# US 2022/0093279 A1 · Interactive diagnostics of connected appliance devices

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2022/0093279 A1
- Document ID: US20220093279A1
- Status: Published application
- Kind: A1
- Application number: 17/028,129
- Filed: 2020-09-22
- Published: 2022-03-24
- Earliest claimed: 2020-09-22
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Radu Marinescu; Akihiro Kishimoto
- Web page: https://patentgrep.com/patent/US20220093279A1

## CPC

- G06N 5/02: Knowledge representation; Symbolic representation
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- G16Y 40/10: Detection; Monitoring
- G16Y 40/20: Analytics; Diagnosis
- G16Y 40/35: Management of things, i.e. controlling in accordance with a policy or in order to achieve specified objectives
- G06N 3/04: Architecture, e.g. interconnection topology
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16Y: INFORMATION AND COMMUNICATION TECHNOLOGY SPECIALLY ADAPTED FOR THE INTERNET OF THINGS [IoT]

## Abstract

A historical log of a plurality of a type of Internet-connected appliances is collected. a dynamic probabilistic graphical model of the time-lined interactions between subsystems and components of the appliances is generated. The probabilistic graphical model is trained by applying machine learning techniques and applying expert knowledge. A natural language processing (NLP) conversational user interface for diagnostic queries regarding operational errors is provided, and responsive to receive a diagnostic query in a conversational format from a user regarding a particular type, make, and model of an appliance of the plurality of the type of Internet-connected appliances, a diagnostic response is provided to the user interface using the natural language processing conversational format.

## Claims

### Claim 1

A method for automated diagnostics of appliances, the method comprising:
collecting, respectively, by one or more processors, from a plurality of a type of network-connected appliances, a historical log of sensor data and events recorded over a timeline;
generating, by one or more processors, a dynamic probabilistic graphical model of the type of network-connected appliances, wherein generating the dynamic probabilistic graphical further comprises:
mapping, by one or more processors, data of a sensor of a plurality of sensors monitoring operational activity of the type of network-connected appliances to a variable of a plurality of variables, respectively;
mapping, by one or more processors, an event of the type of network-connected appliances, within the timeline, to a set of variables and corresponding sensor values associated with the timeline;
applying, by one or more processors, a structure learning algorithm to the variables and timeline corresponding to the event; and
applying, by one or more processors, expert knowledge of an interaction between one or more subsystems and components of the network-connected appliances, respectively, as supervised machine learning techniques; and
responsive to receiving a diagnostic query regarding a diagnosis of the type of network-connected appliance, generating, by one or more processors, a response based on the diagnosis.

### Claim 2 (depends on 1)

The method of claim 1, further comprising:
providing, by one or more processors, a natural language processing (NLP) conversational user interface for diagnostic queries regarding the operational activity of the type of network-connected appliance.

### Claim 3 (depends on 1)

The method of claim 1, wherein the plurality of the type of appliances perform a same operational activity.

### Claim 4 (depends on 1)

The method of claim 1, wherein the event of the type of network-connected appliances, respectively, includes one or more selected from a group consisting of: malfunction of the appliance, a state of condition of the appliance, a condition of a component of the appliance, a condition of a subsystem of the appliance, and an estimated prediction of a malfunction at a future time.

### Claim 5 (depends on 1)

The method of claim 1, further comprising:
training, by one or more processors, the dynamic probabilistic graphical model by applying machine learning techniques to the historical log of operational activity collected from the plurality of the type of network-connected appliances.

### Claim 6 (depends on 1)

The method of claim 1, wherein mapping of the data of a sensor of the plurality of sensors monitoring operational activity of the type of network-connected appliances to a variable of the plurality of variables, respectively, includes associating a time reference of the data of the sensor from the timeline.

### Claim 7 (depends on 1)

The method of claim 1, wherein the type of network-connected appliance includes a particular manufacturer of the appliance.

