function fusion of Automatic-Dependent Surveillance–Broadcast (ADS–B) and Traffic Collision Avoidance System (TCAS) is used as a typical example to analyze fusion system during the aircraft climbing or landing state. In this paper, the concept and model of multi-source MD–MI–MC are proposed for safety analysis of integrated avionics system. A fusion model with a variable sampling Variational Bayesian–Interacting Multiple Model (VSVB–IMM) algorithm is used to analyze. At last, a set of theory system and evaluation standards including the positive and negative earning analyses are built based on the presented MD–MI–MC theory and mechanism of integrated avionics.

## Keywords

System safety Mishap dilution (MD) Mishap implication (MI) Mishap confusion (MC) Information fusion## 1 Introduction

With the advancement of science and technology, avionics system is developing towards complexity and integration. Avionics system integration [1] which is a system technology can improve task effectiveness, function efficiency, and resource utilization of system. Information fusion [2], which includes function information fusion, processing information fusion, and sensor input fusion, is a core process of avionics system integration.

For avionics system integration, the composition and structure of avionics system integration technology are described by Wang [3]. The impact of spatial and temporal integration choices on the communication performance is brought to light by Badache [4] to realize optimized system organization for avionics system integrating. The avionics software safety [5] is very important to warrant the safe operation of an avionics system. Many safety requirements are imposed by various standards and industrial regulations that must be met by the avionics software.

For information fusion, multisensor fusion and integration [6] is a rapidly evolving research area. The advantages gained through the use of redundant, complementary, or more timely information in a system can provide more reliable and accurate information. ADS–B [7] is known as the foundation of free flight by Federal Aviation Administration (FAA). TCAS [8] is airborne traffic alarm and collision avoidance system which is independent of air-traffic control on the ground. The fusion of ADS–B and TCAS data can improve the prediction accuracy of TCAS system [9], increase the rate of true alarm, and decrease the rate of false and missed alarm. There are many benefits in airspace alarm accuracy and flight safety by combining with TCAS II and ADS–B system.

However, when the integration of avionics system brings benefits, it will also lead to an increase in system complexity. The system failure in integrated avionics system may become dilution, implication, and confusion. Some researches focus on safety analysis methods. Ding [10], who combines the advantages of support vector machine theory (SVM) and genetic algorithm (GA), proposes a smart fusion prediction method GASVM to evaluate the safety of the remaining life of the electronic device (RUL) prediction. Shen [11] introduced the systematic method of constructing security cases for the system through the Goal Structuring Notation (GSN), which opened up new research ideas for the certification of avionics systems.

In addition, for the safety design of the system, Xu Xianliang et al. [12] designed a security-centric IMA software architecture design theory, using the dangerous scene to evaluate the security of the IMA structure. From the perspective of airworthiness, the security issues of the IMA system were studied by CAI [13]. The shortcomings of traditional airworthiness certification and the main difficulties in integrating airworthiness certification of modular avionics are discussed. In addition, the problem between the authentication method and development practice is solved.

In this paper, the concept and model of multi-source MD–MI–MC are proposed for safety analysis of integrated avionics system. A fusion model with a VSVB–IMM algorithm is used to analyze. A theory system and evaluation standards are tested by real-engineering process of integrated surveillance avionics system. The positive and negative benefits of MD–MI–MC can be obtained quantitatively based on the presented integrated avionics theory and mechanism.

## 2 Theory of multi-source MD–MI–MC

### 2.1 Hazard theory

Figure 1 shows that the hazards and mishaps are two independent states of the same phenomenon, and the two are connected through a state transition. These states can also be considered as prior and post-mortem states. The hazard is an event that may occur in one end of the state spectrum, and in the other end of the spectrum is an event that has actually occurred. A suitable metaphor is water, which has its own molecular formula, but can be liquid or solid, and the key to its conversion is temperature.

- (1)
Specific hazard components involved.

- (2)
Mishap risk represented by hazard components.

The hazard components are the items that constitutes the hazard, and the mishap risk is the probability of the mishap occurring and the severity of the loss caused by the mishaps.

The probability and severity of accidents can be analyzed both in qualitative and quantitative way. Time is introduced into the concept of danger as a factor in probability calculation. The definition of a hazard component is more complex. The hazards only include those elements that must be included, and are sufficient to make the mishap happen.

- (1)
Hazardous element (HE). This is the basic resource that creates incentives for harm, such as the explosives used in the system.

- (2)
Initiating mechanism (IM). This is the trigger that causes the occurrence of the state transition.

- (3)
Target and threat (T/T). This part refers to the person or thing that is vulnerable, and describes the severity of the accident. It can also be understood as the consequences of the accident and the series of injuries and losses it anticipates.

### 2.2 Modes and mechanisms of multi-source MD–MI–MC

#### 2.2.1 Modes and mechanisms of MD

In the functional information fusion, MD leads to risk source composition and strength weakened. In addition, it also leads to the probability of safety accidents decreased.

Functional information fusion may lead to the hazardous element weakened and the triggering probability lower, so that the mishap will not occur or the seriousness of mishap is significantly reduced. The above process is called MD.

#### 2.2.2 Modes and mechanisms of MI

The MI in the fusion of functional information results in the non-characterized or non-existent hazard sources. It can lead to the occurrence of a security incident, where the hazard cannot be found.

#### 2.2.3 Modes and mechanisms of MC

The modes and mechanisms of MC are shown in Fig. 6. The small circles with color represent the harmful elements, trigger mechanisms, and results, respectively. The area of the circle indicates the probability of its occurrence. Two phenomena of MC are mishap not happen and mishap happen, but hazard source confused is shown in Fig. 6.

### 2.3 Models of multi-source MD–MI–MC

- (1)
A mishap can be a source of danger, but the source of the danger is not just a mishap. Only consider the source of hazard, and do not consider human factors, environments, etc. (man–machine loop).

- (2)
The initiating mechanism is random and can be considered as a random variable. In this sense, a mishap caused by a hazard can be seen as a specific probability of a security incident.

*G*(

*X*,

*Y*) = 0, where:

- (1)
*X*which is the state parameter of the system may be multidimensional in practical problems. In mathematical model, it is simplified as a one-dimensional parameter. In addition, in this model,*X*can be understood as system-state instability. The larger*X*is, the higher the instability is, and the easier it is to trigger a security mishap. - (2)
*Y*is the Initiating Mechanism. It is a random variable. It is also multidimensional in practical problems. There exists a probability density function*P*(*Y*). It is simplified to 1 dimension in mathematical modeling, and in the simplified model, without loss of generality. We assume that the larger y, the more difficult the accident. Mathematically, it is expressed as a monotonically decreasing function, whose probability density function*P*(*Y*) is*Y*.

*G*(

*X*,

*Y*) is also called the mishap trigger decision function. As shown in Fig. 7, when

*G*(

*X*,

*Y*) > 0, there is a security accident, and when

*G*(

*X*,

*Y*) < 0, no security incident occurred. Function

*G*(

*X*,

*Y*) is the core of the model. It describes the relationship of hazard sources, triggering factors, and security events. The specific form of this function is determined by the specific physical problem.

*X*,

*Y,*and

*G*are the three elements in the model of the mishap trigger decision function (corresponding to the hazard triangle).

*N*independent sources of information that can be used independently (independently distributed) before functional information fusion. When the

*i*th information source is used alone, it can be described according to the previous model system:

*X*

_{1},…

*X*

_{N};

*Y*

_{1},…

*Y*

_{N};

*G*

_{1},…,

*G*

_{N}) to (

*X*

_{f},

*Y*

_{f},

*G*

_{f}):

The model is a general model of functional information fusion for avionics systems.