### Claim 8

A method for applying a natural language query to a probabilistic graphical model, the method comprising:
receiving, by one or more processors, a natural language query regarding an event of an appliance;
receiving, by one or more processors, sensor data from the appliance including a timeline associated with the sensor data;
creating, by one or more processors, a set of observations in which an observation pairs a variable of the appliance with a value from a corresponding sensor of the sensor data from the appliance, wherein variables of the set of observations correspond to respective variables in a probabilistic graphical model, and the respective variables of the probabilistic graphical model include respective values;
mapping, by one or more processors, the natural language query regarding the event of the appliance to a logical expression of the probability of the event of the appliance included in the natural language query, given the value of the respective variables of the set of observations associated with the natural language query;
generating, by one or more processors, a conditional probability query by mapping the respective variables to the set of observations included in the natural language query; and
applying, by one or more processors, the conditional probability query to the probabilistic graphical model using an inference algorithm.

### Claim 9 (depends on 8)

The method of claim 8, further comprising:
determining, by one or more processors, whether the probability of the event of the appliance exceeds a pre-determined threshold; and
responsive to the probability of the event of the appliance exceeding the pre-determined threshold, returning, by one or more processors, a response indicating a positive likelihood of the event.

### Claim 10 (depends on 9)

The method of claim 9, wherein the positive likelihood of the event of the appliance is expressed as a probability of the event of the appliance occurring within a designated time frame.

### Claim 11 (depends on 8)

The method of claim 8, wherein a neural network model maps natural language terms of the natural language query of components and a state of the appliance to a conditional probability expression.

### Claim 12 (depends on 8)

The method of claim 8, further comprising:
receiving, by one or more processor, a query a natural language query regarding a prediction of an event of an appliance;
applying, by one or more processor, a dynamic Bayesian network (DBN) inference algorithm in which a transition function depends only on an immediately preceding timeslice of the timeline of sensor data; and
predicting, by one or more processors, a response to the conditional probability query, based on the inference algorithm.

### Claim 13 (depends on 12)

The method of claim 12, wherein the inference algorithm is a variable elimination algorithm.

### Claim 14

A computer system for automated diagnostics of appliances, computer system comprising:
one or more computer processors;
one or more computer-readable storage media, and program instructions stored on the one or more computer-readable storage media, the program instructions comprising:
program instructions to collect, respectively, from a plurality of a type of network-connected appliances, a historical log of sensor data and events recorded over a timeline;
program instructions to generate a dynamic probabilistic graphical model of the type of network-connected appliances, which includes:
program instructions to map data of a sensor of a plurality of sensors monitoring operational activity of the type of network-connected appliances to a variable of a plurality of variables, respectively;
program instructions to map an event of the type of network-connected appliances, within the timeline, to a set of variables and corresponding sensor values associated with the timeline;
program instructions to apply a structure learning algorithm to the variables and timeline corresponding to the event;
program instructions to apply expert knowledge of an interaction between one or more subsystems and components of the network-connected appliances, respectively, as supervised machine learning techniques; and
responsive to receiving a diagnostic query regarding a diagnosis of the type of network-connected appliance, program instructions to generate a response based on the diagnosis.

### Claim 15 (depends on 14)

The computer system of claim 14, further comprising:
program instructions to provide a natural language processing (NLP) conversational user interface for diagnostic queries regarding the operational activity of the appliance.

### Claim 16 (depends on 14)

The computer system of claim 14, wherein the plurality of the type of appliances perform a same operational activity.

### Claim 17 (depends on 14)

The computer system of claim 14, wherein the event of the type of network-connected appliances, respectively, includes one or more selected from a group consisting of: malfunction of the appliance, a state of condition of the appliance, a condition of a component of the appliance, a condition of a subsystem of the appliance, and an estimated prediction of a malfunction at a future time.

### Claim 18 (depends on 14)

The computer system of claim 14, further comprising:
program instructions to train the dynamic probabilistic graphical model by applying machine learning techniques to the historical log of operational activity collected from the plurality of the type of network-connected appliances.

### Claim 19 (depends on 14)

The computer system of claim 14, wherein the program instructions to map the data of a sensor of the plurality of sensors monitoring the operational activity of the type of network-connected appliances to a variable of the plurality of variables, respectively, includes associating a time reference of the data of the sensor from the timeline.

### Claim 20 (depends on 14)

The computer system of claim 14, wherein the type of network-connected appliance includes a particular manufacturer of the appliance.