#### 2.3.1 MD model

#### 2.3.2 MI model

#### 2.3.3 MC model

*X*

_{i},

*Y*

_{i}, and

*G*

_{i}are given, the fused

*X*

_{f},

*Y*

_{f}, and

*G*

_{f}become random variables. The approximate ranges and distributions can be inferred, but specific values cannot be determined. Suppose we know that the fused

*X*

_{f},

*Y*

_{f}, and

*G*

_{f}, the state of each subsystem cannot be pushed back as well. The uncertain correspondence due to fusion can be understood as confusion. When confusion occurs, the specific benefits of its integration are also uncertain, as shown in Fig. 10.

## 3 Cases of multi-source MD–MI–MC

### 3.1 MD–MI–MC verification method

In the aircraft climbing or landing process, TCASII and ADS–B-fused system are taken as an example to study safety of information fusion. According to the functional information fusion model, the impact on the security and warning modes of TCASII and ADS–B is summarized. In this case, *X* is defined as the state of subsystem such as ADS–B or TCAS data without fusion process and *X*′ is the output of fused system. Based on the safety feature model of the hazard triangle, by comparing the posterior probability \( P\left( {R|X} \right) \) before fusion and the posterior probability \( P\left( {R|X^{{\prime }} } \right) \) after fusion, the positive and negative benefits of multi-source MD–MI–MC are indicated. The position parameter \( X_{i} = \left\{ {x_{i} } \right\},i = 1,2, \ldots ,L \) can be obtained by the subsystem sensors such as ADS–B and TCAS. After fused by the fusion mechanism *F*, the output of fused system is *X*′ which is defined as *X*′ = *F*(*X*_{i}). The output of system is *R*_{j} in which *R*_{j} = 0 means no alarm and *R*_{j} = 1 means alarm. The posterior probability \( P\left( {R|X} \right) \) can be obtained as the following equation:

*P*(

*X*) is the prior distribution of

*X*which can be obtained by mishap mechanism or experimental simulation. \( P\left( {X|R_{j} } \right) \) is the probability distributions of

*X*in alarms or no alarms status. \( P\left( {R_{j} |X} \right) \) is the probability of alarms or no alarms when the input condition is

*X*.

*L*= 2. The process of fusion is in Eqs. (5, 6):

*X*

_{i}is independent. ADS–B system uses GPS modules to achieve its position information and TCAS system uses Mode S to achieve its position information. The prior distribution of

*X*′ in fusion is \( P\left( {X^{{\prime }} } \right) = \mathop \sum \limits_{i = 1}^{L} A_{i} P(X_{i} ) \). The posterior probability \( P\left( {R|X^{{\prime }} } \right) \) can be obtained as the following equation:

*X*′ which can be obtained by mishap mechanism or experimental simulation. \( P\left( {X^{{\prime }} |R_{j} } \right) \) is the probability distributions of

*X*′ in alarms or no alarms status. \( P\left( {R_{j} |X^{{\prime }} } \right) \) is the probability of alarms or no alarms when the input condition is

*X*′.

- (1)
MD

The experiment environment is configured into the alarm area. The posterior probability \( P\left( {R_{j} = 1|X} \right) \) before fusion and the posterior probability \( P\left( {R_{j} = 1|X^{{\prime }} } \right) \) can be obtained as Eqs. (4) and (7). When \( P\left( {R_{j} = 1|X} \right) < P\left( {R_{j} = 1|X^{{\prime }} } \right) \), it means that the leak alarm rate reduced. It is obvious that information fusion brings positive benefits to the system.

The experiment environment is configured into the no alarm area. The posterior probability \( P\left( {R_{j} = 0|X} \right) \) before fusion and the posterior probability \( P\left( {R_{j} = 0|X^{{\prime }} } \right) \) can be obtained as Eqs. (4) and (7). When \( P\left( {R_{j} = 0|X} \right) < P\left( {R_{j} = 0|X^{{\prime }} } \right) \), it means that the false alarm rate reduced. It is obvious that information fusion brings positive benefits to the system.

- (2)
MI

The experiment environment is configured into the alarm or no alarm area. The posterior probability \( P\left( {R_{j} |X} \right) \) before fusion and the posterior probability \( P\left( {R_{j} |X^{{\prime }} } \right) \) can be obtained as Eqs. (4) and (7). When \( P\left( {R_{j} |X} \right) = P\left( {R_{j} |X^{{\prime }} } \right) \), it means that the source of mishap is not characterized and does not appear. It is obvious that information fusion brings positive or negative benefits to the system.

### 3.2 Flight scenario definition for simulation cases

### 3.3 Verification for MD–MI–MC

*T*= 1 s. Flight position: 98°00′00″E, 29°00′00″N, 4502 m height. It climbs from 300 m height and then cruises at constant height. The initial position of Intruder: 106°00′00″E, 29°00′00″N, 300 m height. TCAS’s observation noise standard deviation is 50 m/s, and ADS–B’s observation noise standard deviation is time-varying. Flight path simulation is shown in Fig. 15.

Statistics of early alarm and hysteresis alarm during TA (35–45S), RA (< 35S) in 200 experiments

Alarm type | System categories | ||
---|---|---|---|

TCAS | ADS–B | Fused system | |

False alarm (TA) (frequency) | 667 | 335 | 248 |

Leak alarm (TA) (frequency) | 747 | 389 | 256 |

False alarm (RA) (frequency) | 778 | 247 | 131 |

Leak alarm (RA) (frequency) | 790 | 254 | 120 |

#### 3.3.1 MD in 16–40 nm area

Figure 12 shows that MD can be found in 16–40 nm distance between this aircraft and the intruder. In this area, the decision is made by ADS–B and TCAS fusion system. The TCAS, ADS–B, and fusion system are chosen to be three independent processes to verify MD by injecting into the method proposed in “MD–MI–MC verification method”.

*N*

_{TCAS}. Independent simulations’ number of ADS–B is

*N*

_{ADS–B}. Independent simulations’ number of fusion system is

*N*

_{fusion}. According to Eqs. (4), (8), (9), and (10), the value of \( P\left( {R_{j} = 1|X_{\text{TCAS}} } \right) \) can be obtained:

According to Table 1, the data are substituted into the Eqs. (4), (8), (9), (10), (11), (12), (13), (14), (15), and (16). The results are shown as follows: \( P\left( {R_{j} = 1|X_{\text{TCAS}} } \right) \approx 0.49963 \), \( P\left( {R_{j} = 1|X_{\text{ADS - B}} } \right) \approx 0.49965 \), and \( P\left( {R_{j} = 1|X_{\text{fusion}} } \right) \approx 0.49977 \). In this flight scenario, \( P\left( {R_{j} = 1|X_{\text{TCAS}} } \right) \) < \( P\left( {R_{j} = 1|X_{\text{fusion}} } \right) \) and \( P\left( {R_{j} = 1|X_{\text{ADS - B}} } \right) \) < \( P\left( {R_{j} = 1|X_{\text{fusion}} } \right) \).

According to theory defined in 3.1, it means that the leak alarm rate reduced. It is obvious that information fusion brings positive benefits to the system. In 16–40 nm distance between this aircraft and the intruder, MD is found.

#### 3.3.2 MC in 16–40 nm area

Statistics of early alarm and hysteresis alarm during TA (35–45S), RA (< 35S) in 300 experiments

Alarm type | System categories | ||
---|---|---|---|

TCAS | ADS–B | Fused system | |

False alarm (TA) (frequency) | 647 | 1011 | 448 |

Leak alarm (TA) (frequency) | 734 | 614 | 492 |

False alarm (RA) (frequency) | 827 | 3058 | 531 |

Leak alarm (RA) (frequency) | 825 | 1147 | 526 |

Statistics of leak alarm TA (35–45S) in 300 experiments

Number of experiments | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
---|---|---|---|---|---|---|---|---|---|---|

TCAS | 747 | 553 | 674 | 890 | 459 | 386 | 961 | 768 | 662 | 658 |

ADS–B | 389 | 249 | 423 | 186 | 369 | 353 | 502 | 276 | 441 | 363 |

Fusion system | 256 | 301 | 189 | 236 | 301 | 469 | 613 | 452 | 376 | 421 |

According to Table 3, \( P\left( {R|X} \right) \) and \( P\left( {R|X^{{\prime }} } \right) \) can be obtained from the Eqs. (4), (8), (9), (10), (11), (12), (13), (14), (15), and (16). Comparing with the posterior probability \( P\left( {R|X} \right) \) before fusion and the posterior probability \( P\left( {R|X^{{\prime }} } \right) \) after fusion in ten experiments, there are only seven times that \( P\left( {R_{j} = 1|X_{\text{TCAS}} } \right) \) < \( P\left( {R_{j} = 1|X_{\text{fusion}} } \right) \) and 4 times that \( P\left( {R_{j} = 1|X_{\text{ADS - B}} } \right) \) < \( P\left( {R_{j} = 1|X_{\text{fusion}} } \right) \). Therefore, the relationship between \( P\left( {R|X} \right) \) and \( P\left( {R|X^{{\prime }} } \right) \) is uncertain. The source of failure mode is confused and uncertain. Therefore, in 16–40 nm distance between this aircraft and the intruder, MC is found.

#### 3.3.3 MI in 4–16 nm area

Figure 12 shows that MD can be found in 4–16 nm distance between this aircraft and the intruder. In this area, the decision is ADS–B subsystem priority. The ADS–B and fusion system is chosen to be two independent processes to verify MI by injecting into the method proposed in “MD–MI–MC verification method”.

According to Eqs. (4), (11), (12), and (13), the value of \( P\left( {R_{j} = 1|X_{\text{ADS - B}} } \right) \) can be obtained. Fusion system is priority to use ADS–B data when both ADS–B and TCAS data valid. Therefore, \( P\left( {R_{j} = 1|X_{\text{fusion}} } \right) = P\left( {R_{j} = 1|X_{\text{ADS - B}} } \right) \). It means that the source of TCAS failure is not characterized and does not appear. In 4–16 nm distance between this aircraft and the intruder, MI is found.

## 4 Conclusions

- (1)
The concept of MD–MI–MC is first defined in functional information fusion of avionics system.

The integrated technology brings benefits to avionics system, and meantime, it also brings MD–MI–MC and many uncertainly problems. MD in functional information fusion causes the mishap source composition and strength weakening which can also lead to the occurrence probability of accident decrease. In this case, the accident may not happen and the mishap is in diluted state. MI in functional information fusion causes the mishap source not be characterized and not revealed. In this case, the mishap source cannot be found when the accident occurs. MC in functional information fusion causes the mishap source in random concurrency or disappearance. In this case, the mishap source is confused and uncertain.

- (2)
The model of multi-source MD–MI–MC is established based on hazard theory.

Based on the hazard theory, hazards are composed of three basic components: Hazardous Element (HE), Initiating Mechanism (IM), and Target and Threat (TT). The MD–MI–MC model is built, respectively, in the abstract concept based on the general model of functional information fusion of avionics system. In addition, the model can describe the MD–MI–MC process quantitatively with the mishap initiating decision function

*G*(*X*,*Y*). - (3)
The functional fusion of Automatic-Dependent Surveillance–Broadcast (ADS–B) and Traffic Collision Avoidance System (TCAS) is used as a typical example to analyze fusion system during the aircraft climbing or landing state.

Based on the model of multi-source MD–MI–MC, a set of theories system is built to analyze the fusion of ADS–B and TCAS model. A fusion model with a VSVB–IMM algorithm in integrated surveillance avionics system is used to integrated display and the fusion system can significantly increase the range of traffic collision avoidance. The warning mode and the safety impact of ADS–B and TCAS II fusion system are proposed. According to Hazard Triangles Theory (HTT), the positive and negative benefits of MD–MI–MC are obtained from the theory system in functional information fusion system.

In this paper, the concept and model of multi-source MD–MI–MC are proposed for safety assessment of integrated avionics system. First, it is based on the general functional information fusion model, and then, the VSVB–IMM fusion method of ADS–B and TCAS II in integrated surveillance avionics system is used to analyze. At last, a theory system and evaluation standards are tested by real-engineering process of integrated surveillance avionics system. The positive and negative benefits of MD–MI–MC can be obtained quantitatively based on the presented integrated avionics theory and mechanism.

## Notes

### Acknowledgements

This paper is sponsored by National Program on Key Basic Research Project (2014CB744903), National Natural Science Foundation of China (61673270), Shanghai Pujiang Program (16PJD028), Shanghai Industrial Strengthening Project (GYQJ-2017-5-08), Shanghai Science and Technology Committee Research Project (17DZ1204304) and Shanghai Engineering Research Center of Civil Aircraft Flight Testing.

## References

- 1.Wang G, Gu Q, Wang M, Wang Y (2014) Research on integrated technology and model in avionics system. In: Digital avionics systems conference, pp 4A2-1–4A2-11Google Scholar
- 2.Mahler RP (2014) Advances in statistical multisource-multitarget information fusion. Artech House, NorwoodzbMATHGoogle Scholar
- 3.Wang G (2012) Integration technology for avionics system. In: Digital avionics systems conference, pp 7C6-1–7C6-9Google Scholar
- 4.Badache N, Jaffres-Runser K, Scharbarg JL, Fraboul C (2013) End-to-end delay analysis in an integrated modular avionics architecture. In: Emerging technologies and factory automation, pp 1–4Google Scholar
- 5.Wu J, Yue T, Ali S, Zhang H (2013) Ensuring safety of avionics software at the architecture design level: an industrial case study. In: International conference on quality software, pp 55–64Google Scholar
- 6.Luo RC, Ying CC, Chen O (2002) Multisensor fusion and integration: algorithms, applications, and future research directions. IEEE Sens J 2(2):107–119CrossRefGoogle Scholar
- 7.Kunzi F (2011) ADS-B benefits to general aviation and barriers to implementation. Massachusetts Institute of Technology, CambridgeGoogle Scholar
- 8.Kuchar JE, Drumm AC (2007) The traffic alert and collision avoidance system. Linc Lab J 16(2):277Google Scholar
- 9.Wang Y, Xiao G, Dai Z (2017) Integrated display and simulation for automatic dependent surveillance-broadcast and traffic collision avoidance system data fusion. Sensors 17(11):2611CrossRefGoogle Scholar
- 10.Ding C, Xu J, Xu L (2013) ISHM-based intelligent fusion prognostics for space avionics. Dialogues Cardiovasc Med Dcm 29(1):200–205Google Scholar
- 11.Shen X, Bai Y (2014) Architectural considerations in integrated modular avionics (IMA) system safety case construction. IEEE Aerosp Electron Syst Mag 29(10):26–33CrossRefGoogle Scholar
- 12.An X (2012) Safety-centered architecture design method for IMA software. Comput Sci 39(3):128Google Scholar
- 13.Cai Y, Wang Z, Ou X, Zhu L (2011) Approach civil integrated modular avionics airworthiness certification by iterative incremental certification process. In: International conference on information science and engineering, pp 148–151Google Scholar
- 14.Ericson CA (2005) Hazard analysis techniques for system safety. Wiley, HobokenCrossRefGoogle Scholar
- 15.Deng Zili (2005) Optimal estimation theory with application: modeling, filtering, information fusion estimation, Chap. 6. Harbin Institute of Technology Press, Harbin, pp 377–385Google Scholar